The Architecture of Trypanosoma brucei editosomesSuzanne M. McDermotta, Jie Luob, Jason Carnesa, Jeff A. Ranishb,1, and Kenneth Stuarta,1

aCenter for Infectious Disease Research, Seattle, WA 98109; and bInstitute for Systems Biology, Seattle, WA 98109

Edited by Alberto Carlos Frasch, Universidad de San Martin and National Research Council, San Martin-C.P., Argentina, and approved August 30, 2016(received for review June 26, 2016)

Uridine insertion and deletion RNA editing generates functionalmitochondrial mRNAs in Trypanosoma brucei. Editing is catalyzed bythree distinct ∼20S editosomes that have a common set of 12 pro-teins, but are typified by mutually exclusive RNase III endonucleaseswith distinct cleavage specificities and unique partner proteins. Pre-vious studies identified a network of protein–protein interactionsamong a subset of common editosome proteins, but interactionsamong the endonucleases and their partner proteins, and their in-teractions with common subunits were not identified. Here, chem-ical cross-linking and mass spectrometry, comparative structuralmodeling, and genetic and biochemical analyses were used to definethe molecular architecture and subunit organization of purified edi-tosomes. We identified intra- and interprotein cross-links for all edi-tosome subunits that are fully consistent with editosome proteinstructures and previously identified interactions, which we validatedby genetic and biochemical studies. The results were used to create ahighly detailed map of editosome protein domain proximities, lead-ing to identification of molecular interactions between subunits, in-sights into the functions of noncatalytic editosome proteins, and aglobal understanding of editosome architecture.

cross-linking | proteomics | Trypanosoma brucei | RNA editing | editosome

The kinetoplastid Trypanosoma brucei is the causative agent ofhuman African trypanosomiasis, which is transmitted by the

tsetse fly and is a health threat to millions of people in sub-Saharan Africa. Kinetoplastids are named for their distinctivemitochondrial DNA network, known as the kinetoplast DNA,which is composed of interlocked DNA maxicircles and minicircles(1, 2). Several identical maxicircles encode two ribosomal RNAs(rRNAs) and messenger RNAs (mRNAs) for 18 mitochondrialproteins, 12 of which undergo posttranscriptional RNA editing (3–9). RNA editing was first described in kinetoplastids (6, 8, 9), whereit is essential (10) and involves insertion and deletion of uridines(Us) to generate translatable mitochondrial transcripts. The se-quence information for editing is specified by ∼60-nucleotide guideRNAs (gRNAs) that are encoded in thousands of heterogeneousminicircles (11–17). Editing occurs by rounds of coordinated cata-lytic steps that require a number of different enzymes: cleavage ofthe mRNA substrate by endonucleases, addition of Us by 3′ ter-minal uridylytransferase (TUTase) or removal of Us by U-specific 3′exonuclease (exoUase) at insertion or deletion editing sites, re-spectively, and rejoining of mRNA fragments by RNA ligases. Theenzymes that catalyze RNA editing in T. brucei are in ∼20S edito-some complexes that also contain proteins with no known catalyticfunctions (16, 18–28).There are three similar versions of ∼20S editosomes that have a

common set of 12 proteins (SI Appendix, Fig. S1 and Table S1).However, they are compositionally and functionally distinct in thateach contains a different single RNase III endonuclease with auniquely associated specific partner protein, and a different edit-ing site cleavage specificity (22–26, 29) (SI Appendix, Fig. S1 andTable S1). One complex contains the kinetoplastid RNA edit-ing endonuclease (KREN)1/kinetoplastid RNA editing protein(KREP)B8 protein pair and kinetoplastid RNA editing exo-nuclease (KREX)1 and cleaves deletion editing sites. The othertwo complexes contain the KREN2/KREPB7 or KREN3/KREPB6protein pairs and cleave insertion sites, albeit with different pref-erences. Direct interactions between the endonucleases and their

partner proteins with each other, and with common editosomeproteins, have not yet been identified.The common set of proteins contains two related proteins,

KREPB4 and KREPB5, each of which has a U1-like zinc-fingermotif, a Pumilio/fem-3 mRNA binding factor (PUF) motif, and adegenerate RNase III domain (30, 31). Mutation of KREPB4and KREPB5 residues that are universally conserved in allknown catalytic RNase III domains has no effect on editing or invitro cleavage of editing sites (30), whereas equivalent muta-tions in the RNase III domains of the KREN1, KREN2, orKREN3 endonucleases eliminate in vivo editing and in vitrocleavage of editing sites (22). Because all characterized RNaseIIIs function as dimers to cleave double-stranded RNA (dsRNA)(32, 33), and because the endonucleases are each present as asingle copy per editosome (23), we have therefore hypothesizedthat the noncatalytic RNase III domain of KREPB4 and KREPB5forms a heterodimeric RNase III active site with the editing en-donucleases. Both KREPB4 and KREPB5 are also required forstructural integrity of editosomes and are essential for cell growthand editing in vivo. Knockdown of KREPB4 expression results inloss of ∼20S editosomes and editosome proteins (34), whereas lossof KREPB5 results in a dramatic loss of editosomes and edito-some components in bloodstream form (BF) cells (35), withretention of components and of complexes with smaller S valuesin insect, procyclic form (PF) cells (36). No interactions forKREPB4 were identified in previous studies, and KREPB5 wasshown only to interact with one protein, KREPA3 (37) (SI Ap-pendix, Fig. S1). Thus, although KREPB4 and KREPB5 are es-sential for editosome integrity, very little is known about howthey interact with other editosome proteins.Common editosome proteins also include the related KREPA1 to

-A6 proteins, which have C-terminal oligonucleotide/oligosaccharide

Significance

Trypanosoma brucei is a deadly kinetoplastid parasite thatcauses the human and veterinarian diseases African sleepingsickness and nagana. We combine chemical cross-linking andmass spectrometry, structural modeling, genetics, and bio-chemistry to define the global architecture of RNA editing“editosome” complexes in these parasites. Editosomes areunique to kinetoplastids, which also include Trypanosoma cruziand Leishmania parasites that cause Chagas disease andLeishmaniases, respectively. Editosomes are essential for par-asite viability and are promising drug targets. This work cre-ates a comprehensive, highly detailed map of editosomeprotein proximities and interactions, and furthers our under-standing of editosome protein function in T. brucei and otherkinetoplastids, thus aiding the potential development of newtherapeutics for the treatment of several different parasiticdiseases.

Author contributions: S.M.M., J.L., J.C., J.A.R., and K.S. designed research; S.M.M., J.L., andJ.C. performed research; S.M.M., J.L., and J.C. analyzed data; and S.M.M., J.L., J.C., J.A.R.,and K.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. Email: jranish@systemsbiology.org or ken.stuart@cidresearch.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1610177113/-/DCSupplemental.

E6476–E6485 | PNAS | Published online October 5, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1610177113

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

binding (OB)-fold motifs (31). KREPA1, -A2, and -A3 also eachhave two C2H2 zinc fingers. A combination of affinity purification,fractionation, yeast two-hybrid, and coimmunoprecipitation (co-IP)experiments previously showed that KREPA1 and KREPA2 arecomponents of two different, stable heterotrimeric subcomplexesthat are present in all editosomes and mapped a number of protein–protein interactions within these subcomplexes (SI Appendix, Fig. S1and Table S2). KREPA2 directly interacts with both the exo-nuclease KREX2 and the kinetoplastid RNA editing ligase (KREL)1to form a ∼5–10S deletion subcomplex that catalyzes precleaveddeletion editing (28, 37–39) (SI Appendix, Table S2). Furthermore,KREPA2 can enhance the enzymatic activity of KREL1 (39). In asimilar fashion, KREPA1 directly interacts with the TUTaseKRET2 and the ligase KREL2 to form a ∼5–10S insertion sub-complex that catalyzes precleaved insertion editing (28, 37, 38) (SIAppendix, Table S2), and can enhance the enzymatic activity ofKRET2 (28).The KREPA proteins KREPA3, -A4, and -A6 are each es-

