The Expanded Universe of Prokaryotic Argonaute Proteins

Sergei Ryazansky,a Andrey Kulbachinskiy,a Alexei A. Aravina,b

aInstitute of Molecular Genetics, Russian Academy of Sciences, Moscow, RussiabDivision of Biology and Biological Engineering, California Institute of Technology, Pasadena, California, USA

ABSTRACT Members of the ancient family of Argonaute (Ago) proteins are presentin all domains of life. The common feature of Ago proteins is the ability to bindsmall nucleic acid guides and use them for sequence-specific recognition—andsometimes cleavage— of complementary targets. While eukaryotic Ago (eAgo) pro-teins are key players in RNA interference and related pathways, the properties andfunctions of these proteins in archaeal and bacterial species have just started toemerge. We undertook comprehensive exploration of prokaryotic Ago (pAgo) pro-teins in sequenced genomes and revealed their striking diversity in comparison witheAgos. Many pAgos contain divergent variants of the conserved domains involved ininteractions with nucleic acids, while having extra domains that are absent in eAgos,suggesting that they might have unusual specificities in the nucleic acid recognitionand cleavage. Many pAgos are associated with putative nucleases, helicases, andDNA binding proteins in the same gene or operon, suggesting that they are in-volved in target processing. The great variability of pAgos revealed by our analysisopens new ways for exploration of their functions in host cells and for their use aspotential tools in genome editing.

IMPORTANCE The eukaryotic Ago proteins and the RNA interference pathways theyare involved in are widely used as a powerful tool in research and as potential ther-apeutics. In contrast, the properties and functions of prokaryotic Ago (pAgo) pro-teins have remained poorly understood. Understanding the diversity and functionsof pAgos holds a huge potential for discovery of new cellular pathways and noveltools for genome manipulations. Only few pAgos have been characterized by struc-tural or biochemical approaches, while previous genomic studies discovered about300 proteins in archaeal and eubacterial genomes. Since that time the number ofbacterial strains with sequenced genomes has greatly expanded, and many previ-ously sequenced genomes have been revised. We undertook comprehensive analysisof pAgo proteins in sequenced genomes and almost tripled the number of knowngenes of this family. Our research thus forms a foundation for further experimentalcharacterization of pAgo functions that will be important for understanding of thebasic biology of these proteins and their adoption as a potential tool for genomeengineering in the future.

KEYWORDS Ago, RNA interference, genome editing, horizontal gene transfer,prokaryotic Argonaute proteins

Argonaute (Ago) proteins play the key role in the RNA interference (RNAi) pathwaysin eukaryotes. All known eukaryotic Agos (eAgos) bind small RNA molecules and

use them as guides for sequence-specific recognition of long RNA targets. Uponrecognition of the target, it can be cleaved by the intrinsic endonuclease activity of theAgo protein (1–4). Alternatively, Ago proteins, especially the members of the family thatlack nuclease activity, can recruit partner proteins to the target RNA, leading to itsdegradation and/or repression of its translation (5–7). Recognition of nascent RNA by

Received 3 September 2018 Accepted 20November 2018 Published 18 December2018

Citation Ryazansky S, Kulbachinskiy A, AravinAA. 2018. The expanded universe ofprokaryotic Argonaute proteins. mBio9:e01935-18. https://doi.org/10.1128/mBio.01935-18.

Editor Eleftherios T. Papoutsakis, University ofDelaware

Copyright © 2018 Ryazansky et al. This is anopen-access article distributed under the termsof the Creative Commons Attribution 4.0International license.

Address correspondence to Sergei Ryazansky,s.ryazansky@gmail.com, or Alexei A. Aravin,aaa@caltech.edu.

RESEARCH ARTICLEMolecular Biology and Physiology

crossm

November/December 2018 Volume 9 Issue 6 e01935-18 ® mbio.asm.org 1

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

some eAgos can also lead to modification of the chromatin structure—DNA andhistone methylation— of the target locus (8–10).

The proteins that belong to the Ago family are also present in the genomes of manybacterial and archaeal species (11, 12). Structural and biochemical studies of prokaryoticAgos (pAgos) have provided key insights into the mechanisms of RNAi in eukaryotesand revealed that Ago proteins directly bind short nucleic acid guides and can cleavecomplementary targets (13–21). The same studies showed that pAgos can associatewith short DNA guides and preferentially recognize DNA targets, in contrast to allknown eAgos (17, 20, 22–28). Despite these differences, solved structures of severalpAgos and eAgos combined with their sequence alignments revealed a conserveddomain organization of these proteins (reviewed in references 12 and 59). All eAgosand all pAgos except one that were experimentally characterized to date possess fourdomains that are organized in a bilobal structure, with N- and PAZ (PIWI-Argonaute-Zwille) domains forming one lobe and MID (Middle) and PIWI (P-element InducedWimpy Testis) domains forming another lobe (16, 18, 20, 29–34). The nucleic acids arebound between the lobes; the MID and PAZ domains interact with the 5= and 3= endsof the small nucleic acid guide, respectively, and the PIWI domain contains an RNaseH-like fold with a catalytic tetrad of conserved amino acid residues involved in thetarget cleavage (18, 19, 23–25, 27, 28). The N-domain is the least conserved in Agoproteins; it was proposed to facilitate RNA duplex unwinding during guide loading byeAgos (35) and prevent extended duplex formation during target recognition by somepAgos (23).

Until now only a few pAgos were characterized by structural or biochemicalapproaches (see Fig. 1B). At the same time, earlier genomic studies revealed that up to32% of the Archaea and 9% of the Eubacteria with sequenced genomes contain genesencoding proteins from the Ago superfamily and showed that the diversity of pAgos isfar greater than that of eAgos (11, 12). Indeed, many pAgos contain substitutions of keycatalytic residues in the PIWI domain (and are probably inactive), and a large class ofpAgos lack the N- and PAZ domains (“short pAgos”), while many pAgos containadditional domains absent in eAgos, either in the same protein or as a part of putativeoperons. Furthermore, previous studies of the genomic context of pAgo genes revealedthat catalytically inactive pAgos are usually associated with several types of nucleasesfrom the SIR2, TIR, PLD, Mrr (which include proteins with novel RecB-like and RecB2domains), or Cas4 family (11, 12). Mrr and Cas4 are related to a highly divergent familyof PD-(D/E)XK nucleases (36, 37). Furthermore, a subclass of pAgos was found inassociation with Cas1 and Cas2 nucleases within the CRISPR loci (24). The diversity ofpAgos revealed by genomic studies together with few examples of biochemicallycharacterized proteins suggested new cellular functions for these proteins (38–41).Beyond understanding functions of pAgos in their host cells, analysis of pAgos mayyield new proteins that can be potentially harnessed for biotechnology, in particular asan alternative to the CRISPR/Cas genome editing tools (42).

Previous genomic studies identified a nonredundant set of 261 pAgos (11, 12).However, since their discovery the number of bacterial strains with sequenced ge-nomes has more than tripled, and many previously sequenced genomes have beenrevised, resulting in the removal of some pAgo genes from the databases. Therefore, wehave undertaken comprehensive exploration of the diversity of pAgos using the latestversion of the RefSeq protein database for bioinformatic search. The results of thesearch almost tripled the number of identified pAgo genes. Our analysis supports thepreviously reported separation of pAgos into short pAgos and long pAgos (which bythemselves can be separated into two distinct clades) and reveals several subclasseswith significant differences in the domain organization. These include variations in theMID and PAZ domains involved in guide binding and substitutions of catalytic residuesin the PIWI domain. A large fraction of pAgos, including those that lack intrinsicendonuclease activity, are associated with putative nucleases and other DNA bindingproteins. Overall, the great diversity of pAgos revealed by our analysis opens a way for

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 2

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

exploration of their biochemical properties and cellular functions, as well as their use aspotential tools in genome engineering.

RESULTSThe expanded set of prokaryotic Argonaute proteins. To reveal the diversity of

pAgos and to get new insights into their evolution, we performed comprehensiveanalysis of proteins that belong to the Ago family in sequenced genomes of prokaryoticspecies. For this, we searched 116 million proteins fetched from the RefSeq NCBIprotein database using known pAgos as queries by PSI-BLAST. We found 1,010 pAgosencoded in 1,385 completely and partially sequenced genomes of 1,248 eubacterialand archaeal strains (Fig. 1; see also Table S1 in the supplemental material). In total,pAgos were found in 17% of eubacterial and 25% of archaeal genera (Fig. 1A). Themajority (1,186, or 95%) of archaeal and eubacterial strains carry only one pAgo gene(876 pAgos, or 86.7% of all proteins). However, the genomes of 57 strains carry twopAgo genes, four strains carry three pAgo genes, and one strain encodes four pAgos.pAgos are in general randomly distributed in different prokaryotic clades as the number

1/71/71/21/61/51/32/51/3

2/252/111/61/72/22/112/44/103/82/45/82/44/233/154/874/129/18

2/714/25

11/1024/6023/20119/14721/246

27/4617/55

42/9333/156

43/8760/343

88/325

ChlorobiaDeferribacteresRhodothermia

SaprospiriaThermoleophilia

ZetaproteobacteriaBalneolia

DehalococcoidiaMethanomicrobia

NitrososphaeriaNitrospiraOpitutae

RubrobacteriaAquificae

ArchaeoglobiAnaerolineae

VerrucomicrobiaeMethanococciAcidobacteriia

ThermococciThermoproteiSpirochaetia

DeltaproteobacteriaThermotogaeChitinophagia

DeinococciPlanctomycetia

SphingobacteriiaCytophagia

ClostridiaBacilli

ActinobacteriaHalobacteria

BacteroidiaFlavobacteriia

BetaproteobacteriaCyanobacteria

GammaproteobacteriaAlphaproteobacteria

long pAgo short pAgo

Found pAgo proteins

Cla

sses

0 100 200

20

FIG 1 Phylogenetic analysis of pAgo proteins. (A) The numbers of identified short and long pAgo proteins in the classes of Eubacteria and Archaea. Pinkasterisks near the clade names indicate the archaeal classes. The numbers on the right of each bar depict the number of genera encoding pAgos versus thetotal number of genera with the complete or partially sequenced genomes in the corresponding class. (B) The circular phylogenetic tree of the nonredundantset of pAgos constructed based on the multiple alignment of the MID-PIWI domains. The pAgo proteins were annotated as follows, from the inner to the outercircles: the type of protein, in which long-A pAgos are colored in green (truncated variants without the PAZ domain, light green), long-B pAgos are light green(truncated variants without PAZ, green), and short pAgos are orange; the type of the PIWI domain, depending on the presence of the catalytic tetrad DEDX;the type of the 5=-end guide binding motif in the MID domain (the first two conserved residues are indicated); the presence and the type of the PAZ/APAZdomains in pAgos; and the superkingdom to which the corresponding pAgo belongs. Biochemically characterized pAgos (or the most similar redundanthomologs for RsAgo and TtAgo) with resolved structures are highlighted in red. The scale bar represents the evolutionary rate calculated under the JTT�CATevolutionary model.

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 3

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

of genera in each class that encode pAgos in their genomes correlates with the totalnumber of genera in the class (Fig. 1A; Fig. S1). The nonredundant set of pAgos withthe level of similarity of less than 90% comprises 721 proteins, which is almost triple thenumber of previously identified pAgo proteins (11, 12).

As described below, the vast majority of pAgos have the MID and PIWI domains, butmany proteins lack the N- and PAZ domains conserved in eAgos. Therefore, we usedmultiple alignments of the MID-PIWI domains to construct a phylogenetic tree of pAgos(Fig. 1B; Fig. S2). The tree revealed three large clades of pAgos, while two proteins fromthermophilic archaea Thermoproteus uzoniensis and Vulcanisaeta moutnovskia could notbe unambiguously classified as belonging to any of these clades due to their highdivergence. Following Makarova and coauthors (11), we call one of these clades “shortpAgos” (Fig. 1B), as all proteins in this clade lack the N- and PAZ domains present ineAgos. The majority of pAgos that belong to the other two clades include the PAZdomain, and we therefore designate them as long-A and long-B. The diversity of pAgosvastly exceeds the diversity of eAgos that were previously shown to form a singlebranch on the phylogenetic tree of long pAgos (as illustrated in Fig. 1B) (11, 12).

The three clades of pAgos can be found in both archaeal and eubacterial species.Furthermore, the majority of prokaryotic classes that include large numbers of generaencode both short and long pAgos. However, some classes have a clear bias towardshort or long pAgo types; thus, genomes of Alphaproteobacteria predominantly encodeshort pAgos, while Cyanobacteria predominantly encode long pAgos (Fig. 1A). pAgosthat belong to different clades can be found in the same genome (for the strains thatencode several proteins). For example, Enhydrobacter aerosaccus ATCC 27094 encodesthree short and one long-A pAgo, Parvularcula bermudensis HTCC2503 and Halomi-cronema hongdechloris C2206 encode two short and one long-B pAgos, and Methylo-microbium agile ATCC 35068 contains one short and one long-B pAgo. Generally, thephylogenetic tree of pAgos has little similarity with the phylogenetic tree of hostspecies built using classic molecular markers such as rRNA genes, thus confirmingpreviously published observations (11, 12) and suggesting that pAgos have mostlyspread through horizontal gene transfer (HGT). Notably, a few pAgos that have beenbiochemically studied almost all belong to the long-A clade, while only two pAgos fromthe long-B clade and no proteins from the short pAgo clade have been characterizedto date (indicated in Fig. 1B).

Domain architecture of pAgos. To gain further insight into the structural diversityof pAgo, we analyzed their domain architecture. The exploration of multiple alignmentsof PIWI, MID, and PAZ domains was used to identify the main structural and functionalfeatures of pAgos. We did not include the N-domain in our analysis because it has thelowest level of conservation, which impedes its identification in diverse pAgos. UsingInterProScan and CDD-batch programs and the Pfam and Superfamily databases, wealso searched for other domains in pAgo proteins. This analysis allowed us to proposea comprehensive classification of the pAgo domain organization (Fig. 2).

As justified by their name, we found that all members of the short pAgo clade lackthe PAZ domain implicated in binding of the 3= end of a guide molecule. The majorityof short pAgos (470, or 81%) have only two domains, MID and PIWI. Six short pAgos(�1%) are further truncated and lack the MID domain, suggesting that they may beunable to interact with a guide. In addition to the MID and PIWI domains, 106 shortpAgos (18%) that all belong to a single monophyletic branch encompass the APAZ(“analog of PAZ”) domain (Fig. 1B and 2A). The APAZ domain has no sequence similarityto PAZ and was previously described as a domain always associated with short pAgoseither as a part of the same protein or present in the same operon with short pAgos(11). Almost all (102 out of 106) APAZ-containing short pAgos also carry domainsrelated to the SIR2 family (SIR2_1 subfamily) (Fig. 1B and 2A), which are considered tobe bona fide nucleases (11, 12).

