In the complex genetics of Alzheimer’s, only a handful of genes by themselves have a massive impact on risk. ApoE4 and TREM2 boost a person’s odds of late-onset AD from three- to 12-fold, and the three autosomal-dominant genes APP, PS1, and PS2 bring on AD with near certainty. All other AD genes confer much smaller odds. Or so we thought. Growing evidence now suggests that a sixth gene, the endocytic receptor SORL1, deserves a place among this set. Several recent studies have identified pathogenic SORL1 variants that segregate with the disease in families. Now, researchers led by Richard Mayeux at Columbia University in New York report finding 17 rare loss-of-function SORL1 variants in a large whole-exome sequencing study. The variants occurred exclusively in people with cognitive impairment or AD, never in healthy controls. This strengthens the case that they cause disease. AD patients carrying these mutations developed the disease seven years earlier on average than did noncarriers. The findings appeared online May 24 in the Annals of Clinical and Translational Neurology. “This elevates SORL1 to one of the major AD risk genes,” Mayeux said.

  • Rare loss-of-function mutations in SORL1 appear to cause AD.
  • More common SORL1 missense mutations boost risk as much as ApoE4.
  • This makes SORL1 one of the top six AD risk genes.

Other researchers in the field agreed. “There is no longer any doubt SORL1 is a major genetic determinant of Alzheimer's disease, probably by participating in a central pathophysiological pathway,” Jean-Charles Lambert at Institut Pasteur de Lille, France, wrote to Alzforum. SORL1 protein, also known as SORLA or LR 11, shunts amyloid precursor protein (APP) into non-amyloidogenic processing pathways, and loss of SORL1 leads to higher Aβ levels in cell culture experiments. “If this were to be confirmed [in vivo], it would obviously be additional genetic evidence supporting the amyloid cascade hypothesis,” Lambert added (full comment below).

SORL1 first turned up as an AD risk gene in GWAS studies (Rogaeva et al., 2007), and SNPs in or around the gene were found to associate with AD in multiple ethnic groups (Lee et al., 2007). Brain SORL1 levels are low in people with mild cognitive impairment or AD (Scherzer et al., 2004; Dodson et al., 2006; Sager et al., 2007). Meanwhile, a protective SNP boosts SORL1 expression and lowers Aβ in cell culture (Mar 2015 news). The gene was first implicated in familial AD when French geneticists led by Dominic Campion identified SORL1 variants in seven out of 29 families with unexplained early onset AD (Apr 2012 news). 

Moving Up? SORL1 belongs near higher-impact variants, not GWAS hits, said Richard Mayeux, who placed a red star in this diagram to estimate an aggregate location for all SORL1 variants. [Image adapted from Manolio et al., 2009.] 

For their part, Mayeux and colleagues did not set out to look for SORL1 variants. They were simply searching for some of the missing heritability of AD, i.e., that large fraction of cases that is not accounted for by the genetic variants known to date. First author Neha Raghavan performed an unbiased screen of whole-exome sequencing data from people with AD and healthy controls participating in the Washington Heights-Inwood Community Aging Project (WHICAP) and the Alzheimer’s Disease Sequencing Project (ADSP). WHICAP follows people who are 65 or older and initially free of dementia, while ADSP enrolls cases and controls 60 and older. The authors also added data from 6,395 controls seen at Columbia University, for a total of 6,965 cases and 13,252 controls in the study.

Genetic variants that exert large effects on disease tend to be rare, perhaps because natural selection weeds the most damaging mutations out of the population. For this reason, Raghavan and colleagues focused on ultra-rare coding variants present in the general population in fewer than one person in 10,000. They looked for any loss-of-function mutations that were enriched in cases versus controls. To increase their chance of finding genes, they combined the signal from different loss-of-function variants in the same gene, a technique known as gene-based collapsing analysis.