sential for the structural integrity of editosomes. Knockdown ofKREPA3 results in a dramatic loss of editosome complexes inBF cells, whereas smaller complexes accumulate in PF cells (36,40), and mutational analysis in BF cells showed that theKREPA3 OB-fold is vital for the interaction of KREPA3 withthe editosome (40). PF knockdown of KREPA4 or KREPA6also results in loss of editosome complexes (41–43). Yeast two-hybrid analyses and reconstitution experiments further indicatedthat KREPA3, -A4, and -A6 are part of an OB-fold network thatthat physically links the heterotrimeric insertion and deletionsubcomplexes within ∼20S editosomes. KREPA6 can directlyinteract with the KREPA1 and KREPA2 OB-folds, KREPA4,and with the KREPA3 OB-fold (37, 38) (SI Appendix, Fig. S1).The KREPA3 OB-fold also interacts with the KREPA2 OB-fold(37). The remaining KREPA family protein, KREPA5, was morerecently identified as a common editosome subunit, and thus itwas not included in previous analyses and has not yet been func-tionally characterized. A number of other proteins, KREPB9,KREPB10, and MEAT1, were also identified more recently andhave been shown to copurify with the majority of proteins in ∼20editosomes (27, 44).Thus, significant progress has been made to unravel the

mechanism of RNA editing in T. brucei, the editosome proteininventory, and the functions of a number of individual editosomeproteins. However, several key questions remain outstanding.Primary among these are a global understanding of the molec-ular interactions within editosomes, in addition to interactionsbetween editosomes and other mitochondrial RNA processingcomplexes. Previous work described a total of 11 direct inter-actions between 10 common proteins, but these predominantlyencompass the insertion and deletion subcomplexes, and theKREPA proteins linking them (37–39, 45–47) (SI Appendix, Fig.S1 and Table S2). Other associations, for example between theendonucleases and the insertion subcomplex, and between en-donuclease partner KREPB proteins and the deletion sub-complex, have been inferred (29), but no direct interactions havebeen shown.Knowledge of molecular interactions within editosomes is

crucial to fully understand how editosomes discriminate amongthousands of editing sites in multiple transcripts, and how themultiple catalytic steps of RNA editing are coordinated. Endo-nuclease interactions with other editosome proteins are partic-ularly key, as substrate recognition and cleavage by theseproteins are likely a major point of regulation during editing, andin differential editing between parasite life-cycle stages. To ap-proach this problem, we have used chemical cross-linking com-bined with mass spectrometry (CXMS), together with geneticand biochemical analyses, to determine molecular proximitiesand interactions within and between editosome components.Using this approach we identified numerous intra- and inter-protein cross-links for all editosome components, including thosefor which there was no prior information, such as the endonu-cleases KREN1 to -3, the endonuclease partner proteins

KREPB6 to -8, KREPB4, KREPA5, and KREPB10. All cross-links are fully consistent with editosome protein structures, andwith all previously described protein–protein interactions be-tween editosome subunits. Our work leads to a global, highlydetailed map of editosome protein domain proximities and anunderstanding of editosome protein interactions and functions,particularly between the editing endonucleases, their associatedpartner proteins, and other common editing proteins.

ResultsCXMS of Editosomes. Editosomes were affinity purified from PFT. brucei using tandem affinity (TAP)-tagged KREN1 and alsoTAP-tagged KREPB5 (SI Appendix, Fig. S2). Western analysis of10–30% (vol/vol) glycerol gradient fractions of KREN1– andKREPB5–TAP eluates revealed that we had isolated ∼20Scomplexes, and that ∼5–10S subcomplexes were also present inour preparations (SI Appendix, Fig. S2A). The subcomplexes inKREPB5–TAP eluates predominantly contained KREPA2 andKREL1 and were deletion subcomplexes. In contrast, the sub-complexes in KREN1–TAP eluates primarily contained insertionsubcomplexes, as seen by the greater abundance of KREPA1compared with KREPA2 and KREL1. These observations areconsistent with previous studies that indicated that KREPB5 ismost stably associated with the deletion subcomplex (30) and theendonucleases primarily associate with the insertion subcomplex(23, 29). The compositions of the complexes and subcomplexes inthe two different purifications provided a rationale for subsequentanalysis of both KREN1–TAP and KREPB5–TAP eluates.We used CXMS to identify pairs of lysine and N-terminal

residues that are in spatial proximity to each other in bothKREN1–TAP and KREPB5–TAP preparations (Fig. 1 and SIAppendix, Fig. S3). This approach identifies the specific cross-linked amino acid residues and uses distance constraints of thecross-linker to define the residues’ proximity, thus revealingpotential domain–domain interactions (48). Each ∼20S edito-some component, except KREPA4, contains many lysine resi-dues distributed throughout the polypeptides (SI Appendix,Table S3), making editosomes good substrates for this tech-nique. The TAP-tag purified complexes were cross-linked usingthe homo-bifunctional, amine reactive cross-linking reagentbis(sulfosuccinmidyl)suberate (BS3). The cross-linked sampleswere digested with trypsin, analyzed by mass spectrometry,and a database containing all known editosome proteins wassearched to identify cross-linked peptides. KREN1–TAP andKREPB5–TAP complexes were each cross-linked with both2 mM and 5 mM BS3 (SI Appendix, Fig. S2 B and C), and thereactions were independently analyzed for protein cross-links.Results from both BS3 concentrations were then combined foreach complex (SI Appendix, Table S4 and Datasets S1 and S2).Confidently identified cross-links (Datasets S1 and S2) wereused to assemble site-specific linkage maps of all of the cross-linked residues between and within the editosome proteins(Fig. 1 and SI Appendix, Fig. S3). From the KREN1–TAP pu-rified editosome complexes we identified 81 unique interlinks(cross-links between different proteins) for 30 different proteinpairs, and 54 unique intralinks (cross-links within the sameprotein) for 8 individual proteins (SI Appendix, Table S5). Fromthe KREPB5–TAP purified complexes, we identified 278 in-terlinks between 68 protein pairs, and 247 intralinks within 19proteins (SI Appendix, Table S5). Taken together, these com-bined datasets yielded a total of 345 unique interlinks between77 protein pairs, and 279 unique intralinks within 19 proteins(Fig. 1 and SI Appendix, Fig. S3 and Table S5).To validate our cross-linking approach, we mapped cross-

linked lysine pairs onto known 3D crystal structures (SI Appen-dix, Fig. S4 and Table S6) (45–47, 49–51) and comparativemodels (SI Appendix, Figs. S5 and S6 and Table S7, and DatasetS3) of editosome proteins. Comparative models for parts of allsubunits were built using ModPipe (52), HHPred (53), Modeler(54), and Swiss-model (55), which generally identified templatessimilar to those previously used for comparative modeling for a

McDermott et al. PNAS | Published online October 5, 2016 | E6477

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

number of proteins (Dataset S3) (56). The BS3 cross-linker has alinker arm of 11.4 Å when fully extended and can cross-link tworesidues whose Cα atoms are up to 30 Å apart (57). We mea-sured the distances between the Cα atoms in the cross-linkedlysine residues for 59 intralinks, and one interlink that could bemapped onto known structures and structural models (SI Ap-pendix, Figs. S4–S6 and Tables S6 and S7). Of the 60 cross-linksanalyzed in this way, 56 have Cα-Cα distances <30 Å (SI Ap-pendix, Fig. S7) and include all cross-links that were mappedonto known crystal structures. Taken together, these results in-dicate that our cross-linking data provide reliable restraints forvisualizing and modeling the subunit architecture of T. bruceieditosomes.

Architecture of the Heterotrimeric Insertion and Deletion Dubcomplexes.We compared cross-links identified here by CXMS with previouslypublished protein–protein interaction data to further validate ourapproach. All previously described interactions between editosomesubunits are also found in our cross-linking data (SI Appendix,Table S8); this is particularly evident within the heterotrimeric in-sertion and deletion subcomplexes (Fig. 2 A and B and SI Appendix,Table S2). Among the deletion subcomplex subunits KREPA2,

KREL1, and KREX2, we detected many cross-links betweenKREPA2 and KREL1, predominantly involving the KREPA2 re-gion that is flanked by the two zinc-finger motifs and the KREL1C-terminal and ligase domains (Fig. 2 A and B). These cross-linksreinforce the importance of the KREL1 C-terminal and ligasedomains, which were previously shown to be important for KREL1interaction with KREPA2, and integration into editosomes (37, 39,58). We observed extensive cross-linking between KREPA2 andKREX2, primarily between the same KREPA2 region and theN-terminal exonuclease domain of KREX2 (Fig. 2 A and B).We also identified cross-links between KREL1 and KREX2that further support the conclusion that KREPA2, KREL1, andKREX2 form a stable subcomplex. Among the insertion sub-complex subunits KREPA1, KREL2, and KRET2, we foundthat KREPA1 was extensively cross-linked to both KRET2 andKREL2, predominantly involving the KREPA1 region flankedby the two zinc-finger motifs (Fig. 2 A and B). Cross-links werealso identified between KRET2 and KREL2, providing furtherevidence that these three subunits form a tightly interlinkedsubcomplex. We did not observe any cross-links involving theN-terminal 200 amino acids of KREPA1 with other subunits,consistent with previous reconstitution studies that showed that

Fig. 1. (A) BS3 interprotein cross-linking map of editosome complexes. Domains are highlighted as indicated based on previous studies or bioinformaticspredictions. Black dots distributed across proteins indicate the positions of lysine residues. Cross-links observed in the KREPB5–TAP complexes are shown asred lines, KREN1–TAP complexes as gray lines, and both KREPB5–TAP and KREN1–TAP complexes as purple lines. Black dotted boxes outline the trimericdeletion (KREX2, KREPA2, and KREL1) and insertion (KRET2, KREPA1, and KREL2) subcomplexes. Colored boxes outline the endonuclease proteins with theirassociated proteins that are unique to each of the three types of 20S editosome. Green represents KREN3 and KREPB6; red represents KREN2 and KREPB7; andblue represents KREN1, KREPB8, and KREX1. (B) Network diagrams of all interprotein cross-links within editosome complexes. Interlinks involving KREN1,KREN2, and KREN3 have been separated for clarity. Network edge width is proportional to the number of interprotein cross-links.