In contrast to short pAgos, the majority (378, or 90%) of pAgos that belong to thelong pAgo clades have the domain architecture similar to eAgos, including PAZ, MID,

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 4

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

and PIWI domains (the N-domain was not included in the analysis). However, 31 pAgosthat undoubtedly belong to long pAgos based on alignments of their MID-PIWIdomains lack the PAZ domain, including previously characterized AfAgo from Archaeo-globus fulgidus (13–15). These truncated long pAgos are scattered at different positionson the phylogenetic tree, suggesting multiple independent cases of the loss of thisdomain during evolution of long pAgos. One of these truncated proteins also lacks theMID domain. On the other hand, 15 long pAgos that have the PAZ domain also carryadditional putative nucleic acid binding domains at their N termini: 7 pAgos containDNA-binding domains (Schlafen, PF04326, or lambda-repressor-like, SSF47413) and8 pAgos contain a nucleic acid binding domain with the OB-fold (SSF50249) (Fig. 1Band 2B).

Analysis of the diversity of MID, PIWI, and PAZ domains revealed significant varia-tions in their structures characteristic for different groups of pAgos (Fig. 1B and 2), asdescribed below. In particular, subsets of proteins from both clades of long pAgoscontain MID* domains with substitutions of key residues involved in interactions withthe guide 5= end; many long-A pAgos and all long-B and short pAgos contain PIWI*variants with substitutions of essential catalytic residues; and most long-B pAgoscontain incomplete variants of the PAZ* domain involved in the 3=-guide interactions(Fig. 2).

Binding of nucleic acid guides in the MID domain. The vast majority (1,001, or99% of all proteins) of pAgos contain the MID domain that has been implicated inbinding of the 5= end of guide molecules. The resolved 3D structures of pAgos,including AfAgo (pAgo of A. fulgidus) (13–15, 22), RsAgo (Rhodobacter sphaeroides) (25,28), TtAgo (Thermus thermophilus) (17, 19, 23), and MjAgo (Methanocaldococcus jann-aschii) (27), have shown that the MID domain anchors the phosphorylated 5=-guidenucleotide by a set of conserved amino acid residues (Fig. 3). These interactions arecomplemented by the PIWI domain that provides additional residues for 5=-guide anddivalent cation binding (13, 15, 19, 23, 25, 27, 28).

247

218

102

5

3

1

(HK)-MID

(HK)-MID

FIG 2 The diversity of domain architecture of pAgo proteins. Various types of domain organization of short (A) and long (B) pAgo proteins are schematicallyshown. The numbers on the right of each type are the frequency of its occurrence in the redundant set of proteins. Based on the presence of the DEDX tetradin the PIWI domain and the putative 5=-P guide binding motif in the MID domain, all long pAgos can be separated into four types shown here. Green PIWIdomains carry the active DEDX catalytic tetrad while turquoise PIWI* domains do not contain the canonical tetrad. Yellow MID domains contain different typesof the 5=-end guide binding motif as indicated, while light-yellow MID* domains carry substitutions of the crucial amino acid residues in this motif. Orange PAZdomains have the full-sized pocket responsible for 3=-end guide binding, while light-brown PAZ* variants do not have the second subdomain. The OB-fold isthe nucleic acid-binding domain SSF50249; the DNA-binding domains are Schlafen domain with AlbA_2 (PF04326) or lambda-repressor-like domain (SSF47413).

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 5

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

Four conserved residues of the MID domain (Y/R, K, Q, and K in crystallized pAgos,e.g., Y463, K467, Q478, and K506 in RsAgo in Fig. 3C) and a bound divalent cation (Mg2�

or Mn2�) form a network of hydrogen bonds anchoring the 5=-phosphate, and also thethird phosphate of the guide molecule, into this basic pocket (13, 15, 18, 25, 27).Multiple sequence alignment of the MID domains shows that these residues are highlyconserved in most pAgos (Fig. 3A and B). Accordingly, substitutions of these residueswere shown to disrupt pAgo-guide interactions (13, 14, 18, 25, 27). Two additionalsemiconserved residues (e.g., T and N in RsAgo) (Fig. 3) contribute to binding of thephosphate group and the base of the nucleotide in the second position of the guide,respectively (Fig. 3C). Together, these residues constitute a conserved six-amino-acidmotif that is found in most pAgos (YKQTNK consensus for long pAgos, with the mostconserved residues shown in bold; indicated with red asterisks in Fig. 3A and B).

Interestingly, we found amino acid variations at the first two positions of this motif,which might be important for the function of different groups of pAgos. The first two

* * * * * *

** * * *

439458120364192743169

42168172413206188

50176174373

71379377377378378380377

484507164411251790216100215220458257239100227228418116424422422423423425422

1 2 3 4 5 6 7 8 9 10

02

4

position

bits

bits

bits

bits

1 2 3 4 5 6 7 8 9 10

02

4

1 2 3 4 5 6 7 8 9 10

01

23

1 2 3 4 5 6 7 8 9 10

02

4

long pAgos

short pAgos

long pAgos with MID-OH

pAgos with MID*

BA* * * * **

* * * * **

* * * **

C

YK-type

HK-type

MID-OH

RK-type

1 2 3

4 5

6

7 8

9

3 10

FIG 3 The structure of the 5=-end guide binding motifs in the MID domain. (A) A fragment of the multiple alignment of MID domains of several typical pAgoproteins with different types of the 5=-end guide binding motifs. Red asterisks are the positions of amino acid residues involved in the binding of the 5=-P-guideend. Black asterisks are the positions of amino acid residues involved in the binding of the 5=-OH end of a guide in the MID-OH pAgos. MID domains with theYK-type 5=-P-end binding motif belong to the long pAgos; MID domains of the HK type belong to the short pAgos; MID domains of the RK type are characteristicof short pAgos but can also be found in some long pAgos. The multiple alignments were manually edited to bring them in conformity to the structures shownin panel C. The species names, NCBI taxon_id, the types of pAgos (long, short, and long truncated without the PAZ domain), and corresponding accessionnumbers are shown at the left of the alignment. (B) The sequence logo of amino acid residues involved in the formation of the 5=-end binding pocket for longpAgos, short pAgos, MpAgo-like pAgos with MID-OH, and long pAgos with MID* domain variants. The combined logo motif including positions characteristicfor canonical MID and MID-OH domains is shown. Red and black asterisks are amino acid positions specific for the 5=-P (MID) and 5=-OH (MID-OH) binding motifs,respectively. (C) Structures of binary complexes of RsAgo, TtAgo, and MpAgo with guide molecules: 5=-P guide RNA for RsAgo, 5=-P guide DNA for TtAgo, or5=-OH guide RNA for MpAgo. For RsAgo and TtAgo, only residues from the conserved YKQTNK motif are indicated; for MpAgo, MID-OH-specific residuesinvolved in interactions with the 5=-guide end are also shown (labeled with asterisks). PDB IDs are shown next to each structure.

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 6

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

residues in the motif are usually YK in long pAgos; subsets of long-A pAgos contain KR,RK, HK, or YY combinations, while long-B pAgos are almost exclusively YK variants. Thearomatic ring of the first Y stacks on the first base at the 5= end of a guide, therebycontributing to the binding of the first nucleotide within the MID pocket (13, 15, 25, 27,28). In contrast to tyrosine, R is positively charged and not aromatic and so unable tostack the first base of the guide, as exemplified in the structure of TtAgo (Fig. 3C) (18,23); the structure of any pAgo with histidine at the first position of the motif remainsunknown.

The conserved residues implicated in binding of the 5=-phosphorylated guide endare present in the majority (80%) of long pAgos (Fig. 1B). However, 84 (20%) of longpAgos have at least one amino acid substitution at the most conserved positions in the5= binding motif in their MID* domain; namely, they do not contain Y/H/R at the firstposition of the YKQTNK motif and/or K/R/Y at the second position, and/or Q at the thirdposition, and/or K at the last position. A separate subgroup of these proteins (MID-OH),which form a tight clade on the phylogenetic tree (Fig. 1B), includes pAgo fromMarinitoga piezophila (MpAgo) that was recently shown to bind guide molecules with5=-OH instead of 5=-P ends (24). The multiple sequence alignment showed that thisgroup includes 9 pAgos (compared to 3 proteins identified in the previous study [24])and that positions of amino acid residues involved in 5=-guide interactions in theseproteins overlap the positions of the 5=-P-end binding motif in other pAgos (Fig. 3A andB). However, in comparison with most pAgos, these residues form a more hydrophobicpocket for 5=-OH binding (Fig. 3C). In addition, the MID pocket of MpAgo does not bindMg2� ions (Fig. 3C) (24). The absence of guide contacts with the Mg2� ion is in partcompensated by interactions between the phosphates of second and third nucleotidesof guide RNA with lysine residues specific for this group of pAgos (K403 and K418 inMpAgo; Fig. 3B and C). While the first tyrosine of the YKQTNK motif that stacks withthe first nucleotide base in the guide molecule in other pAgos is replaced with ahydrophobic residue in the MID-OH domain, its role is taken by a precedingaromatic residue (Y or F) that stacks from the other side of the same base (Y379 inMpAgo; Fig. 3B and C).

Another 75 pAgos with noncanonical MID* variants, none of which is characterizedto date, belong to several distinct branches on the phylogenetic tree of pAgos, andsome of them are located close to the MID-OH pAgos (Fig. 1B). The alignment of thesesequences showed that in general they share the same YKQTNK motif, although theconservation of individual positions is much lower than for canonical pAgos (see logoin Fig. 3B; alignment in Data Set S1). It remains to be established whether these proteinsinclude any additional subgroups with noncanonical MID* domains of different spec-ificities, as already revealed for MpAgo, or whether they just represent divergentvariants of the same canonical motif.

In contrast to long pAgos, all short pAgos have the canonical 5=-P-end binding motif(Fig. 3B). Interestingly, almost all of them contain HK and RK residues at the first twopositions of this motif, instead of YK in long pAgos, further supporting their separationinto a distinct clade from long pAgos (Fig. 1, 2B, and 3A and B). Though none of shortpAgos was biochemically characterized to date, the conservation of key residues in theMID domain strongly suggests that these proteins bind 5=-phosphorylated guidenucleic acids similarly to most long pAgos and eAgos.

Endonuclease activity of the PIWI domain. Structural and biochemical studies ofseveral pAgos and eAgos demonstrated that the conserved tetrad of amino acidresidues in the PIWI domain (DEDX, where X is D, H, or K) is responsible for endonu-cleolytic cleavage of the target upon its recognition by the Ago-guide complex (18–20,24, 40). Similarly to previous reports (11, 12), we found that all pAgos that belong to theshort clade lack a canonical DEDX catalytic tetrad in the PIWI domain (PIWI* variants)(Fig. 2A). Similarly to short pAgos, all proteins in the long-B clade also lack the catalytictetrad. In contrast, the majority of long-A pAgos (79%) have the canonical DEDXcatalytic tetrad in the PIWI domain, suggesting that they possess endonucleolytic

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 7

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

activity. However, this branch also harbors pAgos with substitutions in the catalytictetrad, suggesting several independent events of the loss of endonuclease activity inthis branch.

Binding of the guide 3= end in the PAZ domain. It was previously shown that theamino acid sequences of the PAZ domains of the pAgo proteins are divergent, but theirstructural folds are similar. The full-length PAZ domain has two subdomains, eachconsisting of two nucleic acid binding regions, which are oriented to form a hydro-phobic pocket that anchors the 3= end of the guide strand in the binary pAgo-guidecomplex (16, 18, 20, 28); a typical structure of PAZ is illustrated for TtAgo in Fig. 4A. Thefirst subdomain consists of an oligosaccharide binding fold (OB-fold)-like structure withone or two helices on one side and includes nucleic acid binding regions R1 and R4. Thesecond subdomain consists of an �-helix and a �-hairpin or loop structure, sometimesfollowed by another �-helix, and includes nucleic acid binding regions R2 and R3(Fig. 4A) (16, 18, 20). In TtAgo, all four regions contribute to anchoring of the guide 3=end in the PAZ pocket (Fig. 4A) (18). However, some pAgos, such as RsAgo and MpAgo,contain incomplete variants of the PAZ domain that lack the second subdomain,including regions R2 and R3, and therefore do not possess the PAZ pocket (24, 25, 28).Nevertheless, structural analysis of the binary guide-MpAgo complex revealed that theguide 3= end can still be bound in the incomplete PAZ domain, although with adifferent orientation relative to TtAgo (24). These differences may possibly affect thekinetics of guide binding, the stability of binary guide pAgo complexes, or their abilityto recognize complementary nucleic acid targets (see Discussion).

We determined how often the PAZ domain without the second subdomain (referredto here as PAZ*) is present among long pAgos. As noted above, the structural elementsof PAZ involved in nucleic acid binding are organized by the four poorly conservedregions, R1 to R4 (Fig. S3, as described in reference 28), with the DNA-binding �-helixin the second subdomain formed by stretches of 7 to 15 amino acids from region R3(illustrated for TtAgo and Pyrococcus furiosus PfAgo in Fig. 4). Thus, we analyzed thepresence of this region as a signature of the second subdomain of PAZ in pAgos. Usingmultiple sequence alignments of the identified PAZ domains, we found that R3 is found

FIG 4 PAZ domains in pAgo proteins. (A) The structure of the 3=-guide binding pocket in TtAgo (PDBID: 3DLH). Structural regions R1 to R4 forming the pocket are shown in surface representation. (B) Ageneral scheme of the PAZ and PAZ* domain structures in long pAgos. All PAZ domains can have R1, R2,R3, and R4 regions. PAZ* is defined as variants of PAZ lacking full-length region R3. No regions R1, R2,and/or R4 could be detected in multiple alignments of some PAZ or PAZ* variants (shown by brokenboxes), but their actual presence needed to be tested due to their low conservation. The numbers showhow many long pAgos have corresponding domain structures. The pAgos with resolved structureshaving PAZ and PAZ* domains are indicated. (C) The three-dimensional structures of PAZ (TtAgo andPfAgo) and PAZ* (MpAgo and RsAgo) variants in crystallized pAgos and their binary complexes with DNA(for TtAgo) or RNA (for MpAgo) guide molecules. Tyrosine and lysine/arginine residues from R3 in TtAgoand PfAgo, probably involved in guide binding, are indicated. PDB IDs are shown next to each structure.

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 8

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

in more than half of the PAZ domains in long pAgos, including most long-A pAgos(Fig. 1B and 4A; alignment in Data Set S2), indicating their canonical two-lobe structure.The rest of PAZ* domains are smaller and do not have R3, suggesting that they havethe one-subdomain organization without the nucleic acid binding pocket, as previouslyobserved for RsAgo and MpAgo. Interestingly, the PAZ* domain is found in almost alllong-B pAgos (Fig. 1B), suggesting that it was present in the common ancestor of thisgroup.