Using these methods, variants in only one gene—SORL1—achieved exome-wide significance. Nineteen of the AD cases carried such a variant. One control did too, but that person had mild cognitive impairment, suggesting AD might be afoot. Most variants occurred in only one person in the cohort, for a total of 17 different SORL1 loss-of-function mutations.

As in other studies, SORL1 mutations were not confined to a particular ethnicity, appearing equally often in Caucasians, African-Americans, and Caribbean Hispanics. As additional evidence of their pathogenicity, Mayeux noted that these mutations are virtually absent from the genomes of cognitively healthy people, with only one of the 17 appearing in the Exome Aggregation Consortium (ExAC) database, which selects against childhood genetic diseases but not age-related diseases. None of the disease-associated SORL1 variants appeared in the smaller but more stringent Healthy Exome (HEX) database maintained by Alzforum.

Commenters consider the data in this paper to be strong. “This is an exciting paper, which, together with previous studies, clearly motivates further studies on SORL1 as an important player in AD pathogenesis,” Caroline Graff at the Karolinska Institute in Stockholm wrote to Alzforum (full comment below). Håkan Thonberg, also at Karolinska, added that the paper adds evidence for SORL1 being causative for AD.

Mayeux noted that this is the first population-based study to demonstrate a genome-wide association between SORL1 loss-of-function mutations and AD. Previous studies of highly pathogenic mutations focused on families with early onset disease (Verheijen et al., 2016; Thonberg et al., 2017; Gómez-Tortosa et al., 2018). A recent case report describes the first person known to be missing both copies of SORL1. He developed AD at 55, younger than when either of his parents had, though still in the same range as SORL1 heterozygotes (Le Guennec et al., 2018). 

SORLA (red) occupies neuronal vesicles in mouse hippocampus. [Image courtesy of Caglayan et al., 2014, Science Translational Medicine/AAAS.]

Given these findings, should SORL1 be considered autosomal-dominant? Researchers have discussed this question for a while (Aug 2017 news), but told Alzforum they still need to analyze more extensive family trees before they can draw this conclusion. “We don’t know yet if it is 100 percent penetrant,” Mayeux cautioned. Nonetheless, Henne Holstege at VU University Medical Center in Amsterdam believes that more data will ultimately confirm SORL1 as the fourth autosomal-dominant AD gene, with at least some of its mutations being fully penetrant. She compared SORL1 to presenilin 2. Not all of those mutations are fully penetrant, either.

Other SORL1 variants have less severe effects. Some missense mutations, which change a single amino acid, also associate with AD. Mayeux and colleagues reported that SORL1 missense mutations were enriched in cases versus controls in their study, appearing in 1.8 percent of the former and 1 percent of the latter. Other studies have associated specific SORL1 missense mutations with late-onset AD; notably, these variants boosted Aβ levels in cell culture (Cuccaro et al., 2016; Vardarajan et al., 2015). In a family with several ApoE4 homozygote members, inheriting a SORL1 missense mutation in addition to ApoE4 appeared to trigger earlier AD onset and increase the penetrance of ApoE4 (Louwersheimer et al., 2017). Holstege calculated in her recent study that rare, pathogenic SORL1 missense mutations confer about a 12-fold increased risk of AD, similar to ApoE4 homozygosity (Holstege et al., 2017). 

In this, SORL1 follows a pattern seen for other GWAS hits, for example ABCA7, noted John Hardy at University College London. With the ABCA7 gene, rare loss-of-function mutations have far stronger effects than the common variants in the same gene found in large association studies. Researchers suspect that more such genes will be found, and help explain the heritability of AD. Some researchers, including Hardy, expressed surprise that Mayeux’s study turned up only SORL1. “It’s interesting that the yield of novel genes was so low,” Hardy wrote (full comment below).

Why did the association between SORL1 and familial AD take such a long time to find? Ideally, scientists want to determine penetrance for each variant in large and informative families. In the case of SORL1, such pedigrees have been hard to come by thus far. But evidence is accumulating. “In clinical files of SORL1 mutation carriers in our clinic, we often see early onset AD in the first and second degree,” Holstege wrote. She added, “Nevertheless, proving that the SORL1 mutation segregates in these families can still be complicated, as not all family members can be approached for genetic testing or, in some cases, hesitate to participate in genetic segregation studies.”