E6478 | www.pnas.org/cgi/doi/10.1073/pnas.1610177113 McDermott et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

the N-terminal ∼200 amino acids of KREPA1 are dispensablefor interaction with KREL2 and KRET2 (37).Taken together, these results are completely consistent with

the presence of the two stable heterotrimeric subcomplexes, andwith previous protein–protein interaction studies (SI Appendix,Fig. S1 and Table S2). As indicated by cross-linking here (Fig.2A), and supported by previous yeast two-hybrid analysis (SIAppendix, Table S2), the region flanked by the two zinc-fingermotifs in both KREPA1 and KREPA2 is a particularly importanttopological feature for heterotrimeric subcomplex interactions ofKREPA1 and KREPA2. This region, which we now term theanchor domain, does not possess obvious homology to anyknown structural motifs, but appears to be responsible for unit-ing KREL1 and KREX2 in the deletion subcomplex or KREL2and KRET2 in the insertion subcomplex.In addition, we also identified cross-links between the in-

sertion and deletion subcomplexes, albeit at a lower frequencythan cross-links within the individual subcomplexes (Fig. 2B). Inparticular, we observed cross-links between KREL1 with allthree insertion subcomplex components, and between KREL2with all three deletion subcomplex components. This observationindicates that the insertion and deletion subcomplexes are inclose physical proximity, and advances our understanding ofheterotrimeric subcomplex interactions beyond the previouslydescribed data.

“A” Protein Network That Links the Insertion and Deletion Subcomplexes.A number of editosome proteins, including KREPB4, KREPA3,and KREPA6, are thought to have essential roles in the mainte-nance of editosome integrity. In accordance with this, we detectedcross-links between each of KREPB4, KREPA3, and KREPA6 withboth insertion and deletion subcomplex components, indicating thatthese proteins are in proximity to, and may physically bridge, theheterotrimeric subcomplexes (Fig. 2B). We identified cross-linksbetween KREPB4 with KREPA2, KREPA1, KRET2, and KREL2.We further identified cross-links between KREPA3 with KREPA1and KREPA2, whereas KREPA6 cross-linked to KREPA2,KREX2, KREPA1, and KRET2 (Fig. 2B). KREPA5 also formedcross-links to both KREPA1 and KREPA2, providing evidence forthe position of KREPA5 within editosomes (Fig. 2B). We also ob-served cross-links between KREPA3 and KREPA6, and KREPA6,KREPA5, and KREPB4 also all cross-linked to each other (Fig.2B), signifying that KREPB4 and the “KREPA” proteins togetherform a network that links the insertion and deletion subcomplexes.We found comparatively few cross-links involving KREPA4,

which could be because of the small number of lysine residues inthis protein (SI Appendix, Table S3). The majority of cross-linksthat we observed for KREPA4 were to KREPB4, and we alsofound single cross-links between KREPA4 and KREPA6 (Fig. 2 Band C) and KRET2 (Fig. 2B). Yeast two-hybrid analysis hadpreviously indicated that KREPA4 directly interacts withKREPA6 (SI Appendix, Fig. S1) (37). Together with this previousdata, our CXMS data reinforces the conclusion that KREPA4 isalso part of the KREPB4/KREPA protein network that bridgesthe heterotrimeric subcomplexes.Previous in vitro and structural studies suggested that the in-

teractions between the KREPA proteins are mediated via theirOB-folds (37, 45–47, 49, 59). Indeed, here we identified aninterprotein cross-link that maps onto structural models of aKREPA6–KREPA3 OB-fold dimer (SI Appendix, Fig. S4).Furthermore, our data showed cross-links between the KREPA1OB-fold domain and the OB-folds of KREPA3, -A6, and -A5(Fig. 2C). However, cross-links between KREPA2 to otherKREPA proteins predominantly involved a region in KREPA2just N-terminal to the OB-fold. The previously described inter-actions of the OB-fold–containing proteins, which did not in-clude KREPA5, led to the suggestion of a “five OB-fold center”in the core of the editosome that might perform a critical scaf-fold function (45–47). In this model, KREPA6 and the OB-foldsof three other KREPA proteins are arranged and interact as a“shifted heterotetramer,” whereas another OB-fold interactswith KREPA6 in a different, as yet unknown, manner. Here, wehave identified cross-links between all six of the KREPA pro-teins, including KREPA5, most of which are within or are inproximity to the OB-fold domains of these proteins (Fig. 2C).This work therefore suggests a different model for an OB-foldinteraction network that incorporates KREPA5, which we pre-dict might also be essential for editosome integrity.

Endonuclease Association with Other Editosome Proteins. Directinteractions between the endonucleases and proteins commonto all editosomes, in addition to their partner proteins, are alikely point of regulation during editing but have not yet beendescribed. Using CXMS, we identified a number of interlinksbetween each of the three editing endonucleases and othereditosome proteins (Fig. 3 A and B). The KREN1–TAPpreparation isolated only KREN1 editosomes, in which wedetected 28 interlinks involving KREN1. The KREPB5–TAPpreparation isolated all editosome complexes, in which wedetected 21, 28, and 15 interlinks involving KREN1, KREN2,and KREN3, respectively. These cross-links are consistent withthe presence of one type of editing endonuclease per complex(22, 23), as we did not observe cross-linking between KREN1,KREN2, and KREN3 (Fig. 3 A and B).In agreement with previous studies (23, 29), the majority of

cross-links involving all endonucleases were to the insertionsubcomplex subunits KREPA1 and KRET2 (Fig. 3 A and B).

Fig. 2. CXMS results for heterotrimeric insertion and deletion subcomplexesin editosomes. (A) Cross-linking maps for the identified interprotein cross-links within the deletion subcomplex (Upper), and the insertion subcomplex(Lower). Anchor, anchor domain in KREPA1 and KREPA2; CTD, C-terminaldomain. (B) Network of interprotein cross-links within the insertion anddeletion subcomplexes, and those that bridge the insertion and deletionsubcomplexes. (C) A cross-linking map of interprotein cross-links betweenthe KREPA proteins.

McDermott et al. PNAS | Published online October 5, 2016 | E6479

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

The C-terminal domains of all three endonucleases in particularare in proximity to the KREPA1 OB-fold, and the Poly(A) Po-lymerase (PAP) catalytic domain of KRET2 (Fig. 3B). We ad-ditionally identified a small number of cross-links between theendonucleases and deletion subcomplex components, indicat-ing that the endonucleases can bridge the heterotrimeric sub-complexes to some extent (Fig. 3 A and B). KREN1 and KREN2also cross-linked to the OB-fold of KREPA3, whereas all threeendonucleases formed cross-links with KREPA6 and KREPB4,which also bridge the heterotrimeric subcomplexes (Fig. 3 Aand B).The proximity of KREPB4 to all endonucleases is intriguing

given that KREPB4 is one of two editosome proteins, in additionto KREPB5, previously shown to contain a degenerate non-catalytic RNase III domain (30, 31). We detected cross-links, andtherefore proximity, between the RNase III domains of each ofthe endonucleases, and the region flanked by the U1-like zincfinger and RNase III domain, and the C-terminal domain inKREPB4 (Fig. 3 A and B). Thus, equivalent domains of thedifferent endonucleases are in proximity to the same regions ofKREPB4, indicating that each of the endonucleases interactssimilarly with KREPB4. We identified only a single cross-linkbetween KREPB5 and an endonuclease, KREN2, suggestingthat KREPB5 is not a major partner of the editing endonucle-ases in PF cells (Fig. 3 A and B).