Phylogenetic analysis of APAZ domain-containing proteins. Previous analysesidentified the so-called APAZ domain that is present in some short pAgos as well as ingenes contained in putative operons containing short pAgos (11). The APAZ domainwas proposed to be the functional analog (but not homolog) of the PAZ domain (11)or the N-domain (43) that is absent in short pAgos. To gain insight into the diversity andphylogeny of APAZ-containing proteins, we iteratively searched the RefSeq proteindatabase by DELTA-BLAST using the sequences of known APAZ domains as queries. Intotal, we found 5,385 protein hits. However, the vast majority (4,753, or 88%) of thefound proteins belonged to the HisG (PRK01686) and EIIB families (TIGR01996). Wetherefore performed the reverse DELTA-BLAST search using them as queries to test ifthese protein families were indeed related to the other APAZ proteins; however, we didnot find any APAZ-containing proteins within obtained hits. Moreover, the hisG pro-teins were found only among BLAST hits of the SIR2-APAZ protein from Sphingomonaswittichii RW1, while EIIB proteins were found only among BLAST hits of the TIR-APAZprotein from Chlorobium phaeobacteroides BS1. Thus, we concluded that HisG and EIIBfamilies were revealed artificially and are not related to the APAZ proteins, and weexcluded them all from the further phylogenetic analysis. The remaining set of APAZdomains included 632 proteins.

The multiple alignments of APAZ domains were used to build their phylogenetictree, which revealed that the APAZ proteins can be separated into four large groupsdesignated Ia, Ib, IIa, and IIb and a fifth group III of 18 proteins from several archaealspecies that have remote similarity to groups I and II (Fig. 5; alignment in Data Set S3).All five groups are present in putative operons with pAgos. We further analyzed thedomain architectures of APAZ-containing proteins using InterProScan and the Pfamand Superfamily databases. The first group of APAZ proteins designated Ia includes 117proteins that in addition to APAZ contain the SIR2 domain (the “SIR2-APAZ” type). Themajority of proteins in this group are short pAgos that also contain the MID-PIWIdomains; however, 31 proteins from this group lack the MID-PIWI domains. Since thesevariants are scattered along different branches in this group, they might have inde-pendently lost their MID-PIWI parts. In contrast, while a significant part of APAZproteins from the four other groups are associated with pAgos, only one of them ispAgo itself. Most proteins that belong to group Ib (150) are related to group Ia and alsocontain SIR2 besides APAZ, but do not have MID and PIWI. There are also severalproteins in both groups Ia and Ib that lack SIR2 or any other domains (the “APAZ” type;4 and 6 proteins for Ia and Ib, correspondingly). Most proteins of group IIa (122) arecharacterized by the presence of the TIR domain on their N termini (“the TIR-APAZ”type); one of them is short pAgo (also Fig. 2); 16 proteins in this group lack TIR domains.Group IIb, which encompasses 225 proteins, is diverse: 112 proteins contain unchar-acterized domain DUF4365 (“DUF4365-APAZ”), and 13 have the SIR2 domain, while 100do not have any extra domains (“APAZ”). Our analysis of the DUF4365 domain hasshown that it corresponds to the recently identified RecB-like domain of the Mrrsubfamily of PD-(D/E)XK nucleases (see below) (11). Finally, the proteins from group III,similarly to a major part of proteins from group IIb, do not have any other domainsexcept APAZ (the “APAZ” type).

APAZ-containing proteins are found almost exclusively in eubacterial species. Theanalysis revealed tight reciprocal association between APAZ-containing proteins andshort pAgos. Indeed, although only one branch of short pAgos contains the APAZdomain, the majority of other short pAgos that do not have APAZ themselves contain

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 9

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

APAZ-containing genes in their operons (Fig. 1B and 2). Reciprocally, 397, or 63%, of allgenes with the APAZ domain are positioned close (within 10 genes on the samegenomic strand) to short pAgos, likely in the same operons (Fig. 5) (see below). Thus,APAZ proteins are likely an integral component of functional pathways mediated byshort pAgos.

WP

_082

8220

67.1

|439

92|B

rady

rhiz

obiu

m li

aoni

ngen

se

WP_093992232.1|1819565|Loktanella m

arina

WP

_081

8527

74.1

|112

5973

|Bos

ea s

p. 1

17

WP_091963848.1|1761772|Bradyrhizobium sp. err11

WP

_093

0391

41.1

|128

7727

|Ros

eova

rius

azor

ensi

s

WP

_090

0726

29.1

|655

353|

Coh

aesi

bact

er m

aris

flavi

WP_062680196.1|222|Achrom

obacter

WP

_014

1327

57.1

|108

2930

|Pel

agib

acte

rium

WP_

0975

4471

4.1|

1184

720|

Rhi

zobi

um a

nhui

ense

WP_018016394.1|2426|Teredinibacter turnerae

WP_081730962.1|68287|Mesorhizobium

WP

_097

0827

96.1

|570

013|

Rho

doba

cter

sp.

JA

431

WP_0

2919

3119

.1|5

9839

|Pae

niba

cillu

s al

gino

lyticu

s

WP_053571900.1|1353886|Caballeronia cordobensis

WP_0

8729

2483

.1|1

9655

88|A

naer

ofilu

m s

p. A

n201

WP_007215639.1|246787|B

acteroides c

ellulosil

yticu

s

WP_087738802.1|1701573|Paraburkholderia piptadeniae

WP

_006

7462

22.1

|108

010|

Thio

alka

livib

rio p

arad

oxus

WP

_056186121.1|1736259|Methylobacterium

sp. Leaf113

WP_038943875.1|375|Bradyrhizobium

japonicum

WP_

0788

1466

9.1|

4846

7|Pr

osth

ecob

acte

r deb

ontii

WP_039858603.1|205844|Novosphingobium pentaromativorans

WP_093081846.1|1884369|Sphingobium sp. AP50

WP_

0756

3405

0.1|

1672

749|

Rhi

zobi

um rh

izos

phae

rae

WP_

1002

3990

5.1|

1983

755|

Sphi

ngob

ium

sp.

LB1

26

WP

_051044811.1|91459|Methylobacterium

sp. B1

WP_036051529.1|999627|Sedimentitalea nanhaiensis WP_089318528.1|1610492|Pontibacter ummariensis

WP

_078056299.1|1234595|Pacificim

onas flava

WP_006685036.1|133448|Citrobacter youngae

WP_038205789.1|151755|Xenophilus azovorans

WP_008781127.1|469592|Bacteroides sp. 3_1_19

WP

_067969880.1|83263|Am

inobacter aminovorans

WP

_094291737.1|2004485|Acidovorax sp. K

ND

SW

-TSA

6

NP

_951611.1|243231|Geobacter sulfurreducens

WP_061958301.1|82633|Cupriavidus pauculus

WP_018273545.1|29448|Bradyrhizobium elkanii

WP

_050

0385

98.1

|223

7|H

aloa

rcul

a

WP_0

7309

9040

.1|98

8|Lee

uwen

hoek

iella

marino

flava

WP_096483475.1|223967|Methylobacterium populi

WP_096462202.1|1675686|Sulfurifustis variabilis

WP

_082

7381

20.1

|136

87|S

phin

gom

onas

WP_

0917

8116

4.1|

1761

778|

Burk

hold

eria

sp.

yr2

81

WP_065998852.1|1886027|Mucilaginibacter sp. PPCGB 2223

WP_063002581.1|508464|Nocardia mikamii

WP_075285887.1|339|Xanthomonas campestris

WP

_082731713.1|13687|Sphingom

onas

WP_037187103.1|1500259|Rhizobium

sp. YR519

WP

_087

2528

36.1

|196

5647

|Dra

ncou

rtel

la s

p. A

n57

WP

_071

1456

52.1

|185

2362

|Bac

tero

ides

ihua

e

WP_

0676

9636

7.1|

4996

56|E

ryth

roba

cter

sp.

AP2

3

WP_079605750.1|1437360|Bradyrhizobium erythrophlei

WP_

0507

8387

7.1|

5164

66|B

urkh

olde

ria s

p. H

160

WP

_056152149.1|1736266|Duganella sp. Leaf126

WP_003807800.1|518|Bordetella bronchiseptica

WP

_013632722.1|151895|Pseudopedobacter saltans

WP_092021235.1|488535|Marinobacter zhejiangensis

WP_037435767.1|380|Sinorhizobium

fredii

WP_037716570.1|68239|Streptomyce

s mira

bilis

WP_013299755.1|208216|Parvularcula bermudensis

WP_062419304.1|229921|Levilinea sa

ccharolyt

ica

WP_035838021.1|765699|Defluviim

onas indica

WP

_059

6229

37.1

|605

52|B

urkh

olde

ria v

ietn

amie

nsis

siral

ucis

evsa

nomi

dnuv

erB|

6721

4|1.

1525

7348

0_P

W

WP

_026932797.1|279359|Gram

ella echinicola

WP

_092684840.1|1513892|Rhodopseudom

onas pseudopalustris

WP

_098

0634

74.1

|146

9170

|Lon

gim

onas

hal

ophi

la

WP_028849834.1|37926|Thermocrispum municipale

WP_052117900.1|205844|Novosphingobium pentaromativorans

WP_062785679.1|13688|Novosphingobium

capsulatum

WP_050399989.1|630921|Bradyrhizobium embrapense

WP

_052

3032

32.1

|605

48|P

arab

urkh

olde

ria g

ram

inis

WP_0

5603

4460

.1|17

3636

6|Chr

yseo

bacte

rium sp

. Lea

f404

WP_004271609.1|28077|Nitrospirillum amazonense

WP

_052114702.1|172045|Elizabethkingia m

iricola

WP_092996169.1|1566262|Rhizobium sp. NFR07

WP

_081215336.1|13690|Sphingobium

yanoikuyae

WP_029005631.1|281091|Azorhizobium doebereinerae

WP

_089

6991

20.1

|660

521|

Hal

ogra

num

gel

atin

ilytic

um

WP_050771919.1|404236|Maritimibacter alkaliphilus

WP

_059401471.1|1603291|Alicycliphilus sp. B

1

WP_0

2815

3625

.1|3

75|B

rady

rhizo

bium

japo

nicu

m

WP_026042451.1|1225184|Pantoea sp. A4

WP_056209640.1|1736485|Flavobacterium sp. Root186

WP_100788469.1|817|Bacteroides fragilis

WP_0

4756

4307

.1|1

5744

09|M

esor

hizo

bium

sp.

F7

WP

_082442752.1|311410|Labrenzia alba

WP

_083

7167

55.1

|125

0539

|Pel

agib

aca

abys

si

WP

_090

2706

87.1

|657

014|

Lito

rimic

robi

um ta

eane

nse

WP_064816696.1|1703960|Rhizobium sp. N113

WP_093164034.1|1882828|Variovorax sp. YR216

00B082

CJSL.ps

muibozihroseM|6337821|1.660086320_

PW

WP

_073035593.1|337701|Roseovarius pacificus

WP_082576376.1|1736568|Lysobacter sp. R

oot604

WP_054211403.1|1526658|Bosea vaviloviae

WP_089334214.1|1411120|Hymenobacter mucosus

WP_016393047.1|70775|Pseudomonas plecoglossicida

WP_

1000

8505

1.1|

1524

80|B

urkh

olde

ria a

mbi

faria

WP_068222132.1|333140|Roseivirga spongicola

WP

_044

0159

94.1

|153

9298

|Tre

pone

ma

sp. O

MZ

838

WP_092637302.1|1884352|Rhizobiales bacterium GAS113

WP_056160041.1|1839|Nocardioides

WP

_033317603.1|279369|Sphingopyxis baekryungensis

WP_045057112.1|1827278|Aliterella atlantica

WP_027412346.1|279356|Aquimarin

a muelle

ri

WP

_008140723.1|387090|Bacteroides coprophilus

WP_008615787.1|453852|Joostella marina

WP

_081

6247

27.1

|330

35|B

laut

ia p

rodu

cta

WP_093590751.1|235987|Streptomyces jietaisiensis

WP

_085937812.1|225324|Enhydrobacter aerosaccus

WP_036017302.1|102129|Leptolyn

gbya sp

. PCC 7375

WP_081903350.1|58117|Microbispora rosea

WP_068185907.1|1857568|M

acellib

acteroides s

p. HH-Z

S

WP_072017557.1|371042|Erwinia typographi

WP_0

0256

6035

.1|2

0847

9|[C

lost

ridiu

m] b

olte

ae

WP_0

8209

6051

.1|1

6035

55|C

andid

atus

Nitr

osot

enuis

cloa

cae

WP

_074129929.1|1680160|Bradyrhizobium

sp. NA

S96.2

WP

_026782276.1|257440|Pleom

orphomonas koreensis

NP_952413.1|243231|Geobacter sulfurreducens

WP_092119516.1|1855396|Porphyromonadaceae bacterium KH3CP3RA

WP_082722381.1|1385591|Burkholderia sp. BDU6

WP_031452536.1|1232437|Desu

lfobacu

la sp. T

S

WP_083808461.1|940615|Granulicella tundricola

esne

parb

memu

iboz

ihry

dar

B|12

9036

|1.1

8919

9380

_P

W

WP_0

2716

0975

.1|1

0409

84|M

esor

hizob

ium sp

. WSM

1293

WP_024810211.1|1305735|Oceanicola sp. HL-35

WP_009507644.1|794903|Opitutaceae bacterium TAV5

WP

_011

0210

17.1

|221

4|M

etha

nosa

rcin

a ac

etiv

oran

s

WP_036212047.1|1415754|Marinobacter sp. MCTG268

WP_044447509.1|589873|Alteromonas australica

WP

_025

4931

97.1

|123

2455

|Lac

hnos

pira

ceae

bac

teriu

m V

E20

2-12

WP_042565353.1|1587524|Flavo

bacteriu

m sp. M

EB061

WP_097031611.1|439529|Rhodobacter ovatus

WP_010424308.1|177400|Anaerophaga thermohalophila

WP_077684787.1|1332264|Tessaracoccus sp. NSG39

WP_084600957.1|80878|Acidovorax tem

perans

WP_080811272.1|1209493|Halom

icronema hongdechloris

WP_082469977.1|1736242|M

ethylobacte

rium sp

. Leaf86

WP_091133619.1|549386|Microvirga guangxiensis

WP_077230992.1|29448|Bradyrhizobium

elkanii

WP_077293451.1|187304|Labrenzia aggregata

WP_020809957.1|357|Agrobacterium

WP_071240727.1|1882733|Thalassospira sp. MIT1004

WP_083942961.1|266127|Sphingom

onas soli

WP_080801769.1|1183401|Agrobacterium genomosp. 1

WP_101603479.1|817|Bacteroides fragilis

WP_068818385.1|268141|Phormidesmis priestleyi

WP_051821414.1|1463855|Streptomyces sp. N

RRL F-5065

WP_0

8547

4560

.1|5

6106

1|Sph

ingob

acte

rium

psy

chro

aqua

ticum

WP

_023727136.1|1287285|Mesorhizobium

sp. LSH

C412B

00

WP

_052

0271

61.1

|118

7851

|Rho

dovu

lum

sp.

