Tuesday, December 25, 2012

All I want for Christmas is a cure for HIV

Commonly known as CCR5-Δ32 (CCR5 delta 32), rs333 isn’t a true SNP - it’s a deletion of 32 nucleotides in the CCR5 chemokine receptor gene. It was considered important enough though for dbSNP to assign it one of the earliest 3 digit SNP identifiers, and for 23andMe to have added it to their custom content under the name i3003626. What makes this variation so interesting is that people who carry 2 copies of the deletion are highly resistant to HIV infection. Wikipedia has a great explanation of this variation and it’s likely spread from survivors of earlier European plagues.

Timothy Ray Brown was suffering from both HIV and leukemia, and during the course of treatment for his leukemia, researchers in Berlin realized there might be a way to take advantage of rs333’s ability to resist HIV. After chemotherapy and radiation treatments, the researchers transplanted bone marrow from a donor with 2 copies of the rs333 deletion. These donor cells were able to reconstitute his blood.

The transplant took place in 2007, and doctors waited and watched for years while his body cleaned out the HIV and repaired itself. Finally publishing in 2010 and 2011, they were able to conclude that the 'SNP transplant' cured their patient’s HIV PMID 21148083Evidence for the cure of HIV infection by CCR5Δ32/Δ32 stem cell transplantation”.

Now in 2012 a Phase 2 clinical trial has begun recruiting patients. The protocol indicates donors “must be a 7/8 or 8/8 match at HLA-A, -B, -C, and -DRB1”. This is the same sort of donor matching we described five days ago in Do you see what I MHC so perhaps the trial could also record the rs2281389 and rs1800795 genotypes..

Despite the urgent need for a cure for HIV infection, “the risks associated with chemotherapy and radiation, and the relatively low frequency of ccr532 homozygous individuals, makes it unlikely that allogeneic HSC transplants using cells from ccr532 homozygous donors will become a widespread treatment option, and has prompted attempts to mimic the genetic knockout”.(PMID 22470838)

So this type of transplant only directly applies to the fortunately rare cases where a patient has to undergo chemotherapy and radiation to treat their leukemia, enabling a stem cell transplant to repopulate their blood system. It isn’t a practical solution for large numbers of patients. But a ‘SNP transplant’ can be a cure.

Monday, December 24, 2012

Deck the halls with boughs of HOXB

Prostate cancer ranks as the most commonly diagnosed non-skin cancer in Caucasian men, and it’s the most heritable common cancer. Genome-wide association studies (GWAS) have yielded perhaps 30 – 40 SNPs each associated with a (rather small) increase in risk, and so most inherited risk remains unexplained. A paper published early this year (PMID 22236224) about rs138213197 may have begun a new chapter in the search for inherited prostate cancer risk, and it’s indicative of how new methods can bring about highly significant discoveries.

For almost a decade since linkage evidence emerged indicating that a predisposition to prostate cancer was due to variants in the ch 17q21-22 region, GWAS studies yielded little from the region. The labs of K. Cooney (U Mich) and W. Isaacs (JHU) turned to exome sequencing – specifically, they looked at over 2000 exomes from 200 genes spanning around 15.5MB of ch 17q21-22, and they did this for almost 100 prostate cancer patients. In these exomes, each patient had around 12 novel variants, along with almost 700 already reported in dbSNP. Due to its occurrence in several patients and the role of HOXB13 as a transcription factor involved in prostate development, HOXB13 variant rs138213197(T) was identified. Based on a subsequent case-control study, where this SNP was seen in 72 of ~5,000 patients but in only 1 person out of 1400 controls, the researchers concluded that although the variant is clearly rare, it is recurrent. A carrier’s odds of developing prostate cancer are much higher - 10 to 20 fold - compared to a non-carrier’s risk. Since the publication of this finding early this year, already at least five other labs have confirmed the rs138213197 association with prostate cancer risk in independent populations, totaling over 20,000 patients. While the odds for increased risk are bit lower (ranging from 3 to 8 fold) in these studies, they remain far higher than for previously known markers.