Endonuclease Association with KREPB6, -B7, -B8, and KREX1. In ac-cordance with the unique association of KREN1, KREN2, andKREN3 with KREPB8, KREPB7, and KREPB6, respectively,we identified a small number of cross-links specific to theseprotein pairs (Fig. 3 A and B). The KREN2 or KREN3 RNaseIII domains cross-linked with the C-terminal domains ofKREPB7 or KREPB6, respectively, whereas the KREN1 U1-likezinc finger cross-linked with the N terminus of KREPB8 (Fig.3B). We did not observe cross-linking between noncognatepartner protein pairs: for example, between KREN1 andKREPB6 or KREPB7. This observation provides evidence forproximity between the endonucleases and their uniquely associ-ating partner proteins. Furthermore, we identified degenerateRNase III domains in KREPB6, KREPB7, KREPB8, and ad-ditionally KREPB9 and KREPB10 (Fig. 3C, SI Appendix, Fig.S8, and Dataset S3). Intralinks within KREPB8 and KREPB10are consistent with the modeled structures of the newly identifieddomains (SI Appendix, Fig. S5) and with previous modeling ofeditosome proteins that also detected RNase III domains inKREPB6 and KREPB7 (56). These degenerate RNase III do-mains are therefore also candidates for functioning as intermo-lecular heterodimers with their partner endonucleases, whichrequires further validation by alternative approaches.Consistent with the unique association of KREX1 with KREN1

editosomes (22), the KREN1 C-terminal domain cross-linkedextensively with KREX1 (Fig. 3 B and D), providing evidence fordirect proximity and interaction between these two proteins.Similarly to KREN1, KREX1 cross-links primarily with the in-sertion subcomplex components KREPA1 and KRET2 (Fig.3D). We also identified smaller numbers of cross-links betweenKREX1 and deletion subcomplex components KREX2, KREPA2,and KREL1, in addition to KREPA5 and KREPA6, indicatingthat like KREN1, KREX1 can also bridge the insertion anddeletion subcomplexes (Fig. 3D). Most KREX1 cross-links toother editosome proteins, including KREN1, involve the 5′-3′exonuclease domain (Fig. 3D).

Experimental Validation of Endonuclease Protein ProximitiesIdentified by CXMS. The proximity data obtained by CXMS pro-vides experimentally testable predictions about the protein–protein interactions that we sought to interrogate. To validatethe proximities of different endonuclease domains with othereditosome proteins, we first generated chimeric endonucleasesthat were expressed in PF cells (Fig. 4 A and B). ChimericKREN2-N1 fuses the KREN1 C-terminal domain to the N-ter-minal portion of KREN2 that contains the U1-like zinc finger,RNase III domain, and the PUF motif. KREN1-N2 is the re-ciprocal chimera (Fig. 4A). These chimeric endonucleases haveC-terminal TAP-tags that were expressed (Fig. 4B) and used toisolate the resultant complexes. Western analysis of KREN2-N1–TAP complexes revealed robust signals for KREPA1, KREPA2,KREL1, and KREPA3 in relative amounts similar to the positivecontrol 20S complexes, indicating that the KREN2-N1 chimeracould assemble into editosome complexes (Fig. 4C). Signals forall four proteins were much weaker following purification withKREN1-N2–TAP, and their relative abundances were alteredsuch that KREPA1 was the predominant protein present in thesample (Fig. 4C). This finding suggests that KREN1-N2 maydisproportionately interact with editosome subcomplexes. Massspectrometric analysis of these complexes supported this, as onlypeptides corresponding to KREPA1, KREPA3, and chimericdomains of KREN1 and KREN2 were detected across two dif-ferent experiments (SI Appendix, Tables S9–S13). Mass spec-trometric analysis of the complexes isolated by KREN2-N1–TAPpurification revealed peptides corresponding to chimeric do-mains of KREN1 and KREN2, and to all proteins previouslydetected in KREN2–TAP editosomes (SI Appendix, Tables S9–S13), including the uniquely associating partner proteinKREPB7. As for KREN2–TAP editosomes, peptides corre-sponding to KREPB6 or KREPB8 were not identified. Addi-tionally, a number of peptides corresponding to KREX1 were

Fig. 3. (A) Network diagrams and (B) interprotein cross-linking maps ofKREN1, KREN2, and KREN3 first neighbors. (C) Comparative models of RNaseIII domains in KREPB6, KREPB7, and KREPB8. The Campylobacter jejuni RNaseIII nuclease domain (PDB ID code 3O2R) was used as a template for com-parative modeling of the RNase III domains. Modeled structures are shown inmagenta, and template RNase III domain dimer is shown in gray. Dashedlines denote large loop regions (Dataset S3). (D) Network diagram andinterprotein cross-linking map of KREX1 first neighbors.

E6480 | www.pnas.org/cgi/doi/10.1073/pnas.1610177113 McDermott et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

also detected. Taken together, these results fully align with theCXMS data, and show that the C-terminal domain of KREN1 issufficient for KREX1, but not KREPB8 interaction in KREN1editosomes, whereas the N-terminal portion of KREN2 thatcontains the zinc finger, RNase III, and PUF motifs is sufficientfor interaction with KREPB7.To test the functional consequences of exclusive expression of

the chimeric KREN2-N1 endonuclease, we inserted the codingsequence into the β-tubulin locus of BF KREN1 and KREN2conditional null (CN) cells (24, 26). In these cell lines bothendogenous alleles of either KREN1 or KREN2 have beendeleted, and a tetracycline (tet)-regulatable WT KREN1 orKREN2 allele has been inserted into the rRNA locus. Removal oftet leads to loss of WT KREN expression from the rRNA locus,and hence exclusive expression of the chimeric allele from theβ-tubulin locus. Interestingly expression of KREN2-N1 cancomplement for the loss of KREN2 expression, but not for theloss of KREN1 (Fig. 4D). This finding indicates that KREN2-N1retains KREN2, but not KREN1, function, which is thereforespecified in the portion of KREN2 that contains the U1-like zincfinger, RNase III domain, and PUF motif. This result is poten-tially via the ability of this region to associate specifically withKREPB7, but not KREPB8, as it has been shown that thepresence of KREPB7 is required for KREN2 function (29). Theadditional recruitment of KREX1 by the KREN1 C terminusdoes not interfere with KREN2 activity, nor is it sufficient forKREN1 function.

Differential Association of KREPB4 and KREPB5 with EndonucleasePartner Proteins and Other Editosome Proteins. KREPB4 andKREPB5 have different patterns of cross-linking within edito-somes, suggesting that they are in proximity to different proteinsand are functionally distinct, despite sharing common proteindomains (31). For example, we identified cross-links betweenKREPB4, but not KREPB5, with the endonuclease partnerproteins KREPB6 and KREPB7 (Fig. 5A). The C-terminal do-mains of KREPB6 and KREPB7 form cross-links with the same

regions of KREPB4 as the endonucleases (Fig. 5B), providingfurther evidence that KREPB4, but not KREPB5, is a majorinteraction partner of the editing endonucleases in PFs.Like the endonucleases, KREPB4 forms cross-links with both

insertion (KRET2, KREPA1, and KREL2) and deletion(KREPA2) subcomplex components, whereas KREPB5 formscross-links to KREPA2 (Fig. 5). We did not observe any cross-links between KREPB5 and insertion subcomplex components.Considerable cross-linking between KREPB5 and KREPA2 isconsistent with the preferential association of KREPB5 with thedeletion subcomplex, and was expected given the results offractionation of the KREPB5–TAP complexes used for CXMS(SI Appendix, Fig. S2A).We also identified differential cross-linking between KREPB4

and KREPB5 and the remaining KREPA proteins, with theexception of KREPA6 (Fig. 5). KREPB4 cross-linked to bothKREPA4 and KREPA5 whereas, consistent with previousstudies (SI Appendix, Fig. S1) (36, 37), KREPB5 formed cross-links with the OB-fold of KREPA3 (Fig. 5). The C-terminaldomains of both KREPB4 and KREPB5 are the major sites ofcross-linking to the KREPA proteins (Fig. 5B), includingKREPA1 and KREPA2, suggesting that they are pivotal sitesfor interaction with OB-fold domain proteins in editosomes. TheC-terminal domains of KREPB4 and KREPB5 do not possessdetectable homology to known protein motifs, but a recentrandom mutagenesis screen for amino acid substitutions thatinhibit KREPB5 in BF cells identified two C-terminal substitu-tions that resulted in inhibition of cell growth, RNA editing, andloss of 20S editosome complexes (60), underscoring the impor-tance of this domain for KREPB5 function. Intriguingly, theseC-terminal substitutions, in addition to others in the RNase IIIdomain of KREPB5, did not have the same inhibitory effects onRNA editing in PF parasites (60), despite the fact that KREPB5is essential in both life-cycle stages for cell growth andRNA editing.