PH

10

WP_050631536.1|1654716|Bradyrhizobium viridifuturi

WP_032541222.1|816|Bacteroides

WP_011831497.1|105560|Methylibium petroleiphilum

WP_064102550.1|1805633|Acinetobacter sp. SFA

WP

_041864518.1|40324|Stenotrophom

onas maltophilia

WP_012486312.1|155077|Cellvibrio japonicus

WP

_029134085.1|191960|Sedim

enticola selenatireducens

WP_005783348.1|47883|Pseudomonas synxantha

WP_017505568.1|1229204|alpha proteobacterium L41A

WP_

0854

8289

2.1|

1515

439|

Para

burk

hold

eria

sus

onge

nsis

WP_090227846.1|1007099|Pseudomonas kuykendallii

WP_024973836.1|329|Ralstonia pickettii

WP

_019

0156

71.1

|457

934|

Elio

raea

tepi

diph

ila

WP_016278428.1|671267|Bacteroides sartorii

WP

_014020246.1|104|Cyclobacterium

marinum

WP_048755578.1|1035|Afipia felis

WP_008983418.1|553384|Alishewanella agri

WP_062365449.1|34073|Variovorax paradoxus

WP_100221482.1|80842|Herbaspirillum rubrisubalbicans

WP

_080826259.1|1183411|Agrobacterium

genomosp. 6

WP

_093

2808

27.1

|489

703|

Sol

imon

as a

quat

ica

WP_091979223.1|1502776|Methylobacterium sp. 13MFTsu3.1M2

WP

_013

6370

54.1

|861

208|

Agr

obac

teriu

m s

p. H

13-3

WP

_028

0811

00.1

|413

479|

Sol

imon

as s

oli

WP_082795034.1|33051|Sphingomonas sanguinis

WP

_044394680.1|80878|Acidovorax tem

perans

WP_011767948.1|189918|Mycobacterium sp. KMS

WP

_064683281.1|1138189|Rhizobium

bangladeshense

WP_085933358.1|225324|Enhydrobacter aerosaccus

WP

_046846704.1|113574|Hyphom

icrobium sp. G

J21

WP_049803085.1|375|Bradyrhizobium japonicum

WP

_081

2956

19.1

|384

|Rhi

zobi

um le

gum

inos

arum

WP

_096

3917

61.1

|104

8396

|Hal

open

itus

pers

icus

WP

_027

1015

16.1

|129

8913

|Com

amon

adac

eae

bact

eriu

m U

RH

A00

28

WP

_082657619.1|165697|Sphingopyxis

WP_052530012.1|36809|Mycobacterium abscessus

WP

_068847673.1|1841610|Planctopirus sp. JC

280

WP_1

0099

1967

.1|2

0570

25|S

piros

oma

sp. H

A7

WP_015056745.1|651143|Fibrella aestuarina

WP_083473000.1|269536|Frankia sp. R43

WP

_062

4869

46.1

|171

5989

|Can

dida

tus

Nitr

ospi

ra in

opin

ata

WP

_012

4894

04.1

|379

|Rhi

zobi

um

WP_035240078.1|1380348|Acidobacteriaceae bacterium URHE0068

WP_036267940.1|73780|Methylocaldum szegedienseWP_013685229.1|191579|Fluviicola taffensis

WP_092821445.1|1380357|Rhodospirillales bacterium

URHD0017

WP

_072559535.1|1913578|Sphingopyxis sp. LP

B0140

WP_081684175.1|266854|Granulicoccus phenolivorans

WP_004264290.1|587|Providencia rettgeri

WP_045167303.1|28131|Prevotella in

termedia

WP

_048

0768

61.1

|117

3487

|Hal

orub

rum

sp.

AJ6

7

WP_028337809.1|29448|Bradyrhizobium elkanii

WP_013654325.1|991904|Polymorphum gilvum

WP_093153036.1|420996|Thalassobaculum litoreum

WP_088492884.1|837|Porphyromonas gingivalis

WP_008657175.1|816|B

acteroides

WP

_083

5097

81.1

|121

290|

Hyp

hom

icro

bium

sul

foni

vora

ns

WP

_064304386.1|303|Pseudom

onas putida

WP

_092459739.1|1393122|Crenotalea therm

ophila

WP_0

7306

7332

.1|11

9409

0|Aliif

odini

bius r

oseu

s

WP

_037165528.1|1500300|Rhizobium

sp. CF

258

WP

_055143318.1|39492|[Eubacterium

] siraeum

WP

_097

2007

09.1

|188

4383

|Var

iovo

rax

sp. Y

R75

2

WP_003280119.1|316|Pseudomonas stutzeri

WP_053381025.1|42253|Nitrospira moscoviensis

WP_

0757

9371

1.1|

1141

883|

Mas

silia

put

ida

WP_068094658.1|76978|Novosphingobium rosa

WP

_066

8029

57.1

|406

81|S

phin

gom

onas

asa

ccha

roly

tica

WP_081000780.1|87883|Burkholderia multivorans

WP_048500689.1|232216|Chryseobacterium koreense

WP_084218942.1|328301|Syntrophorhabdus aromaticivorans

WP_071980017.1|1917158|Alteromonas sp. RW2A1

WP_050631537.1|1654716|Bradyrhizobium viridifuturi

WP_078813596.1|48467|Prosthecobacter debontii

WP

_092

0336

21.1

|185

5398

|Por

phyr

omon

adac

eae

bact

eriu

m K

HP

3R9

WP_0

7968

5319

.1|6

8886

7|Oht

aekw

angia

kore

ensis

esneiranacmuibozihrydar

B|540552|1.463053580_P

W

WP

_080902475.1|294|Pseudom

onas fluorescens

WP

_014200913.1|253245|Ow

enweeksia hongkongensis

WP

_047468669.1|1339242|Delftia sp. ZN

C0008

WP_005372458.1|39773|Methylomicrobium

WP_018005262.1|164546|Cupriavidus taiwanensis

WP

_071

9700

84.1

|191

7485

|Sul

fitob

acte

r sp

. AM

1-D

1

WP

_008510818.1|1218315|Brucella inopinata

WP_097674963.1|142585|Bradyrhizobium sp. C9

WP

_049

9067

81.1

|119

434|

Hal

orub

rum

tebe

nqui

chen

se

WP_

0599

4530

4.1|

1503

055|

Burk

hold

eria

terri

torii

WP_010683362.1|39956|Methylobacterium mesophilicum

WP_051880056.1|421072|Chryseobacterium halperniae

WP_044237100.1|122|G

imesia

maris

WP_004308218.1|816|Bacteroides

WP_051600234.1|1280944|Hyphomonas sp. CY54-11-8

WP_013078838.1|88688|Caulobacter segnis

WP_058745238.1|172044|Sphingom

onas yabuuchiae

WP_091379109.1|1881103|Geodermatophilus sp. DSM 45219

WP_012473795.1|1096|Chlorobium phaeobacteroides

WP_072995904.1|192903|Pseudozobellia thermophila

WP_042419826.1|405782|Streptacidiphilus anmyonensis

WP_002993155.1|28453|Sphingobacterium

WP

_090997603.1|414048|Pedobacter insulae

WP_091126757.1|574651|Nocardioides terrae

WP_022545094.1|1400053|Bacteroidales bacterium CF

WP_093736217.1|1839759|Streptomyces sp. D

valAA-14

esneruhtam

muibognihpsovoN|099824|1.245137970_

PW

WP_101944482.1|2067453|Uliginosibacterium sp. TH139

WP_011923975.1|114615|Bradyrhizobium sp. ORS 278

WP_0

9062

4110

.1|14

7743

7|Par

aped

obac

ter i

ndicu

s

WP_102206007.1|372787|Fischerella thermalis

WP_056695931.1|414371|Aureim

onas

WP

_052

5648

96.1

|795

830|

Can

dida

tus

Bro

cadi

a si

nica

WP_099104184.1|1206980|Nostoc sp. 'Peltigera m

alacea cyanobiont' DB3992

WP_016263566.1|286|Pseudomonas

WP_0

8423

4542

.1|1

0017

6|Pap

illiba

cter

cin

nam

ivora

ns

WP

_070666720.1|838|Prevotella

WP_071098581.1|1172030|M

icromonosp

ora sp. W

MMB235

WP_053326746.1|1324352|Chryseobacterium gallinarum

WP

_049

8970

37.1

|689

11|N

atria

lba

chah

anna

oens

is

WP_081705202.1|1416614|G

loeobacter kilaueensis

WP_

0174

5467

8.1|

8084

2|H

erba

spiri

llum

rubr

isub

albi

cans

WP_073904454.1|1490482|Mycobacterium sp. SWH-M3

WP

_074121857.1|1680155|Bradyrhizobium

sp. AS

23.2

WP_069020256.1|1818881|Candidatus Thiodiazotropha endoloripes

WP_099463435.1|1944636|Parabacteroides sp. Marseille

-P3668

WP

_077292092.1|187304|Labrenzia aggregataWP_082670027.1|33051|Sphingomonas sanguinis

WP_083637046.1|1922337|Leptolyngbya

sp. 'h

ensonii'

WP_004748890.1|29498|Vibrio tubiashii

WP

_088664285.1|1317112|Roseovarius sp. 22II1-1F6A

WP_0

8728

7470

.1|1

9655

76|P

seud

oflav

onifr

acto

r sp.

An1

84

WP

_022

7873

70.1

|877

414|

Clo

strid

iale

s ba

cter

ium

NK

3B98

WP

_083779632.1|314267|Sulfitobacter sp. N

AS

-14.1

WP

_015

4491

54.1

|753

09|R

hoda

noba

cter

WP

_092

8388

21.1

|100

5928

|Ros

eova

rius

lutim

aris

WP_023692463.1|1287325|M

esorhizobium sp. LSJC268A00

WP_069707909.1|158500|Novosphingobium resinovorum

WP_090828645.1|1233|Nitrosovibrio tenuis

WP

_079392070.1|286|Pseudom

onas

WP_041534907.1|208216|Parvularcula bermudensis

WP_013833356.1|165696|Novosphingobium

WP_098195322.1|2029983|Chitinophagaceae bacterium 13

WP_0

0887

7326

.1|4

8972

2|M

esor

hizo

bium

met

allid

uran

s

WP

_085984230.1|1131813|Methylobacterium

sp. 88A

WP

_081

4351

15.1

|160

791|

Sph

ingo

mon

as w

ittic

hii

WP

_026

8086

85.1

|861

04|A

reni

bact

er la

teric

ius

WP_0

4809

0856

.1|1

3929

98|C

andid

atus

Met

hano

pere

dens

nitr

ored

ucen

s

WP

_053

0581

40.1

|909

625|

Rub

roba

cter

apl

ysin

ae

WP

_006270760.1|76891|Asticcacaulis biprosthecium

WP_010123961.1|1030157|Sphingom

onas sp. KC8

WP_

0564

9544

3.1|

1735

692|

Sphi

ngom

onas

sp.