In addition to its implications for prostate cancer diagnosis and treatment, the finding of a “rare but recurrent” variant with such a strong effect on cancer risk supports the concept that a significant fraction of the heritability of common diseases will indeed be due to quite rare variants. These variants will not have been detectable by GWAS studies (or the meta-analyses based on them, for that matter), but will now come to light thanks to exome and genome sequencing. Whether it is indeed due to our species population growth from a few million to 7 billion in just 400 generations or for other reasons (1), we each carry a different set of quite rare alleles with quite profound effects.

Sunday, December 23, 2012

Do they know it’s Christmas?

Myriad is in the middle of a multi-year defense of their BRCA gene patents. It has reached the Supreme Court and in 2013 we’re likely to get a significant ruling on the patentability of human genes. This year 23andMe stepped into the battle for the first time. Instead of patenting a whole gene, 23andMe was granted a patent on a specific SNP, rs10513789, and its use in predicting Parkinson’s risk. This may not have been too much of a shock to experts like Dan Vorhaus (who promptly posted a detailed and informative discussion here), but some customers felt this should have earned 23andMe a spot on Santa’s Naughty List. This prompted 23andMe to clarify that they “will not prevent others from accessing their genetic data or its interpretation”.

That’s probably reason enough to earn a spot here on our top 12 SNPs list, but this SNP wasn’t done making news. Late this year National Geographic began their second Genographic project. This time they’re checking 150,000 markers which are informative of ancestry. Medically relevant SNPs aren’t supposed to be part of the collection, but some SNPs have both medical and genealogical associations. In fact, nearly 500 of the NatGeo SNPs are already in SNPedia, many with medical consequences. To our great surprise, one them turns out to be … yes, you guessed it … 23andMe’s very own rs10513789.

Saturday, December 22, 2012

I'm dreaming of a white (matter) Christmas

In the almost 20 years since the discovery of the ApoE4 allele’s association with increased risk for late-onset Alzheimer’s disease, not a single additional SNP was discovered with a comparably strong effect … until this year. And the SNP discovered has interesting implications not only for Alzheimer’s but for how it bolsters a theory about diseases in general.

Publishing back to back in the November issue of NEJM (PMID 23150908, PMID 23150934), both the deCODE and Alzheimer Genetic Analysis Group teams found that a single copy of rs75932628(T), a SNP in the TREM2 gene, increases risk for Alzheimer’s about three-fold. This is comparable to the increased risk associated with an APOE4 allele. But it has a frequency of under 1%, too low to have been used or detected in most genome wide association (GWAS) studies, unlike the APOE4 allele’s frequency of about 15%.

The TREM2 gene encodes a protein known as “triggering receptor expressed on myeloid cells 2”, involved in microglial activity and inflammation. The rs75932638(T) allele somehow leads to reduced TREM2 activity, apparently reducing the microglial-cell based removal of beta amyloid. In addition to the quest for drugs that will reduce amyloid deposition directly, there’s now likely to be more support for drugs stimulating TREM2 activity as well as other molecules of the inflammatory cascade with roles in amyloid clearance (PMID:23237888).

It also turns out that homozygous recessive loss-of-function mutations in TREM2 have previously been associated with Nasu-Hakola disease, a very rare disorder involving early-onset dementia and bone fractures. The addition of another example where inheriting two loss-of-function mutations leads to a severe, early-onset disease, while inheriting only one leads to the late-onset of an apparently different disease, was noted by both sets of authors. It may well be that inheriting single loss-of-function SNPs will be the most common underlying risk factors for late-onset diseases (PMID 20813421).

Exome and genome sequencing are proving effective at uncovering these moderate risk, low frequency variants. In 2013 (and beyond) we hope to see more studies of late-onset diseases in the heterozygous relatives of children diagnosed with loss-of-function homozygous recessive diseases.

Friday, December 21, 2012

Do you know what I know?

