Summary of Interpretation

of sequence variants

schemes- splice variants

David Moore- Regional Genetics Service, Edinburgh

Aims of workshop

• Feedback on 2012/13 NEQAS returns

• Best practice guidelines

– When to mention:

• When is it prudent to mention potential splicing effects?

– What to mention:

• List or scores, generic phrase, prediction of result?

• RNA or other functional studies

• How do clinicians react to splice variants on

reports and recommendations/suggestions

for RNA studies

• Invariant sites

– Acceptor: -1,-2

– Donor: +1,+2

• All others vary

Structure of splice sites

Collected information from 2012 scheme

• 22 labs (76%) classed the -2, -1, +1, +2 variants as class 5 (some with the caveat of classification after literature searches), 3 as “class 4 or 5” and 4 labs would class them as class 4.

• The drop in splice score necessary to deem a variant into class 4 or 5 varied from 2.4% to 100%, with one lab stating that they “didn’t look”- presumably because they do not come across this type of variant.

• The most common response was “over 10% in three programs” which 6 labs (21%) supplied as their answer.

• All laboratories that participated use Alamut to assess pathogenicity. As all labs are utilising the same software this provides a good platform to attempt to attain some parity in assessment.

Exceptions to every rule….

• Exceptions to drop of 10% “rule”

– CFTR c.1584G>A (p.Glu528Glu)

– At a frequency of 1.44% on ESP

– Has been seen in an unaffected father who

also has p.Phe508del

Variants in defined splice sites • Some changes are clearly tolerated and are

recognised polymorphisms – c.732C>T (p.=) myh7 exon 8 (last base of exon)

– 28% on ESP

Literature searches

• Can aid with interpretation – Review of detected variants can prove useful

• Modest changes but experimentally proven – GAA gene c.-32-13T>G, accounts for half of disease

causing alleles in Pompe disease in caucasians.

– SPAST gene c.870+3A>G

Creation of sites rivalling natural site but

outwith traditional splice regions

• Spast c.1729-20T>G

– Segregation analysis

• MYBPC3 c.906-36G>A

– Frank-Hansen et al

– NGS ROI’s

22/06/1947

N/c.1729-20T>G 13/03/1946

N/c.1729-20T>G

14/01/1977

N/c.1729-20T>G

Caution

• c.1005-2A>G.

– As shown by this example, a site in the vicinity

does not necessarily mean it will be used

Missense variants within exons

• TNNT2 c.341C>T p.(Ala114Val)

– These should be routinely examined for potential splicing effects.

– In this case a site is created with a score that rivals the natural splice site.

– This variant was classified as a 3

Other factors to consider

• Disease mechanism?

– Loss of function or gain of function?

• Alternate transcripts?

– In primary transcript?

• Result in out of frame or retain reading frame?

– Predictions of either skipping or inclusion of extra material do not always correlate with experimental data.

• Heterogeneous disorder or single gene?

• Mode of inheritance of disorder?

• All other variants excluded?

• If a site is created, what are the scores of the naturally occurring sites?

Best practice guidelines

• It is acceptable to assign nucleotide changes that disrupt the consensus dinucleotide splice sites (+/-1 and +/-2) as clearly pathogenic (class 5) requiring no further investigation.

• It is important to consider the impact of alternative transcripts if known.

• An NGRL study showed that the better performing tools were capable of a good degree of accuracy but must be used with caution. It also showed that they are likely to be useful beyond the invariant acceptor and donor sites at intron/exon boundaries, but not beyond positions +7 and -10 of exons.

• Laboratories should be aware that any sequence changes (including silent variants) may create new splice sites and as such are amenable to analysis using this software, especially when AG or GT dinucleotide sequences are formed.

• It is considered good practice to use more than one splice-site prediction tool.

• It is not currently possible to set criteria for the change in prediction tool scores which should be considered significant (e.g. a 10% deviation from the wild-type score). This remains a matter for local judgement and agreement.

• Summary: It is acceptable to use in-silico splice site prediction; however, it is unacceptable to base an unequivocal clinical interpretation solely on this line of evidence. It is, however, acceptable to suggest further investigations based on the outcome. If this method of prediction is used it is recommended to arrive at an interpretation based upon a consensus of at least three splice site prediction pro-grammes.

Best practice guidelines- RNA studies

• Where possible, RNA studies are the best means of interpreting the consequences of a splicing mutation. Therefore it is recommended that RNA studies be performed, in the following context:

– If RNA from an appropriate and validated tissue or cell type is available.

– If a variant of uncertain significance is found and prediction programs give an indication for a splice site alteration.

– Where other lines of evidence support an effect on splicing.

• Only if the entire gene has been tested and no other variants have been found.

• In addition we recommend: – In cases where no splice errors are apparent on mRNA analysis, biallelic

expression by molecular analysis of a variant should be proven to conclude that the unclassified variant has no effect such as nonsense mediated decay which may prevent in vitro observation of aberrant splice product.