Lea

f25

WP_

0476

1820

2.1|

398|

Rhi

zobi

um tr

opic

iW

P_065664090.1|1841655|Agrobacterium sp. B131/95

WP_065010823.1|68287|Mesorhizobium

WP_0

1197

1023

.1|1

1032

1|Sin

orhi

zobi

um m

edica

e

WP_

0599

8250

1.1|

1015

71|B

urkh

olde

ria u

bone

nsis

WP_099503355.1|69665|Caulobacter sp. FWC26

WP_091516911.1|998|Flexibacter flexilis

WP

_081

7432

00.1

|288

965|

[Clo

strid

ium

] cla

rifla

vum

WP_099463422.1|1944636|Parabacteroides sp. Marseille-P3668

WP_0

2503

0983

.1|2

0479

9|Nitr

atire

duct

or a

quib

iodo

mus

WP_083657918.1|1892903|Herbaspirillum sp. W

T00C

WP

_092

6623

01.1

|136

7881

|Hal

orie

ntal

is p

ersi

cus

WP_019960613.1|240237|Woodsholea maritima

WP_092048392.1|1576369|Planctomicrobium piriforme

WP

_089355802.1|447679|Ekhidna lutea

WP_085773725.1|655015|Methylocystis bryophila

WP_094530879.1|1980935|Pseudanabaena sp. SR411

WP_008706033.1|1265734|Rhodopirellula maiorica

WP_073409929.1|986|Flavo

bacteriu

m johnso

niae

WP_0

4747

9138

.1|15

0028

3|Chr

yseo

bacte

rium sp

. CF35

6

WP_082340617.1|339|Xanthomonas campestris

WP_035638686.1|566679|Bradyrhizobium

sp. OR

S 375

WP

_092

9101

36.1

|188

2830

|Rhi

zobi

ales

bac

teriu

m G

AS

191

WP

_073365442.1|1124188|Flavobacterium

fontis

WP_013109336.1|120|Planctopirus limnophila

WP

_009301608.1|665942|Desulfovibrio sp. 6_1_46A

FAA

WP_084467687.1|103730|Actinokineospora inagensis

WP_065006787.1|1854057|M

esorhizobium sp. AA22

WP_052744930.1|1538294|Micr

omonospora sp

. HK10

WP_022842741.1|278963|Acidobacteriaceae bacterium TAA166

WP

_082

5495

97.1

|173

6555

|Mes

orhi

zobi

um s

p. R

oot5

52

WP_083869695.1|477641|Modestobacter marinus

WP_083592922.1|1317114|Aurantim

onas sp. 22II-16-19i

WP_049284968.1|286|Pseudomonas

WP_0

0462

4807

.1|2

9371

|[Clo

strid

ium

] ter

mitid

is

WP_095840540.1|2033437|Chitinophaga sp. MD30

WP_046756489.1|1348245|Kordia jejudonensis

WP

_042

6191

22.1

|358

|Agr

obac

teriu

m tu

mef

acie

ns

WP

_008

3879

95.1

|411

361|

Hal

ogeo

met

ricum

pal

lidum

WP_066752989.1|1685010|Chryseobacterium sp. IHBB 10212

WP

_095

4474

52.1

|271

865|

Och

roba

ctru

m s

p. A

44

WP_075061401.1|11

34406|Orn

atilinea apprim

a

WP

_082

7559

86.1

|135

5477

|Bra

dyrh

izob

ium

dia

zoef

ficie

ns

WP

_009043888.1|587753|Pseudom

onas chlororaphis

WP

_095387429.1|2012632|Sphingopyxis sp. G

W247-27LB

WP

_089

7614

43.1

|408

074|

Chi

tinop

haga

terr

ae K

im a

nd J

ung

2007

WP_061535229.1|279058|Collimonas arenae

WP_052259141.1|56459|Xanthomonas vasicola

WP

_052

4942

27.1

|581

33|N

itros

ospi

ra s

p. N

pAV

WP_095565911.1|2024580|Plantactinospora sp. KBS50

WP_092226358.1|1855318|Bradyrhizobium sp. Gha

WP

_010913851.1|2066070|Mesorhizobium

japonicum

WP_099588558.1|40324|Stenotrophomonas maltophilia

WP_075628770.1|464029|Rhizobium oryzae

WP_0

1997

7020

.1|2

47|E

mpe

doba

cter

bre

vis

WP

_074487307.1|1302689|Mucilaginibacter polytrichastri

WP_084522028.1|53432|Nocardia uniformis

WP_044272248.1|1470345|Bacteroides timonensis

WP_015664541.1|44255|Bradyrhizobium oligotrophicum

WP_084625517.1|1207055|Sphingobium sp. C100

WP_087337819.1|357276|Bacteroides dorei

WP_

0122

2225

8.1|

3399

6|G

luco

nace

toba

cter

dia

zotro

phic

us

WP_068133588.1|980254|R

oseim

aritima ulva

e

WP

_064835082.1|396|Rhizobium

phaseoli

WP_071507486.1|305|Ralstonia solanacearum

WP_028095445.1|337702|Pseudodonghicola xiamenensis

sano

mogn

ihp

S|78

631|

1.65

7804

640_

PW

WP_

0236

9906

8.1|

1287

322|

Mes

orhi

zobi

um s

p. L

SJC

265A

00

WP

_036284031.1|622637|Methylocystis sp. ATC

C 49242

WP_068768974.1|1184151|Opitutaceae bacterium TSB47

WP

_068017268.1|674703|Rhodoplanes sp. Z

2-YC

6860

WP_051443965.1|1429916|Afipia sp. P52-10

WP_084182446.1|39687|Mycobacterium austroafricanum

WP_062341847.1|1768752|Novosphingobium

sp. CCH12-A3

WP

_036632782.1|823|Parabacteroides distasonis

WP

_092416776.1|1801620|Collim

onas sp. OK

307

WP

_034679768.1|236814|Chryseobacterium

formosense

WP_061048384.1|221822|Phaeobacter inhibens

WP_073132933.1|947013|Chryseolinea serpens

WP_005988489.1|56460|Xanthomonas vesicatoria

WP_028057923.1|1298857|Solirubrobacter s

p. URHD0082

WP

_098195162.1|2029983|Chitinophagaceae bacterium

13

WP_004290502.1|28111|Bacte

roides eggerth

ii

WP

_081

4449

39.1

|124

|Bla

stop

irellu

la m

arin

a

WP_090603244.1|332977|Parapedobacter koreensis

WP_095589297.1|1590614|Actibacterium

ureilyticum

11rr

e.ps

muib

ozih

ryda

rB|

2771

671|

1.09

7369

190_

PW

WP

_093307966.1|34072|Variovorax

WP_024740054.1|107401|Te

nacibacu

lum mariti

mum

WP_065137571.1|1778|Mycobacterium gordonae

WP

_078

3669

26.1

|280

66|R

hodo

fera

x fe

rmen

tans

WP

_083

2407

32.1

|177

4968

|Met

hylo

cean

ibac

ter

met

hani

cus

WP

_030539670.1|1303256|Sphingobium

sp. DC

-2

WP_059000176.1|1552121|Leptolyngbya sp. NIES-2104

WP_082932868.1|190721|Ralstonia insidiosa

WP_093896586.1|1839774|Streptomyces sp. Ncost-T10-10d

WP

_067673142.1|1822215|Erythrobacter sp. H

I00D59

WP_050591131.1|68287|Mesorhizobium

WP

_004

5941

99.1

|292

82|H

aloa

rcul

a ja

poni

ca

WP_085936213.1|225324|Enhydrobacter aerosaccus

WP_051882337.1|1247963|Parvularcula oceani

WP_083960493.1|980561|Methylomonas lenta

WP_005801097.1|817|Bacteroides fragilis

WP

_096683443.1|1973475|Nostoc sp. N

IES

-2111

WP

_052

1921

92.1

|156

4113

|Sph

ingo

mon

as s

p. A

nt H

11

WP_0

9117

0676

.1|5

5199

6|M

ucila

giniba

cter g

ossy

pii

WP

_005

2771

91.1

|121

7695

|Aci

neto

bact

er s

p. N

IPH

185

9

WP_034728376.1|1233950|Chryseobacterium sp. JM1

WP_062547850.1|1727164|Pedobacter sp. PACM 27299

WP_042087938.1|1492281|alpha proteobacterium Q-1

WP_072766083.1|558155|Arenibacter n

anhaiticus

WP_096614338.1|862134|Sphingomonas ginsenosidimutans

WP_085870580.1|1839|Nocardioides

WP

_097

6138

65.1

|203

5448

|Rhi

zobi

um s

p. C

5

WP

_006019181.1|56946|Afipia broom

eae

WP_064665416.1|182141|Pseudoalteromonas prydzensis

WP

_065656395.1|358|Agrobacterium

tumefaciens

WP_084685017.1|244365|Methylohalobius crimeensis

WP_027853030.1|404770|Marinobacterium litorale

WP

_049858702.1|1681194|Pseudom

onas sp. RIT-P

I-a

WP

_015596988.1|53399|Hyphom

icrobiumdenitrificans

WP_074343878.1|36809|Mycobacterium abscessus

WP_102073885.1|2045210|Pusillimonas sp. JR1/69-3-13

WP_080700186.1|103855|Bordetella hinzii

WP_079791392.1|28901|Salmonella enterica

WP_085809851.1|1917973|Sphingomonas sp. TZW2008

WP

_022727544.1|1121832|Fodinicurvata sedim

inis

WP_080810443.1|1209493|H

alomicronema hongdech

loris

WP_036945621.1|118173|Pseudanabaena sp. PCC 6802

WP

_080

5747

28.1

|294

49|R

hizo

bium

etli

WP

_091535647.1|1780376|Alistipes sp. C

HK

CI003

WP_083871010.1|1288121|Alistipes senegalensis

WP_087406053.1|1965620|Bacteroides sp. An279

WP_051677674.1|1380355|Bradyrhizobium sp. URHD0069

WP

_081912819.1|1522072|Sphingobium

sp. ba1

WP_011940822.1|351604|Geobacter uraniireducens

WP_013801384.1|742013|Delftia sp. Cs1-4

WP_091639649.1|145854|Micromonospora pallida

WP

_081

4419

52.1

|314

344|

Mar

ipro

fund

us fe

rroo

xyda

ns

WP

_064790849.1|502049|Thalassospira profundim

aris

WP

_051294653.1|220343|Gem

mobacter nectariphilus

WP_013596009.1|80867|Acidovorax avenae

WP_0

1438

7263

.1|31

2277

|Flav

obac

terium

indic

um

WP_099076851.1|92942|Nostoc linckia

WP_008137934.1|1144344|Bradyrhizobium

sp. YR681

WP

_051

6025

27.1

|85|

Hyp

hom

onas

WP_

0894

5754

7.1|

2015

357|

Burk

hold

eria

sp.

AU33

803

WP_080522385.1|1420851|Methyloprofundus sedimenti

sirtsulapsano

moduespodohR|6701|1.120405110_

PW

WP_051634474.1|358|Agrobacterium

tumefaciens

WP_057482737.1|270918|Salegentibacter mishustinae

murasonimugel

muibozihR|483|1.748615050_

PW

WP

_092

2827

52.1

|937

334|

Cal

dico

prob

acte

r fa

ecal

is

02S

D.pssano

midnuverB|5552351|1.472567450_

PW

WP_042571950.1|321846|Pseudomonas simiae

WP_083808707.1|1032527|Pedosphaera parvula

WP_

0882

0156

1.1|

382|

Sino

rhizo

bium

mel

iloti

WP

_015

8958

34.1

|330

75|A

cido

bact

eriu

m c

apsu

latu

m

WP

_061934617.1|1638162|Aureim

onas sp. AU

22

WP_057026385.1|108015|Bradyrhizobium yuanmingense

WP_009860218.1|328812|Parabacteroides goldsteinii

WP

_084

4627

63.1

|580

166|

Oce

anib

acul

um p

acifi

cum

WP

_089

7698

62.1

|660

517|

Hal

obel

lus

clav

atus

WP_065578027.1|1196095|Gilliamella apicola

WP_091471579.1|492660|Methylophilus rhizosphaerae

WP

_081147616.1|1703345|Niastella vici

WP

_101

2375

71.1

|177

6741

|Aci

neto

bact

er p

rote

olyt

icus

WP_066514273.1|120107|Sphingobium cloacae

WP_0

6732

5608

.1|1

8540

58|M

esor

hizo

bium

sp.

AA23

WP_073243758.1|983|Flavobacterium flevense

WP_

0101

6514

8.1|

1112

216|

Sphi

ngom

onas

sp.

PAM

C 2

6617

WP

_078

7925

95.1

|238

|Eliz

abet

hkin

gia

men

ingo

sept

ica

WP_057221537.1|106591|Ensifer

WP_016736429.1|396|Rhizobium phaseoli

WP_088892393.1|1962290|Leptolyngbya ohadii

WP_057834160.1|280332|Bradyrhizobium jicamae

WP

_074

7935

83.1

|357

99|N

itros

ospi

ra b

riens

is

WP_060847086.1|270351|M

ethylobacterium aquaticum

WP_099746952.1|48296|Acinetobacter pittii

WP

_014

7461

85.1

|171

437|

Tist

rella

mob

ilis

Tree scale: 2

pAgo operons pAgo fusion

short pAgo

superkingdom

Archaea

Bacteria

APAZ in pAgo operon

Ia

Ib

IIa

IIb

operonpAgo fusionkingdom

proteintype

MID

APAZ

APAZ

APAZ

APAZ

APAZDUF4365

III

FIG 5 Phylogenetic analysis of APAZ domains. The circular phylogenetic tree of the five groups of APAZ domains. The phylogenetic tree was annotated asfollows, from the inner to the outer circles: in the “protein type” circle the APAZ-containing proteins are colored according to their phylogenetic groups anddomain compositions, where each type of protein structure is exemplified by the callouts with corresponding domain schemes; isolated APAZ domains in allgroups are shown in light blue; the “operon” circle indicates whether the APAZ-containing protein is found in a pAgo coding operon; the “pAgo fusion” circleshows APAZ domains which are fused with short pAgo proteins; and the “kingdom” circle indicates the superkingdom to which the correspondingAPAZ-containing protein belongs.

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 10

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

Functional classification of proteins enriched in the genomic context of pAgos.Analysis of the functions of proteins encoded in the same operons with pAgos mightshed light on the molecular pathways involving pAgos. We therefore analyzed thegenomic context of pAgos encoded in the 1,385 genomes of 1,248 strains and exploredthe proteins encoded in their proximity. We defined a window centered on pAgoencompassing 20 genes (i.e., 10 genes upstream and 10 genes downstream of pAgo)in each genome. The 16,274 proteins encoded in these windows were clustered intoorthogroups based on the sequence similarities that resulted in 1,892 orthogroups thathad at least 2 proteins (see Materials and Methods). The orthogroups contained 12,668proteins, and the largest one comprised 298 related proteins (Table S2). We furtherdefined pAgo operons that included only genes that are continuously located on thesame genomic strand as pAgo, which resulted in a set of operons with a mean size of6.4 genes. We separately considered operons of five different groups of pAgos (Fig. 6):

RecB(52%), APAZ(96%)

PD

-(D

/E)X

K p

redi

ctio

n

1. w

/o A

PA

Z2.

with

SIR

2-A

PA

Z3.

long

-A4.

long

-B5.

long

-A w

ith M

ID-O

H

Fraction of predicted PD(E/D)XK proteins in orthogroup

Percentage of operonswith protein(-s) fromthe orthogroup

BLAST Pfam

Annotation of orthogroups (% of proteins)

Prim-pol (SSF56747(82%), SSF48452(73%))

***

***

**

*

*

*

**

*

*Cas4Cas4

SIR2_1-APAZ

Putative DNA-binding domain (Schlafen, AlbA_2), SIR2_1, APAZ

SIR2_2

Repeat Associated Mysterious Proteins (RAMP)

0

Cas6

Ort

hogr

oups

of p

rote

ins

?

?

?

?

RecB2

*

DNA-binding domain

* Nuclease activity

P-loop NTPase domainHelicase activity

Hydrolase/phosphataseSAM-dependent methyltransferases

efflux membrane transport

?

FIG 6 Functional annotations of proteins encoded in pAgo operons. The heat map on the left shows the percentage of operons for each pAgo group(columns) that have at least one protein from the particular orthogroup (rows). The orthogroups are ordered according to the number of proteins in eachgroup, from the largest to the smallest. Since many operons do not contain any proteins included in the orthogroups analyzed here, the total number ofoperons with orthogroup-specific proteins may be less than 100%. The PD-(D/E)XK column shows the fraction of proteins from each orthogroup that arepredicted to be nucleases from the PD-(D/E)XK superfamily. The annotation of proteins was performed by BLAST analysis for identification of APAZ, RecB-like,and RecB2 domains and by InterProScan using the Pfam and Superfamily databases. Only the Pfam annotation of domains is shown (with the exception ofog_209, for which the Superfamily annotation is shown). The percentage in the brackets after the domain name shows the fraction of proteins in thecorresponding orthogroup having this type of domain. Since proteins may contain several domains, the sum of percentages in each row may be greater than100%. Only Pfam or Superfamily domains found in at least 30% of orthogroup proteins are shown. The descriptions of domain functions are shown on theright of the panel. Proteins from several orthogroups do not have any known domains and are marked by “?.” After the manual curation of domain functions,several of the most common functional features were selected, which are shown as colored symbols and described below the table.

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 11

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

(i) short pAgos that are not fused to the APAZ domain, (ii) short pAgos that contain theSIR2-APAZ domains, (iii) all long-A pAgos except the MID-OH variants, (iv) long-B pAgos,and (v) long-A pAgos with the MID-OH domain. Proteins from 47 orthogroups that werepresent in at least 5% of operons in at least one of the five pAgo groups were furtheranalyzed in detail. In Fig. 6, the orthogroups are sorted in descending order of thenumber of proteins in each orthogroup.

We defined the domain structure of proteins in each orthogroup using InterProScanwith Pfam, Superfamily, and CDD databases containing known domain annotations. Foridentification of the APAZ, RecB-like, and RecB2 domains, previously described byMakarova et al. (11) and absent from these databases, we performed PSI-BLAST searchusing previously identified proteins as queries. In addition, we proposed that there maybe other pAgo-associated proteins belonging to the PD-(D/E)XK superfamily of nu-cleases, besides previously identified Mrr (RecB-like and RecB2) and Cas4 proteins (11,12). To check this, we have searched for the characteristic PD-(D/E)XK sequence motifin the orthogroup proteins. The identified domains found in each orthogroup ofproteins and the putative functional types of pAgo-associated proteins are shown inFig. 6.

