Genetics: HLA-DQA*05:01

[ryanc97]

Adding on to this, from Mella and Fluge's CycloME paper:

Frequency of HLA risk alleles (HLA DQB1*03:03 and/or HLAC*07:04) in responders and non-responders during follow-up. P-value from
Fischer’s exact test. HLA, Human Leukocyte Antigen

This was decently predictive of response rate in the paper, those HLA risk positive had a 10 responder to 1 non responder rate and those positive had a 11 responder to 29 non responder rate.

However we are talking about HLA DQA here so I don't know the difference.

But in general I am skeptical about DecodeME or genetics as I don't see how it would impact treatment or intervention. But i dont know much either

 

[forestglip]

Frequency of HLA risk alleles (HLA DQB1*03:03 and/or HLAC*07:04) in responders and non-responders during follow-up.

Yes, the same two alleles that were risk alleles for ME/CFS in this other study:

Human Leukocyte Antigen alleles associated with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS): Fluge, Mella et al 2020

I feel like I must be missing something, because why was there never much interest in this? What's the chance of the same exact two alleles, out of the dozens possible, being significant in both?

I never looked deeply into it, but maybe the populations overlap so it's not a true replication.

 

[jnmaciuch]

That strikes me as unintuitive.

ME/CFS is triggered by a variety of pathogens and my intuition tells me that while specific genetic vulnerabilities for ME/CFS would certainly exist for each of these pathogens, they wouldn't be among the top hits in a GWAS. Variants related to disease process would dominate because these are what ME/CFS patients tend to have in common.

Maybe my mental model is wrong. I'm just an amateur that can't stop thinking about these things.

It's a distinction that can't be easily made, because there are parts of the immune response that are going to be pathogen-specific and parts that are going to be generally anti-viral/anti-bacterial, etc. There's also differences in general cell susceptibility to infection that is somewhat independent of an immune system response, which is what the RABGAP1L finding speaks to. It's been found to be relevant in influenza and strep because it was studied in those contexts, but many viruses use those same mechanisms to infect cells.

So, long story short, it's very possible to have genetic hits that confer broader anti-viral protection (to an extent, since none of those genes are going to be fully protective against infection). And it's also important to keep in mind that the odds ratios are very slim for all the hits--increased susceptibility to just one class of pathogens may still be driving the hit.

 

[ryanc97]

But back to the point, it would seem based on the data the gene HLA DQB1*03:03 and/or HLAC*07:04 is fairly predictive of response on cyclophosphamide, that's my point. Why I have no idea.

 

[wigglethemouse]

I am tempted to think that if anything a DQ signal is nudging things towards the more innate side of T cell recognition, which can use DQ, as seen with MAIT cells and gamma delta cells. It may just indicate a non-specific threshold effect due to perhaps binding outside of the traditional active site but even if it just does that it maybe should not be ignored.

I think that story might tie in well with the BTN2A2 and BTN3A3 seem to have a similar role in gamma delta T cells.

 

[Simon M]

Interesting given they didn’t actually measure DQA1*05:01

As far as I understand they inferred its value by imputation (using linkage disequilibrium) which is ever more interesting and confusing given it didn’t return the expected associations with other linkages.

I hope we will get some higher resolution data from SequenceME to clear this up.

There will be a new analysis much sooner than that. I think what they have to do is to run the HLA imputation cases and controls at the same time (on the super computer, that is). , But that is not official – just that they are going to analyse it again

 

[Jonathan Edwards]

A dumb question. Is the fact that DQ is a heterodimer of A and B chains, both of which are polymorphic and therefore producing a combinatorial mess of protein variants a reason why DQ may turn up in genetic risk studies for no good reason? Throughout the process of seeing these results emerge I have been impressed that there is a clear MHC signal that ultimately focused on DQ. I find it hard to believe this is spurious but we live and learn.

 

[Hutan]

Apologies if this is a stupid question but how do you know which SNPs are measured and which ones are imputed? Do they mention this somewhere in the text or is it somewhere in the data?