Adam Foye’s SNPs have earned a spot on the 12 SNPs of Christmas 2012, even though by agreement we can’t name them publicly yet. The elves at SNPedia were one of the 23 teams taking part this year in the CLARITY Challenge. Sponsored by Boston Children’s Hospital, the goal was to yield useful information for three families with undiagnosable genetic disorders and document the process used. In each family, the genomes of the affected child (Adam Foye in the first family) and both parents were fully sequenced by Complete Genomics, and exome sequenced by a Life Technologies ABI SOLiD. In one family, several extended family members were also exome sequenced. Illustrating the difficulty of the challenge, despite the high volume of state of the art data, the recent press release revealed that so far only one of the three families has received a high confidence diagnosis.

Adam Foye has mutations in in the gene Gap junction beta-2 (GJB2) on chromosome 13. This means each of his parents carries one copy of a non-functional version of this gene and he inherited both of the non-functional forms. This resulted in sensorineural hearing loss and the need to wear hearing aids. GJB2 mutations are well known in the literature, and many have dbSNP identifiers. Nine of the most common are included in the DNA-chip based test from 23andMe.

This enables the identification of some, but not all, causes of this type of hearing loss. Carrier status - such as of Adam Foye’s parents - could also have been determined. Without genotyping the parents, however, the test wouldn’t have been conclusive. We wouldn’t have been able to see if both mutations were on the same strand, leaving one functional copy of the gene, or on different strands, damaging both copies.

Last week 23andMe announced that they’ve begun to perform phasing of their data. By using either of his parents (or their entire database), they can assign most variations as being either on the same or different strands. For the moment phased data isn’t downloadable by their customers, but maybe that will change. If not, there will be inexpensive genomic sequencing in a few years.

Unfortunately it wasn’t hearing loss that triggered Adam’s inclusion in the CLARITY Challenge; it was a more pressing medical concern, a form of muscular dystrophy. For this condition genotyping chip tests would have been insufficient. As publicly announced in CLARITY press releases, mutations in the Titin (TTN) gene were reported by several teams. This gene is named for Titan, the giant Greek god, and the name fits. It produces the largest known protein at 34,350 amino acids in length and has the most (363!) exons.

The winning team has been announced; it’s the coalition of five respected research institutes (Brigham and Women’s Hospital, Massachusetts General Hospital, Partners Laboratory for Molecular Medicine, Brown University, and Utrecht University), and although we were not part of that coalition, we are grateful to have participated. This sort of analysis is not the core strength of SNPedia or our analysis tool, Promethease. We are best suited for reporting on published variants in DTC samples, and communicating information about them to a wider audience. Much of CLARITY emphasized searching for novel mutations and communicating them to clinicians. But the gap between clinicians and the public continues to narrow, and we were curious to see where our tools might be applicable.

Even before this contest, Promethease was able to directly work with much of the data provided by Complete Genomics. Their dbSNP file concisely describes all genotypes already known to dbSNP. That will never be able to catch a new de-novo mutation, but with the October completion of the 1000 Genomes Project dbSNP now covers over 38 million validated mutations. As 2012 draws to a close, 35,600 of them have associated literature in SNPedia. These are the SNPs which are easiest to interpret. For GJB2 we currently have 53 annotated SNPs, and for TTN 5. With time those numbers will continue to rise.

The exome data was more challenging. It was just a collection of short DNA reads, not yet assembled or aligned to the reference genome. Pre-assembled data would have been easier to work with, but might have also contained some mis-assemblies. The reference genome is not a universal truth. In SNPedia's Nightmare Before Christmas we discussed how it varies over time as the community learns more. Promethease doesn’t do this sort of de-novo assembly, and we are not likely to add that feature anytime soon.

Completing this challenge pushed us into new capabilities such as support for a new Family Trio report, which works with all 38M snps, not just the 35k in SNPedia. While it was developed to help us process the CLARITY data, it became publicly available as of Promethease v0.1.144 over 3 months ago. It also made us aware of areas for improvement, such as in detecting CNVs or compound heterozygosity.