Experimental Validation of KREPB5 Protein Proximities Identified byCXMS. To further analyze the in vivo function of the KREPB5C-terminal domain in both BF and PF, and validate its interac-tions with KREPA proteins, we generated a truncation mutant ofKREPB5 that contains the zinc-finger motif, the degenerateRNase III domain, and the PUF motif, but lacks the entireC-terminal region (Fig. 6A). The mutant allele was tagged with aV5-epitope tag and inserted into the β-tubulin locus of BF and PFKREPB5 CN cells (35, 36). Exclusive expression of truncatedKREPB5 in the absence of tet resulted in the cessation of growthof both BF and PF cells, indicating the truncated protein couldnot complement for the loss of KREPB5 in either life-cyclestage, whereas exclusive expression of V5-tagged WT KREPB5resulted in normal growth of both BF and PF cells (Fig. 6B). Theeffects of exclusive expression of C-terminally truncatedKREPB5 on BF and PF RNA editing in vivo were assessed byquantitative RT-PCR (RT-qPCR) (Fig. 6C). The abundance ofpre-edited, edited, and never-edited mitochondrial mRNAs weredetermined in BF and PF cells that were exclusively expressingWT or truncated mutant alleles, relative to the correspondingcells in which the WT tet-regulatable KREPB5 allele was alsoexpressed. Exclusive expression of WT KREPB5 did not alterthe relative levels of mitochondrial mRNAs, whereas exclusiveexpression of the C-terminal truncation mutant resulted in dra-matic reductions in the relative levels of all of the edited tran-scripts analyzed in both BF and PF cells, similar to those observedupon loss of KREPB5 (36). We observed a reduction in BF editedmRNAs of 82% or more, and the levels of edited CYb, COII,COIII, MURF2, and ND3 were below detection. In PF they werereduced by between 83% (ND7) and 97% (COIII), and editedND3 was not detected (Fig. 6C and SI Appendix, Fig. S9A). Thisfinding indicates that the KREPB5 C terminus is essential forRNA editing in both life-cycle stages.The effects of KREPB5 C-terminal truncation on editosomes

in BF and PF were also evaluated by IP of V5-tagged truncated

Fig. 4. (A) Diagram showing U1-like zinc finger, RNase III, and PUF motifs ofKREN1 and KREN2 and the C-terminal regions exchanged in the chimericKREN1-N2 and KREN2-N1 proteins. (B) Expressed TAP-tagged chimeric en-donucleases were detected using PAP antibody, which is specific to theProtein A-tagged component of the tag. (C) Western analysis of calmodulinbinding eluates isolated via TAP-tagged chimeric proteins using themonoclonal antibody mixture to KREPA1, KREPA2, KREL1, and KREPA3.The positive 20S control is from purified fractionated mitochondria.(D) Cumulative growth of cells constitutively expressing the KREN2-N1 chi-meric endonuclease in BF KREN1 or KREN2 CN cells in which the tet-regu-latable WT KREN1 or KREN2 allele was also expressed (E) in the presence oftet, or repressed (R) in the absence of tet.

McDermott et al. PNAS | Published online October 5, 2016 | E6481

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

or WT KREPB5, and by glycerol gradient sedimentation ofmitochondrial lysates. IP with anti-V5 antibodies showed nocoprecipitation of editosome components in BF cells that ex-clusively express truncated KREPB5 (Fig. 6D). These cells alsohave much-reduced levels of editosome components comparedwith cells that express WT KREPB5. In contrast, PF cells thatexclusively express truncated KREPB5 have comparable levelsof editosome components as cells expressing WT KREPB5.KREPA1, KREPA2, KREL1, KREPA3, and KREPA6 didcoprecipitate in PF cells that exclusively express truncatedKREPB5, but to a much-reduced extent compared with cellsexclusively expressing WT KREPB5 (Fig. 6D). Comparison ofthe relative abundances of KREPA1, KREPA2, KREL1, andKREPA3 precipitated by WT and truncated KREPB5 revealed apreferential loss of KREPA3 in the truncated KREPB5 sample.As expected, glycerol gradient analysis revealed that ∼20Scomplexes were lost in BF that exclusively express truncatedKREPB5 (SI Appendix, Fig. S9B). In contrast, complexes wereretained in PF but they were shifted to smaller S values. In bothBF and PF, the V5-tagged truncated mutant did not cosedimentwith ∼20S editosomes, indicating that it could not be in-corporated correctly into editosome complexes. Thus, as pre-dicted by KREPB5 cross-linking, the presence of the KREPB5C-terminal domain is essential for KREPB5 function in both life-cycle stages and for the integration of KREPB5 into editosomecomplexes, where it is a pivotal site for interaction with OB-folddomain KREPA proteins in PF editosomes.

DiscussionEditosomes are ∼20S multiprotein complexes containing theenzymes that catalyze cycles of RNA editing, as well as proteinsthat have no known catalytic functions (16, 18–28). Biochemicaland genetic experiments have predominantly revealed a networkof protein–protein interactions that link two heterotrimericsubcomplexes within editosomes (23, 29, 30, 37–39, 45–47).However, the direct interactions of a variety of editosome pro-teins, including those that act at potentially key points of regu-lation during editing, were previously unidentified. Here, CXMSwas used to understand both the global organization and detailedmolecular architecture of editosome protein complexes. Weidentified numerous intra- and interprotein cross-links for alleditosome components, including those that lacked any directinteraction data, such as the endonucleases KREN1 to -3, theendonuclease partner proteins KREPB6 to -B8, KREPB4,KREPA5, and KREPB10. Identified cross-links are consistentwith previously described editosome protein structures, com-parative protein models, and all characterized protein–proteininteractions between editosome subunits. Furthermore, we ex-perimentally validated new interactions that were predicted byCXMS. This work generates a detailed map of editosome proteindomain proximities and identifies molecular interactions be-tween editosome components, thus providing insights into thefunctions of a number of editosome proteins, and expandingour understanding of the global architecture of editosomecomplexes.We observed numerous cross-links within the editosome het-

erotrimeric insertion and deletion subcomplexes, all of whichwere consistent with previous mapping of direct protein–proteininteractions. Of particular note is the extensive cross-linking ofthe KREPA1 and KREPA2 anchor domains with their sub-complex interaction partners. In contrast, the OB-fold domainsof both KREPA1 and KREPA2 did not have any cross-links withtheir interaction partners within the subcomplexes. This findingsuggests that the anchor domains of KREPA1 and KREPA2 arethe main interaction domains that unite deletion or insertionsubcomplex subunits, and that the OB-fold domains in KREPA1and KREPA2 are available for interactions with proteins otherthan subcomplex components or with RNA substrates (61, 62).Indeed, the availability of the KREPA1 and KREPA2 OB-foldsfor substrate binding is potentially vital for KREL1 and KREL2activity. This is because KREL1 and KREL2 lack their own OB-fold

domain, unlike the majority of DNA ligases and mRNA cappingenzymes that have substrate-binding OB folds in addition to cat-alytic domains. It was previously hypothesized that KREPA1 andKREPA2 provide the OB-fold in trans to promote binding of theligases to RNA substrates (45), and the cross-linking observedhere within the heterotrimeric subcomplexes is consistent withthis hypothesis.Our CXMS data further indicate that the insertion and de-

letion subcomplexes are not fully separated within editosomes, asis typically shown in models of editosome protein organization(e.g., SI Appendix, Fig. S1), but are in close physical proximity.This could have implications for our understanding of sub-complex—and specifically KREL1 and KREL2—function. Asimple model evokes parallel functions for the ligases, withKREL1 within the deletion subcomplex specifically catalyzingligation at deletion editing sites, and KREL2 within the insertionsubcomplex specifically catalyzing insertion editing site ligation.However, previously published data also suggest that KREL2 hasa nonessential role in editing, and that the essential KREL1 isresponsible for ligation at both insertion and deletion sites (63).The close physical proximity of the insertion and deletion sub-complexes observed here is consistent with this hypothesis, as itcould allow KREL1 to ligate at insertion sites following KRET2catalyzed U addition.Previous studies principally described protein–protein inter-

actions between insertion and deletion subcomplex components.Notably, here we have advanced our understanding of editingendonuclease interactions with both their specific partner pro-teins and common editing proteins, which had previously

Fig. 5. (A) Network diagrams and (B) interprotein cross-linking maps ofKREPB4 and KREPB5 first neighbors.