The proteins with APAZ domains are present in the orthogroups og_3, og_4, og_9,and og_40. The proteins from these orthogroups also contain other types of domains,including DUF4365 (og_3), SIR2_1 subtype of SIR2 (og_4), TIR (og_9), and SIR2_1 fusedwith Schlafen domain (og_40). Most of the revealed APAZ-containing proteins arelocated in group 1 operons of short pAgos that do not have their own APAZ domain.In total, 91% of group 1 operons contain proteins with APAZ. In contrast, only 6% ofgroup 2 operons of short pAgo fused with APAZ and SIR2 encode APAZ-containingproteins. Taken together, 93% of short pAgos are associated with the APAZ domain asa part of the same protein or present in the same operon with short pAgos. In contrast,long pAgos are only rarely coencoded with APAZ proteins (2%, 3%, and 0% of operonsfrom groups 3, 4, and 5, respectively).

We found several different types of nucleases and DNA-binding proteins abundantin the pAgo operons (Fig. 6). In total, 17 out of 47 orthogroups (36%) encode proteinswith putative nuclease activity. These proteins are found in 68% of pAgo operons.Ninety-two percent of operons of short pAgos that themselves do not have the APAZand SIR2 domains (group 1 of pAgos) encode different types of nucleases. In contrast,only 32% of operons of short pAgos fused with the SIR2 and APAZ domains (group 2)contain nucleases. For group 3 of long-A pAgos, 45% of pAgos carrying the catalytictetrad have nucleases in their operons, while this number is increased to 53% foroperons of long-A lacking the DEDX tetrad. This number is further increased to 67% forgroup 4 of long-B pAgos, which are all catalytically inactive. The nucleases found in thepAgo operons are quite diverse and include proteins with SIR2 (og_4, og_40, andog_44), TIR (og_9), and RecB2 (og_15) domains; the DUF4365 domain that belongs toRecB-like nucleases (og_3); Cas4 proteins (og_56 and og_60), and phospholipase Dproteins with putative nuclease activity (og_136). A set of pAgo operons also containproteins with putative nucleases of the HNH type with DUF252 and DUF1254 domains(og_28).

Two pAgos from Marinitoga piezophila and Thermotoga profunda, which are repre-sentatives of the small group of long-A pAgos containing the MID-OH domain (Fig. 1B,2B, and 3A), were previously reported to be encoded in the CRISPR loci between theCas1 and Cas2 genes (24). We found that all pAgos containing the MID-OH domainfrom 9 prokaryotic strains (group 5 of pAgos) are encoded within CRISPR loci with Cas1(og_154) and Cas2 (og_209) genes nearby (Fig. 6). Other proteins specific for thesubtype III-B CRISPR locus are also observed in association with MpAgo-like pAgos: Cas6(og_190), primase (og_209), and Cmr (Cas module RAMP) proteins (og_455).

In addition, proteins in several other orthogroups have motifs characteristic for thePD-(D/E)XK nucleases [Fig. 6, the PD-(D/E)XK column]. These proteins include the UvrDhelicase and P-loop NTPase (AAA� subfamily) domains (og_6), SAM-dependent meth-yltransferase domain (og_13), restriction endonuclease of type III (og_21, og_34), DEAD

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 12

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

box helicase and P-loop NTPase (AAA� subfamily) (og_34 and og_37), YhgA-typeputative transposases (og_72), and proteins without known domains (og_257 andog_603). Notably, proteins from three of these orthogroups have predicted helicaseactivity (domains related to the UvrD helicase, og_6, and DEAD box helicase, og_34 andog_37).

Proteins with DNA-binding domains are abundant in operons of short pAgos fusedto the APAZ and SIR2 domains (33%) and of long-B pAgos (33%). In addition, proteinswith DNA-binding domains are present in 19% and 13% of operons of long-A pAgosand short pAgos without APAZ-SIR2 domains, respectively. The DNA-binding domainsin pAgo-associated proteins include the lambda repressor-like DNA-binding domain(og_2), helix-turn-helix DNA-binding domains (og_18, og_49, og_73, and og_79), theSchlafen domain with an AlbA_2 subdomain (og_37), and DNA-binding components ofthe type I restriction-modification system (HsdS) (og_11). Interestingly, the lambdarepressor-like DNA-binding domain (PF01381) from highly represented orthogroupog_2 and the Schlafen DNA-binding domain AlbA_2 (PF04326) from og_37 were foundto be fused with several long-B pAgos (Fig. 2B).

Overall, analysis of proteins in pAgo operons strongly indicates the involvement ofnuclease, helicase, and DNA-binding activities in cellular pathways dependent onpAgos (see Discussion). Beyond nucleases and DNA-binding proteins, 10% of all longpAgo operons encode hydrolases and phosphatases (og_47, og_62, and og_55). Finally,14% of long-B pAgo operons encode multidrug efflux transporters of the resistance-nodulation-cell division (RND) superfamily (og_14 and og_31), but their possible rolesin the functional pathways involving pAgos remain unknown.

In the present analysis, we could not reveal any defined operon compositions forspecific pAgo groups. While the median number of all genes associated with pAgo onthe same strand is 6.4, these numbers are much lower if we consider only proteinsincluded in the orthogoups analyzed here: 2.0 for all short pAgos, 3.0 for each long-Aand long-B pAgo, and 5.0 for the MID-OH long-A pAgos. The small sizes of operons onthe one hand and the large number of orthogroups on the other result in numerouspossible operon configurations, making the construction of the consensus operonstructure impossible. The only exception is the MID-OH long-A pAgos, for which thestructure can be presented as Cas1-Cas2-Cas6-pAgo-Cmr. The same operon schemewas reported for MpAgo previously (24). The list of all observed configurations oforthogroup proteins in pAgo operons is presented in Table S3.

The genomes encoding pAgos do not show decreased transposon copy num-ber. Several eukaryotic Ago proteins were shown to prevent propagation of trans-posons and viruses in the genomes (reviewed in references 44 and 45). Similarly, a fewstudied pAgos were proposed to protect prokaryotic cells from foreign DNA, includingplasmids, transposons, and phages (38, 40, 41). In particular, pAgos may be involved inthe protection of the prokaryotic genome from the invasion of mobile genetic elementsor suppress the activity of genomic copies of such elements. If pAgos are able to reducethe expansion of transposons in their host genomes, then the number of transposoninsertions should be lower in strains that encode pAgo. To test this hypothesis, wecounted the number of transposon insertions in the genomes of strains encoding ornot encoding pAgos from different groups (short pAgos, short pAgos fused to theSIR2-APAZ domains, long-A with PIWI and PIWI* domains, long-B pAgos, and long-ApAgos with the MID-OH domain). We found that the genomes of strains encodingpAgos do not have lower copy numbers of transposons than strains without pAgogenes (Fig. 7A). Moreover, we observed a moderate but significant increase in the meannumber of transposon copies per genome in pAgo-contaning strains for both short andlong classes of pAgos. For the MID-OH pAgos, the number of transposon copies incorresponding genomes was smaller than in pAgo-minus strains, but this differencewas statistically insignificant due to the small number of pAgos in this group. Sincedifferent phylogenetic groups may contain different transposon numbers, we com-pared the transposon content within several genera (Fig. 7B). For this, we selected 32prokaryotic genera having at least 10 strains with sequenced genomes encoding and

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 13

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

not encoding pAgos. For the strains of each genus, the transposon copy numbers weresubjected to log2 transformation and z-scaling. If a strain has a z-score of ��2, thenthere is only a 2.3% chance that its genome has the transposon content typical for thisgenus, and so it can be considered the one that has a significantly lower transposonnumber than average. We have found that only one strain among 108 pAgo-encodingstrains from 32 genera (703 strains in total) has a z-score of ��2 (Fig. 7B). The genomesof all other strains with pAgo have the transposon copy numbers that are typical of thecorresponding genera. Overall, our results indicate that pAgos present in the genomesmay not be able to stop transposon insertions; however, this does not exclude apossibility that pAgos can repress transposon expression.

DISCUSSION

Eukaryotic systems of RNA interference have evolved from several diverse prokary-otic components, including RNase III-like enzymes, RNA-dependent RNA polymerases,helicases, and pAgos, which gave rise to various types of eAgos and RNAi pathways(46). While being highly variable and involved in different interference pathways, alleAgos share the same modus operandi: they always bind small RNA guides to locate andinactivate cRNA targets (or attract other components to modify corresponding codingDNA loci). Perhaps not surprisingly, pAgos are much more diverse, with significantsequence and domain variations (11, 12). Moreover, a few published functional studiesof pAgos demonstrated that, while retaining the same basic principle of guided nucleicacid recognition, they may significantly differ in their modes of action in comparisonwith eukaryotic counterparts (38–41, 47). We have now revealed even a larger diversityof pAgos that allows us to suggest their updated classification and propose possibleevolutionary pathways and cellular functions for this conserved family of nucleicacid-targeting proteins.

*******

*****

193 37 69 30 75 4 5914

2

64

2048

shor

t pA

gos

shor

t pA

gos

with

SIR

2-A

PAZ

long

-A p

Ago

s w

ith P

IWI

long

-A p

Ago

s w

ith P

IWI*

long

-B p

Ago

s

long

-A p

Ago

s w

ith M

ID-O

H

no p

Ago

Tn

copi

es in

gen

omes

(lo

g2)

A B

1/11

11/2

7

3/11

2/17

8/19

15/5

4

2/11

3/31

3/11

1/19

4/13

2/11

10/1

5

1/13

3/12

3/62

5/13

2/40

1/10

1/11

5/12

1/22

1/17

2/23

3/14

2/93

5/29

1/10

1/11

1/28

4/19

1/14

-2

0

2

Aci

dovo

rax

Aci

neto

bact

er

Alte

rom

onas

Bor

dete

lla

Bra

dyrh

izob

ium

Bur

khol

deria

Cal

othr

ix

Clo

strid

ium

Cup

riavi

dus

Ent

erob

acte

r

Geo

bact

er

Kle

bsie

lla

Mes

orhi

zobi

um

Met

hano

sarc

ina

Met

hylo

bact

eriu

m

Myc

obac

teriu

m

Nos

toc

Pae

niba

cillu

s

Par

abur

khol

deria

Pla

noco

ccus

Pre

vote

lla

Pse

udoa

ltero

mon

as

Ser

ratia

She

wan

ella

Sph

ingo

bium

Stre

ptom

yces

Ther

moc

occu

s

Ther

mot

oga

Trep

onem

a

Vibr

io

Xan

thom

onas

Yers

inia

Genus

Tn c

opy

num

ber i

n ge

nom

es (l

og2,

z-s

cale

d)

short pAgos

short pAgos with SIR2-APAZ

long-A pAgos with PIWI

long-A pAgos with PIWI*

long-B pAgos

long-A pAgos with MID-OH

FIG 7 The number of transposon copies in genomes with and without pAgo genes. (A) The violin plot shows the log2-transformed distribution of copy numbersof transposons in the genomes of prokaryotic strains encoding or not encoding pAgos. The black dots with whiskers indicate the means and standarddeviations. The numbers of analyzed genomes for each group of pAgos are shown below each violin. The P values shown above the plot were calculated withthe Wilcoxon two-sided rank test: ****, P value � 1 � 10�4; ***, P value � 1 � 10�3; **, P values � 0.05. (B) The violin plot shows log2-transformed and z-scaleddistribution of copy numbers of transposons in the genomes of prokaryotic strains encoding or not encoding pAgos for selected genera. The colored dotsindicate the transposon content of strain genomes encoding pAgos; the correspondence of pAgo type and dot color is specified above the plot. The numbersof analyzed genomes for each genus are shown above each violin (strains with pAgo/total number of strains).

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 14

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

The diversity of pAgos. Our results revealed that pAgos are widespread in archaealand eubacterial species. The availability of new sequenced genomes allowed identifi-cation of an almost 3-fold-larger set of pAgos than previous analysis. Analysis of thisexpanded set demonstrated that pAgos are highly diverse and can be separated intothree large clades, short, long-A and long-B, which diverged early in evolution (Fig. 1B;see also Fig. S1 in the supplemental material). While this division is based solely onsequence alignments of the MID-PIWI domains, these groups also reveal characteristicvariations in specific structural elements involved in nucleic acid binding and process-ing. Thus, the members of the long-A clade mostly contain both catalytically active PIWIdomains and PAZ domains of normal size. In contrast, long-B pAgos contain inactivePIWI* domains and also usually shortened PAZ* domains, although other combinationswere also observed for both clades (Fig. 1 and 2). Furthermore, proteins from the long-Aclade contain several variants of the 5=-P-guide binding motif while long-B pAgosalmost always contain the canonical YK motif. Short pAgos always contain substitutionsin the active site (PIWI*) and a subset of canonical MID motifs (HK/RK).

The exact relationship between the three clades is ambiguous, with two possiblealternative scenarios (12). In one scenario, short and long pAgos diversified first,followed by a further split of long pAgos into two clades. Alternatively, catalyticallyinactive long-B pAgos and short pAgos may have originated from a common ancestorcontaining substitutions in the active site. The phylogenetic analysis of the MID-PIWIdomains has shown that the evolutionary distance between combined short pAgo andlong-B pAgo clades on the one side and the long-A clade on the other is slightly largerthan between short and combined long-A and long-B clades (Fig. S1). This argue infavor of the second scenario of the evolutionary origin of the three clades of pAgos.

Binding of nucleic acids by pAgos. Tight association with small nucleic acid guidesseems to be a defining feature of pAgos and eAgos alike (reviewed in references 12 and59). Analysis of the nucleic acid binding pockets of the MID domain revealed that themajority of pAgos that belong to all three clades (short, long-A, and long-B) likely bindsmall nucleic acid guides with monophosphorylated 5= ends. At the same time, wediscovered significant variations in the conserved 5=-guide binding motifs in the long-ApAgos, including substitutions of key residues involved in the 5=-phosphate interac-tions (i.e., YK, YY, KR, YR, MID*, and MID-OH variants). While most these variants likelyinteract with 5=-P-guides, some proteins might bind guides with an alternative 5=-endstructure, as already shown for the MID-OH pAgos, or not use nucleic acid guides at all.Interestingly, almost all members of the short pAgo clade have MID variants that areclose to the consensus (HK/RK), suggesting that the canonical MID pocket structure isessential for their function in the absence of the PAZ domain. Future biochemicalcomparisons of the long and short pAgos with various noncanonical MID motifs shouldfind out if the observed differences in the 5=-end pockets may influence the specificity,affinity, and kinetics of their interactions with guide and target molecules.