– Naturally occurring alternate splice variants should be excluded by the testing of matched controls and the use of RNA from appropriate and validated tissue or cell types.

• Summary: Given the high predictive value of RNA studies they must be regarded as essential for the definitive interpretation of putative splicing mutations. However, it is recognised that not all laboratories have the facilities to perform these analyses and that limited expression patterns may mean that the required tissue is not available for analysis.

2012 case and results supplied Question 1 • Helen McMillan (dob 02/08/2004) has been referred by a

Metabolic Consultant as she is presenting with Glycogen Storage Disease type 12. She has a reduction in Aldose levels and is presenting with myopathic symptoms, including muscle weakness and premature muscle fatigue. Helen has been found to have a large deletion by array CGH which has been shown to include exons 1 to 6 of the aldolase A (ALDOA) gene (NM_00034.3). There is no other history of this disease in the family. Variants in this gene act in a recessive manner, so your laboratory has been asked to screen the gene by Sanger sequencing and has detected the heterozygous variant; c.363C>T p.(=).

Key points

• Aldolase A deficiency and a clinical presentation that matches the expected phenotype of GSDXII.

• As a deletion has been detected and the c.363C>T variant introduces a cryptic donor splice site that could be pathogenic, does this confirm diagnosis?

• The points to cause some concern are that exons 1-6 are non-coding and it is purely a splicing prediction.

0

2

4

6

8

10

12

14

16

18

Four Three

and a

half

Three

Number

• Indicators to establish the pathogenicity of the variant

were decided to be whether the variants occurred in-cis

or in-trans and to suggest RNA studies to establish

whether aberrant splicing was occurring.

• 17 (59%) labs scored the variant as a class 4, 1 as a 3.5

and the remaining 11 (38%) as a class 3- giving the

variant an average scoring of 3.6.

2012 case and results supplied

2013 Scheme • Four laboratories mentioned that the variant in question 1 could

potentially have an effect on splicing (see figure on left below). – Only one program predicted the creation of an acceptor splice site.

• This is in contrast to the c.5753A>G p.(Asp191Gly) from question 2 – Predicted to improve a potential acceptor site (3 programs)

– Predicted to create a donor site with moderate scores (4 programs).

c.1022G>A p.(Arg314Gln) from question 1 c.5753A>G p.(Asp191Gly) from question 2

2013 Scheme • Two laboratories mentioned a potential splicing effect

caused by the variant in question 3 RPS6KA3 c.229G>A p.(Gly77Arg). It actually decreases a cryptic site.

• Mentioning that a missense variant may have a potential splicing effect in order to support pathogenicity is not appropriate, as you are trying to additively apply a different mutational mechanism

• It is different if there is a significant splicing effect predicted

2013 Scheme

• As the scheme format differed from 2012,

comments to be included on reports were

requested to be submitted, these included:

– In Silico splicing analysis indicates variant creates a cryptic

splice site in 1 out of 5 programs.

– In addition, one splice prediction software tool (human

splicing finder) predicts the creation of a cryptic splice

acceptor site.

– Possible splicing effect. Further work- RNA analysis for

potential splicing effects

– Alamut predictions suggest that the nucleotide change is

unlikely to affect splicing but this is not definitive.

• Some of the returns included numerical scores

2013-Comments

• These statements only really serve to demonstrate that the variants had been examined for an effect on splicing, however according to Best Practice guidelines, a potential effect should only be mentioned if three or more algorithms agree.

• The suggestion for RNA studies in this case (question 3) was not indicated- the variant was determined to be a class 4 missense variant. Obtaining the appropriate sample from the patient and spending the time, effort and money to see if it has a splicing effect in this case was not deemed useful.

• Unqualified statements such as “1 from 4”, are not useful, as the performance of the algorithms varies.

• Entering values without scales is equally not useful, as they vary between the algorithms

Final points

• Most useful to carry out RNA studies if the result then becomes clinically relevant i.e. confirms diagnosis, allows treatment or either presymptomatic or prenatal testing for relatives.

• Offering RNA studies for variants with a poor predicted effect on splicing may not be the best use of resource, requires re-sampling of the patient (which may be invasive) in order to get the appropriate and may raise expectation and delay final conclusion when not strongly indicated.

• The next step is if mis-splicing is demonstrated- what level is required to confirm the diagnosis?

• Was the assay fit for purpose- did it include all of the necessary motifs?

• Splicing should only be mentioned in the main body of the report, if it is implicated as being the pathogenic mechanism.

– Could be mentioned in the notes section to show it has been carried out

• Mentioning specific algorithms, scores and theorising on potential effects- such as exon skipping, intron retention or inclusion of intronic material, may not be useful for the clinicians and could either be wrong, or misinterpreted. A generic statement such as “is predicted to have an effect on splicing” or “is likely to result in aberrant splicing” with details in the notes section, or a statement “more details on the splice analysis are available on request” might be more useful.