E6482 | www.pnas.org/cgi/doi/10.1073/pnas.1610177113 McDermott et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

remained enigmatic. This study corroborates the presence ofonly one type of endonuclease per editosome complex, and in-dicates that KREN1, KREN2, and KREN3 associate with edi-tosomes in generally similar fashions, as each of the three editingendonucleases has comparable proximities to common edito-some proteins. Indeed, in the example of KREPB4, which cross-links to each of the endonucleases, the same regions of KREPB4are in proximity to corresponding parts of each of the differentendonucleases. Each endonuclease primarily formed cross-linkswith insertion subcomplex components KREPA1 and KRET2,but could also cross-link with KREPA2 (KREN1 and KREN3),and KREL1 (KREN2), indicating that they all bridge the in-sertion and deletion heterotrimeric subcomplexes to some ex-tent. Each endonuclease also formed cross-links with KREPA6

and KREPB4 that similarly bridge the insertion and deletionsubcomplexes, whereas KREN1 and KREN2 also cross-linkedwith KREPA3. RNAi knockdown of KREPA3 in PF results inpartially disrupted editosomes that specifically lack endonucle-ase cleavage activity in vitro (40, 64), supporting endonucleaseproximity and interaction with KREPA3.This study provides evidence for proximity between the en-

donucleases and their KREPB6, KREPB7, KREPB8, andKREX1 partner proteins. We found no evidence for proximitybetween noncognate endonucleases and partner proteins—forexample, KREN1 and KREPB7 or KREPB6—despite ourKREPB5–TAP preparation containing all proteins needed forevery combination, consistent with unique associations be-tween KREN1/KREPB8/KREX1, KREN2/KREPB7, and KREN3/KREPB6. Furthermore, through affinity purification of proteinsthat can associate with chimeric KREN1 and KREN2 endonu-cleases, we were able to validate endonuclease and partnerprotein domain proximities and interactions identified byCXMS. We showed that the N-terminal portion of KREN2 thatcontains the zinc finger, RNase III, and PUF motifs is sufficientfor interaction with KREPB7, whereas the C-terminal domainof KREN1 is sufficient for KREX1, but not KREPB8 in-teraction. We also observed cross-linking between the KREX1exoUase and a range of different editosome subunits in additionto KREN1. This finding is in contrast to the KREX2 exoUase,which primarily cross-linked to KREPA2 and KREL1 within thedeletion subcomplex. KREX2 is dispensable, whereas KREX1is essential for normal cell growth and RNA editing in vivo (65,66). Thus, the distinct cross-linking patterns of KREX1 andKREX2 that we describe herein reflect the differential re-quirement for the two exoUases in vivo, where KREX1 medi-ates most editing exoUase activity, and as a result must interactwith a variety of subunits as editing progresses.The cross-linking evidence reinforces the model that catalytic

RNase III domains in KREN1, KREN2, and KREN3 formheterodimers with proteins that possess a degenerate RNase IIIdomain, particularly with KREPB4. Indeed, herein we alsoidentified degenerate RNase III domains in the endonucleasepartner proteins KREPB6, KREPB7, KREPB8, and additionallyKREPB9, and KREPB10, by comparative protein structuremodeling (SI Appendix, Fig. S8 and Dataset S3). Although weobtained relatively few cross-links between these proteins andthe endonucleases, those that we did identify are consistent withheterodimerization. Affinity purification of chimeric KREN1and KREN2 endonucleases also showed that the endonucleaseN-terminal region that includes the RNase III domain is the siteof association with the relevant endonuclease partner protein,supporting a model in which RNase III domains in endonucleasepartner proteins, in addition to KREPB4, can heterodimerizewith the editing endonucleases. The presence of multiple po-tential RNase III heterodimers within the same complex thatmight recognize different RNA substrates would be un-precedented among known RNase III proteins, and may con-tribute to understanding how different editosomes specificallycleave at numerous different insertion and deletion sites. Suchintermolecular heterodimers could cleave mRNA, and leavegRNA intact. An analogous RNase III mechanism has beendescribed, as an artificial bacterial RNase III heterodimer with asingle functional catalytic site can nick dsRNA (67, 68). Curi-ously, recent data indicate that editing consumes gRNAs, asinactivation of KREPA6 or KRET2 results in gRNA accumu-lation (69). The mechanism responsible for editing-mediatedgRNA turnover is unknown, however, and it may be unrelated tothe cleavage activities of editosomes. A clearer view on gRNAmetabolism will require additional experimentation.Full mechanistic understanding of the progression and control

of cycles of insertion and deletion editing within mitochondrialtranscripts is hindered by the lack of high-resolution structures ofeditosomes. Our CXMS data creates a detailed framework forthe global architecture of editosome complexes, and can there-fore provide context to reinterpret existing low-resolution electron

Fig. 6. (A) Diagram showing the U1-like zinc finger, RNase III, and PUFmotifs, and the C-terminal domain of KREPB5. The C-terminal truncation(CTT) mutant of KREPB5 lacks the C-terminal domain (residues 255–384*),but retains all other motifs. (B) Cumulative growth of BF and PF CN cells thatcontain a constitutively expressed V5-tagged WT or CTT mutant KREPB5allele in addition to the tet-regulatable WT allele that is expressed (E) orrepressed (R). (C) Heat maps showing the log10-transformed abundances ofRNAs from BF and PF KREPB5 CN cells in which either V5-tagged WT or CTTmutant alleles (B5 EE) were exclusively expressed for 3 (BF) or 4 (PF) d relativeto those from cells in which the tet-regulatable WT allele (B5 reg) was alsoexpressed. RNAs were analyzed using qRT-PCR, and their levels normalizedto the level of TERT RNA. The relative abundances of KREPB5, never-edited(COI and ND4), pre-edited, and edited RNAs were determined in quadru-plicate. (D) Anti-V5 IP of editosome complexes from BF and PF cells thatexclusively expressed V5-tagged WT or CTT mutant KREPB5. Anti-V5 im-munoprecipitates from BF and PF cells, and BF and PF input cell lysates wereprobed with monoclonal antibodies against KREPA1, KREPA2, KREL1,KREPA3, and the V5 epitope tag, and a polyclonal antibody against KREPA6.Lysates from KREPB5 CN cells that had no tags and mock IPs without anti-body (−) were used as negative controls.

McDermott et al. PNAS | Published online October 5, 2016 | E6483

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

microscopy data. For example, electron microscopy images ofTAP-purified editosomes from T. brucei and Leishmania tar-entolae revealed bipartite particles composed of two approxi-mately equally sized, but structurally different globular domainsthat are connected via an interface (70, 71). These studies alsoindicated that most editosome proteins are likely present assingle copies in the complex, although the exact stoichiometry ofeditosomes remains unknown. Although the resolution of thesestructures precludes precise assignment of each editosome pro-tein, they are consistent with the insertion and deletion hetero-trimeric subcomplexes being bridged by a group of proteins,including KREPB4, KREPA6, KREPA3, and KREPA4. Usingour CXMS data to orient protein positions, we can expand thathypothesis to include both endonucleases and their partnerproteins. Protrusions from this bridging interface observed byelectron microscopy could correspond to an editing endonucle-ase and associated partner proteins, which we now know cross-link with bridging proteins KREPB4, KREPA6, and KREPA3.Significant structural diversity within the population of electronmicroscopy images of editosomes included a subset of ∼20S com-plexes with additional density. This added density could corre-spond to either KREX1 or to KREPB10, which cross-linkedspecifically with KREN1. Further investigation will be needed todetermine if these hypotheses are correct.The editosome architecture identified in this study can also act

as a reference for future studies to examine differential editingbetween life-cycle stages. A number of transcripts are differen-tially edited between BF and PF cells (3–7, 72–74). The molec-ular basis for this developmental regulation remains unknown,although a growing body of evidence shows that elimination ormutation of the same editosome protein differentially affectsRNA editing in BF and PF parasites, and similar results havebeen observed with other mitochondrial RNA processing proteincomplexes (36, 60, 75). Indeed, KREPB5 is an example of such aprotein, where knockdown (36), or truncation of the C-terminaldomain (reported here) results in a dramatic loss of editosomesand editosome components in BF, but retention of componentsand of complexes with smaller S values in PF. This finding in-dicates that that there are intrinsic differences between BF andPF editosomes, and other complexes as well. If differences in

editosome cross-linking are observed between PF and BF, theycould provide insight into these recently identified developmentalchanges in function.In addition to providing insight into the global editosome ar-

chitecture, the CXMS data from our work can similarly improvestructural modeling of individual proteins. Crystal structureshave previously been solved for only three editosome proteins[KREPA6 (46, 47, 49), KRET2 (50), and KREL1 (51)], and forspecific domains of two others [KREPA1 (45) and KREPA3 (46,47)]. The structures of all other editosome proteins remain un-known. Here, we built comparative models for editosome proteindomains that were generally consistent with intralink distances inCXMS. These models suggest new functions for a number ofproteins, exemplified by the modeling of RNase III domains inKREPB6 to -B10. The combination of known crystal structuresand comparative protein modeling with our cross-linking data willprovide highly reliable restraints toward the first atomic resolu-tion ∼20S editosome structure. This will ultimately provide astructural framework for understanding how editosomes functionduring RNA editing in kinetoplastids, thus aiding the develop-ment of new therapeutics that will treat human African try-panosomiasis, Chagas disease, and leishmaniases.