While all known eAgos use small RNA guides to recognize RNA targets, the majorityof proteins in a relatively small sample of pAgos characterized so far prefer to bindsmall DNA as a guide, except for RsAgo (38) and MpAgo (24), which associate with smallRNAs. Furthermore, all studied pAgos preferably interact with DNA targets. Few aminoacid residues involved in the binding of ribose and deoxyribose backbone of the guideand target molecules were identified in the structures of experimentally characterizedpAgos (17, 23–25, 28); however, these residues are not well conserved. Therefore,sequence analysis is not able to predict preference for DNA versus RNA binding, andthis information has to be obtained experimentally.

The PAZ domain of most studied eAgo and pAgo proteins is composed of twosubdomains that together form the 3=-end guide binding pocket (16, 18, 20, 28).However, our results revealed different organizations of the PAZ domain among longpAgos, indicating that a single conserved structure of PAZ may not be required forguide binding. The structure is not conserved even among the few experimentallycharacterized pAgos. While AaAgo and TtAgo contain both subdomains (18, 20), the

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 15

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

PAZ* domains of MpAgo and RsAgo lack the second subdomain and therefore do notpossess the PAZ pocket (24, 25, 28). While no three-dimensional structure of the binarycomplex is available for RsAgo, the guide 3= end still interacts with PAZ in the binarycomplex of MpAgo but is differently oriented (Fig. 4) (24). Formation of the extendedguide-target base-pairing during target recognition is accompanied by guide releasefrom PAZ (19, 23, 24, 26, 48). Thus, structural characteristics of PAZ may have animportant role in the recognition and binding of nucleic acid targets by pAgo proteins,for example, by influencing the kinetics of target recognition and cleavage (19, 23, 24,26, 48, 49). Interestingly, the incomplete PAZ* domain is characteristic for the long-Bclade of catalytically inactive pAgos, suggesting that the stepwise guide-target anneal-ing controlled by full-length PAZ may be more important in the active long-A pAgos,e.g., for prevention of premature nonspecific target cleavage.

All short pAgos and a small but significant fraction of long pAgos lack the PAZdomain entirely, raising the question of how nucleic acid guides can be bound by theseproteins. Structural and functional analysis of the truncated long-A AfAgo lacking thePAZ domain demonstrated that it can nevertheless bind and stabilize short guide-target duplexes (14, 15, 22). Furthermore, since the guide 3= end is released from thePAZ domain upon target recognition, PAZ may be important for protection of the guidefrom cellular nucleases until the target is found—and the protection may be lessimportant for short pAgos or truncated long pAgos that use DNA guides. However, itremains to be established whether any accessory proteins may facilitate guide andtarget binding by such pAgos. The finding of association between short pAgos and theAPAZ domain led to the suggestion that APAZ might play the role of PAZ in shortpAgos (11). Recently, it was proposed that APAZ may be instead homologous to theN-domain of full-length Agos (43). However, both proposals await experimental vali-dation.

Possible functional activities of pAgo-associated proteins. In contrast to theirability to bind nucleic acids, the endonuclease activity toward the target is not auniversal feature of pAgos: the proteins that are predicted to have the catalytic tetradin the PIWI domain are found in only the long-A clade of pAgos (�18% of all pAgos).However, the lack of catalytic activity in PIWI* does not rule out the possibility that suchpAgos are involved in processing of target nucleic acids. Indeed, our analysis confirmedand expanded previous observations on the association of catalytically inactive pAgowith proteins that might have nuclease activity (11, 12). Markedly, the number ofputative nuclease genes in operons containing catalytically inactive pAgo variants issignificantly higher than for the active pAgos.

The proteins in one branch of short pAgos have the SIR2 domain with potentialnuclease function, while many more pAgos that belong to the short and long-B cladesof catalytically inactive pAgos harbor in their operons separate proteins with predictednuclease activities. The absence of the catalytic tetrad in the PIWI* domain in suchpAgos and their tight association with proteins containing SIR2, TIR2, DUF4365 (novelRecB-like nuclease of the Mrr subfamily), RecB2, Cas4, PLD, and HNH nuclease domainssuggest that these domains might replace the intrinsic endonuclease activity of theAgo protein. Although the APAZ domain is not related to any nuclease, its strongassociation with short pAgos and the proteins with nuclease domains suggests thatAPAZ might also be involved in the nuclease activity. Thus, the functions of targetrecognition and its further processing seem to be separated, with the first stepperformed by pAgos and the second by nucleases encoded in the pAgo operons.Notably, these operons often include predicted helicases and DNA-binding proteins,suggesting their involvement in DNA processing. Future analysis of the functionalactivities of these proteins will be required to uncover the molecular details of pAgoaction.

Possible functions of pAgos in prokaryotic cells. eAgos and their small RNApartners play diverse functions, from the repression of selfish genomic elements suchas transposons to the regulation of host gene expression. Furthermore, eAgos can

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 16

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

direct changes of chromatin marks and even DNA elimination (44, 45). Compared toeAgos, experimental data about pAgo functions are very scarce, as activities of only afew proteins were characterized in vitro and even smaller numbers of proteins (TtAgoand RsAgo) were studied in their host cells in vivo. Available data suggest that pAgosmight participate in protection of prokaryotic genomes against foreign elements suchas plasmids, transposons, and phages (38, 39). However, the only direct results obtainedso far were the ability of TtAgo and RsAgo (and also PfAgo and MjAgo in theheterologous Escherichia coli system) to decrease the amount of plasmid DNA in the cell(38–41). RsAgo was also demonstrated to decrease plasmid transcription and prefer-entially associate with genomic DNA sequences corresponding to transposons and ISelements (38). While the molecular mechanisms of this specific targeting remain to beestablished, we infer that other types of pAgos discovered in our analysis may also beinvolved in host defense against certain types of genetic elements.

Intriguingly, we observed a positive, not negative, correlation between the presenceof various types of pAgos and the number of transposon copies in sequenced genomes.This observation suggests several possibilities. First, transposons may not be the maintargets of all pAgos, most of which may instead target phages and plasmids. Second,the presence of pAgos in genomes with increased transposon numbers might bebeneficial for these strains because pAgos suppress the transposon expression or theirfurther detrimental transpositions. Third, it can be speculated that pAgos might per-form a plethora of other functions in prokaryotic cells, including genetic regulation andDNA repair. Indeed, association of pAgos with multiple DNA-binding and DNA-processing proteins suggests their active involvement in DNA metabolism in prokary-otic cells.

As reported previously (11, 12) and in this work, the pattern of distribution of pAgosamong eubacterial and archaeal species does not correspond to the phylogeny of hostspecies, suggesting their spread by horizontal gene transfer. While we could not builda conserved operon structure for any group of pAgo proteins, their strong associationwith genes encoding proteins involved in DNA processing suggests their commonfunction and joint horizontal transfer. Statistical analysis of the genomic neighborhoodsof pAgos also revealed a significant link to phage resistance systems (11). The functionalroles of such associations should be investigated experimentally. Other genes spreadby HGT include various adaptation and resistance systems such as CRISPR, restriction/modification, drug resistance, and toxin/antitoxin modules that have common featuresof cell protection in the highly diverse and developing environments. While it isattractive to suggest that most pAgos use associated nucleic acid guides for sequence-specific recognition of foreign DNA targets in order to repress them, other scenarios oftheir action are possible as well. For example, pAgos together with their proteinpartners might constitute a suicidal system that is activated under stress conditions orupon viral infections and triggers processing of host DNA unless its activity is sup-pressed. Our study opens new possibilities for detailed analysis of various groups ofpAgos and associated factors, which can reveal their new cellular functions andactivities and possibly adopt pAgos as an efficient tool for genomic manipulations.

MATERIALS AND METHODSProkaryotic protein and genome databases. The set of proteins was downloaded from the NCBI

FTP site in January 2018. Altogether, the database included more than 116 million RefSeq proteins from4,364 taxons with completely sequenced and assembled genomes (“Complete Genome” and “Chromo-some” statuses) and 15,505 taxons with unassembled genomic sequences (“Scaffold” and “Contig”statuses) of Archaea and Eubacteria. The genomic sequences and genome annotations of prokaryoteswere fetched from the NCBI FTP site in January 2018.

Identification of pAgo- and APAZ-containing proteins and analysis of their domain structure.The identification of homologs of already-known pAgo- and APAZ domain-containing proteins wascarried out using the PSI-BLAST and DELTA-BLAST program from the NCBI-BLAST� package, v.2.6.0(-num_descriptions 1000 -num_alignments 1000 -evalue 10E�5 -num_iterations 5) (50). The search wasperformed with five iterations, sufficient for the full convergence of the results. For the search, sequencesof only the PIWI-MID domains of already-known pAgos (12) or APAZ domains (11) were used as queries.The domain architecture of the found proteins was analyzed by the CDD-batch program (https://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi) for the Pfam and CDD databases as well as InterProScan

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 17

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

(v.5.28-67) for the Pfam and Superfamily databases. The proteins were aligned by the MAFFT program,v.7.3 (-ep 0 – genafpair –maxiterate 1000) (51); multiple alignments were manually curated and used forthe extraction of domain borders and features.

Phylogenetic analysis. To construct a nonredundant, representative sequence set for the phyloge-netic analysis, sequences of the PIWI-MID and APAZ domains were clustered using the UCLUST 4.2program (52) with the sequence identity threshold of 90%. The longest sequence was selected torepresent each cluster. The multiple alignment of PIWI-MID domains was carried out by the MAFFTprogram, v. 7.3 (-ep 0 – genafpair –maxiterate 1000) (51). Positions including greater than �0.5 gaps wereremoved from the alignment by trimAl, v.1.4 (53). Phylogenetic analysis was performed using theFastTree program (54) with default parameters, with the WAG evolutionary model and the discretegamma model with 20 rate categories. The tree structure was validated with bootstrap analysis (n � 100).

Analysis of the composition of pAgo operons. The protein-coding sequences of 16,264 genesneighboring all pAgo genes (10 upstream and 10 downstream genes for each pAgo) were clusteredbased on sequence homology. The pairwise similarity was calculated by using the profile-profile metricwith HHblits and HHsearch from the HH-suite (55, 56), as suggested in reference 57. In contrast to themore common sequence-sequence comparison method with using all-against-all BLAST analysis, thecomparison of HMM profiles represents a more sensitive strategy for detecting distant evolutionaryrelationships among diverged proteins. For this, a profile for each sequence was constructed with HHblits(-n 1 -E 1E�3) and the uniclast30_2017_10 HMM database (http://wwwuser.gwdg.de/~compbiol/uniclust/2017_10/). Then, all-against-all profile–profile search was conducted using HHsearch (-b 1 -b1000 -z 1 -Z 1000 -E 1E�5), which provided E values similar to what BLAST does. Clustering of genes intoorthogroups was done with the MCL algorithm implemented in the clusterMaker2 plug-in of Cytoscape3.6.1 (inflation factor � 2).

The annotation of obtained 1,892 protein orthogroups was performed with InterProScan (v.5.28-67)for the Pfam and Superfamily databases. Also, we have performed the PSI-BLAST search of APAZ,RecB-like, and RecB2 domains in the orthogroup proteins. The prediction of nucleases belonging to thePD-(D/E)XK superfamily was carried out with an SVM-based approach using the standalone version of the“PDEXK recognition” program kindly provided by Mindaugas Margelevicius (58). For each pAgo operon,we then determined the orthogroups to which operon genes belonged to. If a pAgo was encoded inseveral genomes, only the genome with the largest operon was selected for final counting. Onlyorthogroups that were present in at least 5% of operons for each pAgo type (minimum three operonsof any type of pAgo operon) were kept.

Counting the transposon copy number in genomes. The transposon insertions in genomes werecounted using genomic annotations of prokaryotic strains fetched from NCBI. Only genomes withstatuses “Complete Genome” and “Chromosome” were taken into account. For each genome, wecounted the number of occurrences of keyword “insertion sequence” or “transposase” in the genomicannotation file. If there were several completed genomes for a given strain, then the median number oflog2-transformed transposon insertions was calculated. If a pAgo gene was found in different strainsof the same species (having the same taxon_id in the NCBI annotation), then the median number oftransposon insertions among all strains was taken. For the comparison of transposon contents of strainsencoding or not encoding pAgos within a genus, the additional z-scaling was performed.

SUPPLEMENTAL MATERIALSupplemental material for this article may be found at https://doi.org/10.1128/mBio

.01935-18.FIG S1, PDF file, 0.01 MB.FIG S2, PDF file, 0.04 MB.FIG S3, PDF file, 0.5 MB.TABLE S1, XLS file, 0.3 MB.TABLE S2, XLSX file, 0.3 MB.TABLE S3, XLS file, 0.03 MB.DATA SET S1, TXT file, 0.02 MB.DATA SET S2, TXT file, 0.1 MB.DATA SET S3, TXT file, 0.1 MB.

ACKNOWLEDGMENTSWe thank Mindaugas Margelevicius for the help with the prediction of PD-(D/E)XK

motifs in proteins.This work was supported by the grant of the Ministry of Education and Science of

the Russian Federation 14.W03.31.0007.

REFERENCES1. Liu J, Carmell MA, Rivas FV, Marsden CG, Thomson JM, Song J-J, Ham-

mond SM, Joshua-Tor L, Hannon GJ. 2004. Argonaute2 is the catalyticengine of mammalian RNAi. Science 305:1437–1441. https://doi.org/10.1126/science.1102513.

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 18

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

2. Miyoshi K, Tsukumo H, Nagami T, Siomi H, Siomi MC. 2005. Slicerfunction of Drosophila Argonautes and its involvement in RISC forma-tion. Genes Dev 19:2837–2848. https://doi.org/10.1101/gad.1370605.

3. Rivas FV, Tolia NH, Song JJ, Aragon JP, Liu J, Hannon GJ, Joshua-Tor L.2005. Purified Argonaute2 and an siRNA form recombinant human RISC.Nat Struct Mol Biol 12:340 –349. https://doi.org/10.1038/nsmb918.

4. Rand TA, Ginalski K, Grishin NV, Wang X. 2004. Biochemical identificationof Argonaute 2 as the sole protein required for RNA-induced silencingcomplex activity. Proc Natl Acad Sci U S A 101:14385–14389. https://doi.org/10.1073/pnas.0405913101.

5. Djuranovic S, Nahvi A, Green R. 2012. miRNA-mediated gene silencing bytranslational repression followed by mRNA deadenylation and decay.Science 336:237–240. https://doi.org/10.1126/science.1215691.

6. Behm-Ansmant I, Rehwinkel J, Doerks T, Stark A, Bork P, Izaurralde E.2006. mRNA degradation by miRNAs and GW182 requires both CCR4:NOT deadenylase and DCP1:DCP2 decapping complexes. Genes Dev20:1885–1898. https://doi.org/10.1101/gad.1424106.