In time we expect to be able to share more results of the CLARITY Challenge and highlight the inspiring work of the other teams. Most importantly, even though the contest submission period has closed, we are hoping that the availability of genomic sequence from these families along with improved methods may yet yield answers for the other two families.

Thursday, December 20, 2012

Do you see what I MHC?

Two independent SNPs discovered this year, rs2281389 and rs1800795, may indicate that the last decade of inconsistent genome-wide association studies hunting for ways to reduce blood cell transplant failures was not for naught. 

Humans have a lot of genetic variation in the human leukocyte antigen (HLA) and this variation creates a somewhat unique signature on the surface of your cells. Your immune system relies on this signature to distinguish its own cells from foreign invaders, and then to target foreigners for attack. This is a great system when you’re trying to fight off a cold, but a problem when you’re trying to find a donor for an organ or tissue transplant. This is why donor registries screen for similar HLA signatures, to minimize the odds of a conflict. There are immune suppressing drugs that lower the odds of rejection, but these come with their own risks. When you’re in need of an organ or tissue donation, you need your immune system working well to help you, not for it to be suppressed.

Graft versus host disease (GVHD) is when the donated cells recognize the recipient as foreign and begin to attack. This affects up to 80% of patients and is a major cause of transplant failure, early complications and death. In the 2012 paper PMID 22837536 researchers looked for SNPs spanning the Major Histocompatibility Complex (MHC) region which didn’t match between 4,000 donor/recipient pairs who were perfectly matched across the 5 major HLA loci. The most significant SNP found, rs2281389, raised the risk and severity of acute GVHD by about 40%. Curiously, it’s the mismatch itself - not any of the possible genotypes - that increases GVHD risk.

Outside the MHC, an independent comprehensive study (PMID 22282500) of 1300 allogeneic hematopoietic cell transplantation (HCT) donors and recipients using previously reported GVHD-associated SNPs concluded that the best replicating - and almost only - variant was IL6 gene SNP rs1800795. This SNP was associated with a 20%-50% increased risk for GVHD, and here, there may be a biological explanation since the rs1800795(G) allele has been associated with increased serum levels of IL-6 and several autoimmune and inflammatory diseases.

Upcoming clinical trials will determine if using these SNPs for improved genetic matching of donors and recipients will lower the incidence of GVHD. And when full genome sequence becomes a routine part of one’s medical record, these type of studies, along with the predictive HLA-typing, will greatly increase the number of potential donors while minimizing rejection risks. As with many Christmas gifts, the new ones will complement and co-exist with the old ones.

Wednesday, December 19, 2012

Christmas in the Heart

rs11591147 is a SNP that warms the heart. This PCSK9 gene SNP has been known since 2005 to be associated with reduced average low density lipoprotein cholesterol (LDL-C), but this year it became the poster child of a set of 9 SNPs used to tackle a key question: to reduce heart disease risk, just how early in life should we start paying attention to LDL-C levels?

The meta-analysis of over 300,000 (!) patients published by Ference et al. [PMID:23083789] concludes that a person who’s lifetime average LDL exposure was reduced by about 40 mg/dl due to one or more of these 9 SNPs has their coronary heart disease risk cut in half. Their models suggest that the earlier you experience lower LDL levels, the better. If you weren’t born with enough of the LDL-lowering variants like rs11591147(T), the sooner you lower your LDL, the better, either by changing your lifestyle (through your diet) or by taking statins. Atherosclerosis isn’t just for old folks – it’s a progressive disease that begins in childhood.

This reminds us of rs4149056, a SNPedia Top 10 SNP pick of 2010. rs4149056(C) alleles and especially rs4149056(C;C) genotypes are associated with higher risk for side effects (muscle pain and degradation) when taking the statin simvastatin. This year, the Clinical Pharmacogenomics Implementation Consortium (CPIC) issued a clinical guideline with dosing recommendations for simvastatin when rs4149056 status is available [PMID: 22617227]. Pharmacogenomics continues to be one of the areas where genomics is directly improving patient health and well-being, and that too warms our hearts.