Materials and MethodsSee SI Appendix, SI Materials and Methods for details of T. brucei growthand genetic manipulation, TAP, IP, glycerol gradient fractionation, Westernblotting, RT-qPCR analysis, and homology modeling.

For CXMS, TAP-purified editosome complexes were mixed with 2 mM or5 mM BS3, quenched, digested with trypsin, and peptides were fractionatedby strong cation-exchange chromatography and analyzed by MS. Cross-linked peptides were identified by searching the MS data against appro-priate databases using pLink (76) and Nexus (48, 77). A 5% false-discoveryrate was used for both pLink and Nexus searches. Details are provided in SI

Appendix, SI Materials and Methods.

ACKNOWLEDGMENTS. The authors thank Peter Cimermancic and AndrejSali for assistance with use of the ModPipe pipeline. This work was sup-ported by National Institutes of Health Grants R01AI014102 (to K.S.) andP50 GM076547 and R21CA175849 (to J.A.R.).

1. Stuart K, Feagin JE (1992) Mitochondrial DNA of kinetoplastids. Int Rev Cytol 141:65–88.

2. Shapiro TA, Englund PT (1995) The structure and replication of kinetoplast DNA.Annu Rev Microbiol 49:117–143.

3. Koslowsky DJ, Bhat GJ, Perrollaz AL, Feagin JE, Stuart K (1990) The MURF3 gene ofT. brucei contains multiple domains of extensive editing and is homologous to asubunit of NADH dehydrogenase. Cell 62(5):901–911.

4. Souza AE, Myler PJ, Stuart K (1992) Maxicircle CR1 transcripts of Trypanosoma bruceiare edited and developmentally regulated and encode a putative iron-sulfur proteinhomologous to an NADH dehydrogenase subunit. Mol Cell Biol 12(5):2100–2107.

5. Souza AE, Shu HH, Read LK, Myler PJ, Stuart KD (1993) Extensive editing of CR2maxicircle transcripts of Trypanosoma brucei predicts a protein with homology to asubunit of NADH dehydrogenase. Mol Cell Biol 13(11):6832–6840.

6. Feagin JE, Jasmer DP, Stuart K (1987) Developmentally regulated addition of nucle-otides within apocytochrome b transcripts in Trypanosoma brucei. Cell 49(3):337–345.

7. Bhat GJ, Koslowsky DJ, Feagin JE, Smiley BL, Stuart K (1990) An extensively editedmitochondrial transcript in kinetoplastids encodes a protein homologous to ATPasesubunit 6. Cell 61(5):885–894.

8. Benne R, et al. (1986) Major transcript of the frameshifted coxII gene from trypano-some mitochondria contains four nucleotides that are not encoded in the DNA. Cell46(6):819–826.

9. Shaw JM, Feagin JE, Stuart K, Simpson L (1988) Editing of kinetoplastid mitochondrialmRNAs by uridine addition and deletion generates conserved amino acid sequencesand AUG initiation codons. Cell 53(3):401–411.

10. Schnaufer A, et al. (2001) An RNA ligase essential for RNA editing and survival of thebloodstream form of Trypanosoma brucei. Science 291(5511):2159–2162.

11. Blum B, Bakalara N, Simpson L (1990) A model for RNA editing in kinetoplastid mi-tochondria: “Guide” RNA molecules transcribed from maxicircle DNA provide theedited information. Cell 60(2):189–198.

12. Blum B, Simpson L (1990) Guide RNAs in kinetoplastid mitochondria have a non-encoded 3′ oligo(U) tail involved in recognition of the preedited region. Cell 62(2):391–397.

13. Pollard VW, Hajduk SL (1991) Trypanosoma equiperdum minicircles encode threedistinct primary transcripts which exhibit guide RNA characteristics. Mol Cell Biol11(3):1668–1675.

14. Pollard VW, Rohrer SP, Michelotti EF, Hancock K, Hajduk SL (1990) Organization ofminicircle genes for guide RNAs in Trypanosoma brucei. Cell 63(4):783–790.

15. Sturm NR, Simpson L (1990) Kinetoplast DNA minicircles encode guide RNAs for ed-iting of cytochrome oxidase subunit III mRNA. Cell 61(5):879–884.

16. Stuart KD, Schnaufer A, Ernst NL, Panigrahi AK (2005) Complex management: RNAediting in trypanosomes. Trends Biochem Sci 30(2):97–105.

17. Aphasizhev R, Aphasizheva I (2011) Uridine insertion/deletion editing in trypano-somes: A playground for RNA-guided information transfer. Wiley Interdiscip Rev RNA2(5):669–685.

18. Panigrahi AK, et al. (2001) Association of two novel proteins, TbMP52 and TbMP48,with the Trypanosoma brucei RNA editing complex. Mol Cell Biol 21(2):380–389.

19. Panigrahi AK, et al. (2001) Four related proteins of the Trypanosoma brucei RNAediting complex. Mol Cell Biol 21(20):6833–6840.

20. Panigrahi AK, Allen TE, Stuart K, Haynes PA, Gygi SP (2003) Mass spectrometricanalysis of the editosome and other multiprotein complexes in Trypanosoma brucei.J Am Soc Mass Spectrom 14(7):728–735.

21. Panigrahi AK, et al. (2003) Identification of novel components of Trypanosoma bruceieditosomes. RNA 9(4):484–492.

22. Panigrahi AK, et al. (2006) Compositionally and functionally distinct editosomes inTrypanosoma brucei. RNA 12(6):1038–1049.

23. Carnes J, Soares CZ, Wickham C, Stuart K (2011) Endonuclease associations with threedistinct editosomes in Trypanosoma brucei. J Biol Chem 286(22):19320–19330.

24. Carnes J, Trotter JR, Ernst NL, Steinberg A, Stuart K (2005) An essential RNase III insertionediting endonuclease in Trypanosoma brucei. Proc Natl Acad Sci USA 102(46):16614–16619.

25. Carnes J, Trotter JR, Peltan A, Fleck M, Stuart K (2008) RNA editing in Trypanosomabrucei requires three different editosomes. Mol Cell Biol 28(1):122–130.

26. Trotter JR, Ernst NL, Carnes J, Panicucci B, Stuart K (2005) A deletion site editingendonuclease in Trypanosoma brucei. Mol Cell 20(3):403–412.

27. Lerch M, et al. (2012) Editosome accessory factors KREPB9 and KREPB10 in Trypano-soma brucei. Eukaryot Cell 11(7):832–843.

28. Ernst NL, et al. (2003) TbMP57 is a 3′ terminal uridylyl transferase (TUTase) of theTrypanosoma brucei editosome. Mol Cell 11(6):1525–1536.

29. Guo X, Carnes J, Ernst NL, Winkler M, Stuart K (2012) KREPB6, KREPB7, and KREPB8are important for editing endonuclease function in Trypanosoma brucei. RNA 18(2):308–320.

E6484 | www.pnas.org/cgi/doi/10.1073/pnas.1610177113 McDermott et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0

30. Carnes J, et al. (2012) Mutational analysis of Trypanosoma brucei editosome proteinsKREPB4 and KREPB5 reveals domains critical for function. RNA 18(10):1897–1909.

31. Worthey EA, Schnaufer A, Mian IS, Stuart K, Salavati R (2003) Comparative analysis ofeditosome proteins in trypanosomatids. Nucleic Acids Res 31(22):6392–6408.

32. MacRae IJ, Doudna JA (2007) Ribonuclease revisited: Structural insights into ribonu-clease III family enzymes. Curr Opin Struct Biol 17(1):138–145.

33. Nicholson AW (2014) Ribonuclease III mechanisms of double-stranded RNA cleavage.Wiley Interdiscip Rev RNA 5(1):31–48.

34. Babbarwal VK, Fleck M, Ernst NL, Schnaufer A, Stuart K (2007) An essential role ofKREPB4 in RNA editing and structural integrity of the editosome in Trypanosomabrucei. RNA 13(5):737–744.