7. Rehwinkel J, Behm-Ansmant I, Gatfield D, Izaurralde E. 2005. A crucialrole for GW182 and the DCP1:DCP2 decapping complex in miRNA-mediated gene silencing. RNA 11:1640 –1647. https://doi.org/10.1261/rna.2191905.

8. Kuramochi-Miyagawa S, Watanabe T, Gotoh K, Totoki Y, Toyoda A, IkawaM, Asada N, Kojima K, Yamaguchi Y, Ijiri TW, Hata K, Li E, Matsuda Y,Kimura T, Okabe M, Sakaki Y, Sasaki H, Nakano T. 2008. DNA methylationof retrotransposon genes is regulated by Piwi family members MILI andMIWI2 in murine fetal testes. Genes Dev 22:908 –917. https://doi.org/10.1101/gad.1640708.

9. Sienski G, Dönertas D, Brennecke J. 2012. Transcriptional silencing oftransposons by Piwi and maelstrom and its impact on chromatin stateand gene expression. Cell 151:964 –980. https://doi.org/10.1016/j.cell.2012.10.040.

10. Verdel A, Jia S, Gerber S, Sugiyama T, Gygi S, Grewal SI, Moazed D. 2004.RNAi-mediated targeting of heterochromatin by the RITS complex. Sci-ence 303:672– 676. https://doi.org/10.1126/science.1093686.

11. Makarova KS, Wolf YI, van der Oost J, Koonin EV. 2009. Prokaryotichomologs of Argonaute proteins are predicted to function as key com-ponents of a novel system of defense against mobile genetic elements.Biol Direct 4:29. https://doi.org/10.1186/1745-6150-4-29.

12. Swarts DC, Makarova K, Wang Y, Nakanishi K, Ketting RF, Koonin EV,Patel DJ, van der Oost J. 2014. The evolutionary journey of Argonauteproteins. Nat Struct Mol Biol 21:743–753. https://doi.org/10.1038/nsmb.2879.

13. Ma J-B, Yuan Y-R, Meister G, Pei Y, Tuschl T, Patel DJ. 2005. Structuralbasis for 5=-end-specific recognition of guide RNA by the A. fulgidus Piwiprotein. Nature 434:666 – 670. https://doi.org/10.1038/nature03514.

14. Parker JS, Roe SM, Barford D. 2004. Crystal structure of a PIWI proteinsuggests mechanisms for siRNA recognition and slicer activity. EMBO J23:4727– 4737. https://doi.org/10.1038/sj.emboj.7600488.

15. Parker JS, Roe SM, Barford D. 2005. Structural insights into mRNArecognition from a PIWI domain–siRNA guide complex. Nature 434:663– 666. https://doi.org/10.1038/nature03462.

16. Song JJ, Smith SK, Hannon GJ, Joshua-Tor L. 2004. Crystal structure ofArgonaute and its implications for RISC slicer activity. Science 305:1434 –1437. https://doi.org/10.1126/science.1102514.

17. Wang Y, Juranek S, Li H, Sheng G, Tuschl T, Patel DJ. 2008. Structureof an argonaute silencing complex with a seed-containing guide DNAand target RNA duplex. Nature 456:921–926. https://doi.org/10.1038/nature07666.

18. Wang Y, Sheng G, Juranek S, Tuschl T, Patel DJ. 2008. Structure of theguide-strand-containing argonaute silencing complex. Nature 456:209 –213. https://doi.org/10.1038/nature07315.

19. Wang Y, Juranek S, Li H, Sheng G, Wardle GS, Tuschl T, Patel DJ. 2009.Nucleation, propagation and cleavage of target RNAs in Ago silencingcomplexes. Nature 461:754 –761. https://doi.org/10.1038/nature08434.

20. Yuan Y-R, Pei Y, Ma J-B, Kuryavyi V, Zhadina M, Meister G, Chen H-Y,Dauter Z, Tuschl T, Patel DJ. 2005. Crystal structure of A. aeolicusargonaute, a site-specific DNA-guided endoribonuclease, provides in-sights into RISC-mediated mRNA cleavage. Mol Cell 19:405– 419. https://doi.org/10.1016/j.molcel.2005.07.011.

21. Yuan YR, Pei Y, Chen HY, Tuschl T, Patel DJ. 2006. A potential protein-RNA recognition event along the RISC-loading pathway from the struc-ture of A. aeolicus Argonaute with externally bound siRNA. Structure14:1557–1565. https://doi.org/10.1016/j.str.2006.08.009.

22. Parker JS, Parizotto EA, Wang M, Roe SM, Barford D. 2009. Enhancement

of the seed-target recognition step in RNA silencing by a PIWI/MIDdomain protein. Mol Cell 33:204 –214. https://doi.org/10.1016/j.molcel.2008.12.012.

23. Sheng G, Zhao H, Wang J, Rao Y, Tian W, Swarts DC, van der Oost J, PatelDJ, Wang Y. 2014. Structure-based cleavage mechanism of Thermusthermophilus Argonaute DNA guide strand-mediated DNA target cleav-age. Proc Natl Acad Sci U S A 111:652– 657. https://doi.org/10.1073/pnas.1321032111.

24. Kaya E, Doxzen KW, Knoll KR, Wilson RC, Strutt SC, Kranzusch PJ, DoudnaJA. 2016. A bacterial Argonaute with noncanonical guide RNA specificity.Proc Natl Acad Sci U S A 113:4057– 4062. https://doi.org/10.1073/pnas.1524385113.

25. Miyoshi T, Ito K, Murakami R, Uchiumi T. 2016. Structural basis for therecognition of guide RNA and target DNA heteroduplex by Argonaute.Nat Commun 7:11846. https://doi.org/10.1038/ncomms11846.

26. Doxzen KW, Doudna JA. 2017. DNA recognition by an RNA-guidedbacterial Argonaute. PLoS One 12:e0177097. https://doi.org/10.1371/journal.pone.0177097.

27. Willkomm S, Oellig CA, Zander A, Restle T, Keegan R, Grohmann D,Schneider S. 2017. Structural and mechanistic insights into an archaealDNA-guided Argonaute protein. Nat Microbiol 2:17035. https://doi.org/10.1038/nmicrobiol.2017.35.

28. Liu Y, Esyunina D, Olovnikov I, Teplova M, Kulbachinskiy A, Aravin AA,Patel DJ. 2018. Accommodation of helical imperfections in Rhodobactersphaeroides Argonaute ternary complexes with guide RNA and targetDNA. Cell Rep 24:453– 462. https://doi.org/10.1016/j.celrep.2018.06.021.

29. Elkayam E, Kuhn CD, Tocilj A, Haase AD, Greene EM, Hannon GJ, Joshua-Tor L. 2012. The structure of human argonaute-2 in complex withmiR-20a. Cell 150:100 –110. https://doi.org/10.1016/j.cell.2012.05.017.

30. Nakanishi K, Weinberg DE, Bartel DP, Patel DJ. 2012. Structure of yeastArgonaute with guide RNA. Nature 486:368 –374. https://doi.org/10.1038/nature11211.

31. Faehnle CR, Elkayam E, Haase AD, Hannon GJ, Joshua-Tor L. 2013. Themaking of a slicer: activation of human Argonaute-1. Cell Rep3:1901–1909. https://doi.org/10.1016/j.celrep.2013.05.033.

32. Schirle NT, Sheu-Gruttadauria J, MacRae IJ. 2014. Structural basis formicroRNA targeting. Science 346:608 – 613. https://doi.org/10.1126/science.1258040.

33. Matsumoto N, Nishimasu H, Sakakibara K, Nishida KM, Hirano T, IshitaniR, Siomi H, Siomi MC, Nureki O. 2016. Crystal structure of silkwormPIWI-clade Argonaute Siwi bound to piRNA. Cell 167:484 – 497.e9.https://doi.org/10.1016/j.cell.2016.09.002.

34. Park MS, Phan HD, Busch F, Hinckley SH, Brackbill JA, Wysocki VH,Nakanishi K. 2017. Human Argonaute3 has slicer activity. Nucleic AcidsRes 45:11867–11877. https://doi.org/10.1093/nar/gkx916.

35. Kwak PB, Tomari Y. 2012. The N domain of Argonaute drives duplexunwinding during RISC assembly. Nat Struct Mol Biol 19:145–151.https://doi.org/10.1038/nsmb.2232.

36. Bujnicki JM, Rychlewski L. 2001. Identification of a PD-(D/E)XK-like do-main with a novel configuration of the endonuclease active site in themethyl-directed restriction enzyme Mrr and its homologs. Gene 267:183–191. https://doi.org/10.1016/S0378-1119(01)00405-X.

37. Kinch LN, Ginalski K, Rychlewski L, Grishin NV. 2005. Identification ofnovel restriction endonuclease-like fold families among hypotheticalproteins. Nucleic Acids Res 33:3598 –3605. https://doi.org/10.1093/nar/gki676.

38. Olovnikov I, Chan K, Sachidanandam R, Newman DK, Aravin AA. 2013.Bacterial argonaute samples the transcriptome to identify foreign DNA.Mol Cell 51:594 – 605. https://doi.org/10.1016/j.molcel.2013.08.014.

39. Swarts DC, Jore MM, Westra ER, Zhu Y, Janssen JH, Snijders AP, Wang Y,Patel DJ, Berenguer J, Brouns SJJ, van der Oost J. 2014. DNA-guided DNAinterference by a prokaryotic Argonaute. Nature 507:258 –261. https://doi.org/10.1038/nature12971.

40. Swarts DC, Hegge JW, Hinojo I, Shiimori M, Ellis MA, Dumrongkulraksa J,Terns RM, Terns MP, van der Oost J. 2015. Argonaute of the archaeonPyrococcus furiosus is a DNA-guided nuclease that targets cognate DNA.Nucleic Acids Res 43:5120 –5129. https://doi.org/10.1093/nar/gkv415.

41. Zander A, Willkomm S, Ofer S, van Wolferen M, Egert L, Buchmeier S,Stockl S, Tinnefeld P, Schneider S, Klingl A, Albers SV, Werner F, Grohm-ann D. 2017. Guide-independent DNA cleavage by archaeal Argonautefrom Methanocaldococcus jannaschii. Nat Microbiol 2:17034. https://doi.org/10.1038/nmicrobiol.2017.34.

42. Hegge JW, Swarts DC, van der Oost J. 2018. Prokaryotic Argonaute

The Diverse World of Prokaryotic Argonautes ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 19

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from

proteins: novel genome-editing tools? Nat Rev Microbiol 16:5–11.https://doi.org/10.1038/nrmicro.2017.73.

43. Willkomm S, Makarova KS, Grohmann D. 2018. DNA silencing by pro-karyotic Argonaute proteins adds a new layer of defense against invad-ing nucleic acids. FEMS Microbiol Rev 42:376 –387. https://doi.org/10.1093/femsre/fuy010.

44. Juliano C, Wang J, Lin H. 2011. Uniting germline and stem cells: thefunction of Piwi proteins and the piRNA pathway in diverse organisms.Annu Rev Genet 45:447– 469. https://doi.org/10.1146/annurev-genet-110410-132541.

45. Hutvagner G, Simard MJ. 2008. Argonaute proteins: key players inRNA silencing. Nat Rev Mol Cell Biol 9:22–32. https://doi.org/10.1038/nrm2321.

46. Koonin EV. 2017. Evolution of RNA- and DNA-guided antivirus defensesystems in prokaryotes and eukaryotes: common ancestry vs conver-gence. Biol Direct 12:5. https://doi.org/10.1186/s13062-017-0177-2.

47. Swarts DC, Koehorst JJ, Westra ER, Schaap PJ, van der Oost J. 2015.Effects of Argonaute on gene expression in Thermus thermophilus. PLoSOne 10:e0124880. https://doi.org/10.1371/journal.pone.0124880.

48. Zander A, Holzmeister P, Klose D, Tinnefeld P, Grohmann D. 2014.Single-molecule FRET supports the two-state model of Argonaute ac-tion. RNA Biol 11:45–56. https://doi.org/10.4161/rna.27446.

49. Jung SR, Kim E, Hwang W, Shin S, Song JJ, Hohng S. 2013. Dynamicanchoring of the 3=-end of the guide strand controls the target disso-ciation of Argonaute-guide complex. J Am Chem Soc 135:16865–16871.https://doi.org/10.1021/ja403138d.

50. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, LipmanDJ. 1997. Gapped BLAST and PSI-BLAST: a new generation of proteindatabase search programs. Nucleic Acids Res 25:3389 –3402. [9254694]https://doi.org/10.1093/nar/25.17.3389.

51. Katoh K, Standley DM. 2013. MAFFT multiple sequence alignment soft-ware version 7: improvements in performance and usability. Mol BiolEvol 30:772–780. https://doi.org/10.1093/molbev/mst010.

52. Edgar RC. 2010. Search and clustering orders of magnitude faster thanBLAST. Bioinformatics 26:2460–2461. https://doi.org/10.1093/bioinformatics/btq461.

53. Capella-Gutierrez S, Silla-Martinez JM, Gabaldon T. 2009. trimAl: atool for automated alignment trimming in large-scale phylogeneticanalyses. Bioinformatics 25:1972–1973. https://doi.org/10.1093/bioinformatics/btp348.

54. Price MN, Dehal PS, Arkin AP. 2010. FastTree 2—approximatelymaximum-likelihood trees for large alignments. PLoS One 5:e9490.https://doi.org/10.1371/journal.pone.0009490.

55. Söding J. 2005. Protein homology detection by HMM-HMM comparison.Bioinformatics 21:951–960. https://doi.org/10.1093/bioinformatics/bti125.

56. Remmert M, Biegert A, Hauser A, Söding J. 2011. HHblits: lightning-fastiterative protein sequence searching by HMM-HMM alignment. NatMethods 9:173–175. https://doi.org/10.1038/nmeth.1818.

57. Bernardes JS, Vieira FR, Costa LM, Zaverucha G. 2015. Evaluation andimprovements of clustering algorithms for detecting remote homolo-gous protein families. BMC Bioinformatics 16:34. https://doi.org/10.1186/s12859-014-0445-4.

58. Laganeckas M, Margelevicius M, Venclovas C. 2011. Identification of newhomologs of PD-(D/E)XK nucleases by support vector machines trainedon data derived from profile-profile alignments. Nucleic Acids Res 39:1187–1196. https://doi.org/10.1093/nar/gkq958.

59. Lisitskaya L, Aravin AA, Kulbachinskiy A. DNA interference and beyond:structure and functions of prokaryotic Argonaute proteins. Nat Com-mun, in press.

Ryazansky et al. ®

November/December 2018 Volume 9 Issue 6 e01935-18 mbio.asm.org 20

on February 26, 2020 by guest

http://mbio.asm

.org/D

ownloaded from