The DecodeME paper says, in a couple of places, that they imputed HLA alleles.

We imputed human leukocyte antigen (HLA) alleles separately using the HLA*IMP:02 algorithm, which the UKB also used for this study’s control individuals (25). Further methodological detail is provided in the Supplementary Methods. It looks as though The HLA alleles tend to be imputed in GWAS because it's very hard to identify them directly.

It seems that that is standard practice, because identifying them directly is very difficult. The presence of an HLA allele is imputed using an algorithm that works off nearby SNPs that are in linkage disequilibrium with the HLA allele. I found the introduction of this paper useful:
HLA imputation and its application to genetic and molecular fine-mapping of the MHC region in autoimmune diseases

Excerpts

The major histocompatibility complex (MHC) region is located at 6p21.3 with spanning approximately 5 Mb in length [1]. The genes encoded by this region are clearly enriched for immune responses and inflammatory pathways [1, 2]. Consistently with its function, genetic variants in the MHC region contribute to the genetics of various human complex traits, especially autoimmune diseases and infectious diseases [3, 4]. The MHC is the region with the highest number of disease associations reported in genome-wide association studies (GWAS) [5]. These associations included those “non-autoimmune diseases,” such as cardiovascular, metabolic, and neurological diseases, implying immune-related mechanisms behind the progression of these diseases and the broader significance of the MHC region [6, 7]. Among the genes densely present in the MHC region, human leukocyte antigen (HLA) genes are considered to explain most of the genetic heritability of MHC. HLA molecules mediate antigen presentation, which is a critical component in triggering the subsequent immune responses; thus, variations in HLA genes have been considered to associate with the risk of immune-related diseases directly. For a representative instance, in type 1 diabetes (T1D), the MHC region explains 42.8% of phenotypic variance, of which HLA-DRB1, -DQA1, and -DQB1 haplotypes account for the most significant proportion at 29.6% [8].

Associations of single nucleotide polymorphisms (SNPs) with phenotypes of interest in GWAS typically do not indicate their direct causal roles but linkage with truly causal variants. To identify such causal variants (i.e., fine-mapping), comprehensive genotyping of regional variations including HLA allelic types for the target individuals is needed. However, the MHC region is one of the most challenging regions of the human genome to genotype because of its high degree of polymorphism and structural variations [9].

**********Thus, HLA typing is conducted with specific approaches, including traditional polymerase chain reaction (PCR)-based methods and next-generation sequencing (NGS). They are so labor-intensive, time-consuming, and expensive that they could not be applied to fine-mapping for large cohorts of GWAS [6, 10]. Subsequently, the genotypes of HLA alleles are indirectly imputed from SNP-level data using a pre-constructed HLA reference panel. HLA imputation has successfully contributed to the fine-mapping of causal HLA variants to delineate of the immunopathology of various diseases. ****************

Beginning with a simple inference using tag SNPs [11, 12], various statistical HLA allelic imputation methods have been developed, each with its advantages and disadvantages for practical use. In this review, we discuss the recent advances and challenges in HLA imputation methods and available HLA reference panels. We also discuss the relationship between the MHC region and autoimmune diseases revealed by the fine-mapping and the current understanding of how HLA variations contribute to disease etiology.

 

[Hoopoe]

A dumb question. Is the fact that DQ is a heterodimer of A and B chains, both of which are polymorphic and therefore producing a combinatorial mess of protein variants a reason why DQ may turn up in genetic risk studies for no good reason? Throughout the process of seeing these results emerge I have been impressed that there is a clear MHC signal that ultimately focused on DQ. I find it hard to believe this is spurious but we live and learn.

I don't see how the polymorphic nature of DQ would lead to spurious associations.

What may be happening is a DQ variant that's not contributing to disease being in linkage equilibrium with another variant that is contributing to disease, and of which we don't yet have knowledge.

Since DQ was determined via HLA imputation, the authors should be well aware of this.