35. Wang B, et al. (2003) TbMP44 is essential for RNA editing and structural integrity ofthe editosome in Trypanosoma brucei. Eukaryot Cell 2(3):578–587.

36. McDermott SM, Guo X, Carnes J, Stuart K (2015) Differential editosome proteinfunction between life cycle stages of Trypanosoma brucei. J Biol Chem 290(41):24914–24931.

37. Schnaufer A, et al. (2010) A protein-protein interaction map of trypanosome ∼20Seditosomes. J Biol Chem 285(8):5282–5295.

38. Schnaufer A, et al. (2003) Separate insertion and deletion subcomplexes of the Try-panosoma brucei RNA editing complex. Mol Cell 12(2):307–319.

39. Mehta V, Sen R, Moshiri H, Salavati R (2015) Mutational analysis of Trypanosomabrucei RNA editing ligase reveals regions critical for interaction with KREPA2. PLoSOne 10(3):e0120844.

40. Guo X, Ernst NL, Stuart KD (2008) The KREPA3 zinc finger motifs and OB-fold domainare essential for RNA editing and survival of Trypanosoma brucei.Mol Cell Biol 28(22):6939–6953.

41. Law JA, O’Hearn S, Sollner-Webb B (2007) In Trypanosoma brucei RNA editing,TbMP18 (band VII) is critical for editosome integrity and for both insertional anddeletional cleavages. Mol Cell Biol 27(2):777–787.

42. Salavati R, et al. (2006) KREPA4, an RNA binding protein essential for editosome in-tegrity and survival of Trypanosoma brucei. RNA 12(5):819–831.

43. Tarun SZ, Jr, et al. (2008) KREPA6 is an RNA-binding protein essential for editosomeintegrity and survival of Trypanosoma brucei. RNA 14(2):347–358.

44. Aphasizheva I, et al. (2009) Novel TUTase associates with an editosome-like complexin mitochondria of Trypanosoma brucei. RNA 15(7):1322–1337.

45. Park YJ, et al. (2012) The structure of the C-terminal domain of the largest editosomeinteraction protein and its role in promoting RNA binding by RNA-editing ligase L2.Nucleic Acids Res 40(14):6966–6977.

46. Park YJ, Hol WG (2012) Explorations of linked editosome domains leading to thediscovery of motifs defining conserved pockets in editosome OB-folds. J Struct Biol180(2):362–373.

47. Park YJ, Pardon E, Wu M, Steyaert J, Hol WG (2012) Crystal structure of a heterodimerof editosome interaction proteins in complex with two copies of a cross-reactingnanobody. Nucleic Acids Res 40(4):1828–1840.

48. Luo J, et al. (2015) Architecture of the human and yeast general transcription andDNA repair factor TFIIH. Mol Cell 59(5):794–806.

49. Wu M, et al. (2011) Structures of a key interaction protein from the Trypanosomabrucei editosome in complex with single domain antibodies. J Struct Biol 174(1):124–136.

50. Deng J, Ernst NL, Turley S, Stuart KD, Hol WG (2005) Structural basis for UTP specificityof RNA editing TUTases from Trypanosoma brucei. EMBO J 24(23):4007–4017.

51. Deng J, Schnaufer A, Salavati R, Stuart KD, Hol WG (2004) High resolution crystalstructure of a key editosome enzyme from Trypanosoma brucei: RNA editing ligase 1.J Mol Biol 343(3):601–613.

52. Eswar N, et al. (2003) Tools for comparative protein structure modeling and analysis.Nucleic Acids Res 31(13):3375–3380.

53. Soding J, Biegert A, Lupas AN (2005) The HHpred interactive server for protein ho-mology detection and structure prediction. Nucleic Acids Res 33(Web Server issue):W244–W248.

54. Sali A, Blundell TL (1993) Comparative protein modelling by satisfaction of spatialrestraints. J Mol Biol 234(3):779–815.

55. Biasini M, et al. (2014) SWISS-MODEL: Modelling protein tertiary and quaternarystructure using evolutionary information. Nucleic Acids Res 42(Web Server issue):W252–W258.

56. Czerwoniec A, Kasprzak JM, Bytner P, Dobrychłop M, Bujnicki JM (2015) Structure andintrinsic disorder of the proteins of the Trypanosoma brucei editosome. FEBS Lett 589(19 Pt A):2603–2610.

57. Merkley ED, et al. (2014) Distance restraints from crosslinking mass spectrometry:Mining a molecular dynamics simulation database to evaluate lysine-lysine distances.Protein Sci 23(6):747–759.

58. Gao G, et al. (2005) Functional complementation of Trypanosoma brucei RNA in vitroediting with recombinant RNA ligase. Proc Natl Acad Sci USA 102(13):4712–4717.

59. Kala S, Moshiri H, Mehta V, Yip CW, Salavati R (2012) The oligonucleotide binding(OB)-fold domain of KREPA4 is essential for stable incorporation into editosomes.PLoS One 7(10):e46864.

60. McDermott SM, Carnes J, Stuart K (2015) Identification by random mutagenesis offunctional domains in KREPB5 that differentially affect RNA editing between lifecycle stages of Trypanosoma brucei. Mol Cell Biol 35(23):3945–3961.

61. Rajkowitsch L, et al. (2007) RNA chaperones, RNA annealers and RNA helicases. RNABiol 4(3):118–130.

62. Kwon SH, et al. (2007) Translation initiation factor eIF1A possesses RNA annealing ac-tivity in its oligonucleotide-binding fold. Biochem Biophys Res Commun 361(3):681–686.

63. Drozdz M, et al. (2002) TbMP81 is required for RNA editing in Trypanosoma brucei.EMBO J 21(7):1791–1799.

64. Law JA, O’Hearn SF, Sollner-Webb B (2008) Trypanosoma brucei RNA editing proteinTbMP42 (band VI) is crucial for the endonucleolytic cleavages but not the subsequentsteps of U-deletion and U-insertion. RNA 14(6):1187–1200.

65. Carnes J, Lewis Ernst N, Wickham C, Panicucci B, Stuart K (2012) KREX2 is not essentialfor either procyclic or bloodstream form Trypanosoma brucei. PLoS One 7(3):e33405.

66. Ernst NL, Panicucci B, Carnes J, Stuart K (2009) Differential functions of two editosomeexoUases in Trypanosoma brucei. RNA 15(5):947–957.

67. Conrad C, Schmitt JG, Evguenieva-Hackenberg E, Klug G (2002) One functional sub-unit is sufficient for catalytic activity and substrate specificity of Escherichia coli en-doribonuclease III artificial heterodimers. FEBS Lett 518(1-3):93–96.

68. Meng W, Nicholson AW (2008) Heterodimer-based analysis of subunit and domaincontributions to double-stranded RNA processing by Escherichia coli RNase III in vitro.Biochem J 410(1):39–48.

69. Aphasizheva I, et al. (2014) RNA binding and core complexes constitute the U-in-sertion/deletion editosome. Mol Cell Biol 34(23):4329–4342.

70. Golas MM, et al. (2009) Snapshots of the RNA editing machine in trypanosomescaptured at different assembly stages in vivo. EMBO J 28(6):766–778.

71. Li F, et al. (2009) Structure of the core editing complex (L-complex) involved in uridineinsertion/deletion RNA editing in trypanosomatid mitochondria. Proc Natl Acad SciUSA 106(30):12306–12310.

72. Feagin JE, Stuart K (1988) Developmental aspects of uridine addition within mito-chondrial transcripts of Trypanosoma brucei. Mol Cell Biol 8(3):1259–1265.

73. Riley GR, Myler PJ, Stuart K (1995) Quantitation of RNA editing substrates, productsand potential intermediates: implications for developmental regulation. Nucleic AcidsRes 23(4):708–712.

74. Koslowsky DJ, Riley GR, Feagin JE, Stuart K (1992) Guide RNAs for transcripts withdevelopmentally regulated RNA editing are present in both life cycle stages of Try-panosoma brucei. Mol Cell Biol 12(5):2043–2049.

75. Aphasizheva I, et al. (2015) Ribosome-associated PPR proteins function as trans-lational activators in mitochondria of trypanosomes. Mol Microbiol 99(6):1043–1058.

76. Yang B, et al. (2012) Identification of cross-linked peptides from complex samples. NatMethods 9(9):904–906.

77. Knutson BA, Luo J, Ranish J, Hahn S (2014) Architecture of the Saccharomyces cer-evisiae RNA polymerase I core factor complex. Nat Struct Mol Biol 21(9):810–816.

McDermott et al. PNAS | Published online October 5, 2016 | E6485

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

0