In her recent study, only the rarest SORL1 variants were pathogenic, suggesting they are selected against. “That was surprising to us, given that AD develops after the reproductive phase. Maybe SORL1 also plays a key role earlier in life,” she told Alzforum.

Each individual SORL1 loss-of-function mutation may be rare, but in the aggregate, they account for more AD cases than do presenilin 1 and 2 combined, Holstege said. Her study identified rare, damaging SORL1 mutations in 2 percent of AD cases. Mayeux noted that SORL1 mutations are more common than pathogenic TREM2 variants, as well, though much less common than ApoE4.

Why are some missense variants pathogenic, and others not? Answering this could open a window into SORL1 biology and suggest therapeutic targets, Jessica Young at the University of Washington in Seattle wrote to Alzforum (full comment below). Mayeux is doing such studies now, expressing pathogenic SORL1 variants in cell lines to examine their functional effects. SORL1 forms part of the retromer, a protein-sorting complex, where it helps direct APP from endosomes back to the trans-Golgi network. This transport prevents cleavage of APP into Aβ (Mar 2005 conference news; Sep 2007 news). SORL1 is also believed to route Aβ peptides to the lysosome for degradation (Feb 2014 news). SORL1 GWAS variants were recently linked to higher amyloid PET signals in MCI and dementia (Jan 2018 news). 

All of these recent data raise the question of whether, and how, to use SORL1 in clinical practice. Many clinicians are loath to use rare mutations until they are clearly shown to be causal. Thonberg, a clinical geneticist at Karolinska, believes this is warranted. “Caution must still be taken when it comes to using this finding in clinical settings, and our advice is not to use it at all for risk assessment,” he wrote to Alzforum (full comment below).

Now that SORL1 pathogenic mutations have been identified, however, many clinicians report finding them in patients, Holstege said. Several have asked her for guidance on how to treat these patients, but the answer is unclear at present. “What can we offer these patients? Should we make them eligible for enrollment in clinical trials and observational studies like DIAN? We need to start this discussion, because we might find more genes like this,” Holstege suggested.—Madolyn Bowman Rogers

Comments

  1. This work is based on an impressive number of sequencing data generated from Alzheimer’s patients and controls, and primarily analyzes loss-of-function (LOF) variants. The major result obtained in this analysis mainly replicates the association of rare variants of SORL1 with a particularly high risk of developing Alzheimer’s disease. These data thus confirm that LOF variants in SORL1 are particularly deleterious (Pottier et al., 2012; Nicolas et al., 2015; Bellenguez et al., 2017; Verheijen et al., 2016; Holstege et al., 2017). 

    Interestingly, these data also indicate that this type of very rare variant is found in cases regardless of their ethnic origin. There is thus no longer any doubt SORL1 is a major genetic determinant of Alzheimer's disease, by probably participating in a central pathophysiological pathway.

    In addition, common variants associated with decreased expression of SORL1 are also associated with increased risk of disease in GWAS studies (Rogaeva et al., 2007; Lambert et al., 2013; Young et al., 2015). It seems thus that a complete loss of function is particularly impacting, leading probably to earlier forms, while a decrease in expression leads to a low level of risk, likely in connection with a more insidious biological impact.

    Currently, it has been proposed that SORL1 underexpression modifies APP processing and potentially leads to Aβ peptide overproduction (Rogaeva et al., 2007). If this were to be confirmed, it would obviously be additional genetic evidence supporting the amyloid cascade hypothesis.

    Of note, beyond SORL1, three other genes could be of interest but these must be validated in independent populations. This indicates that the detection of the association of rare variants, as expected, requires significant statistical power to be able to detect new signals. Strong efforts are still needed.

    References:

    . High frequency of potentially pathogenic SORL1 mutations in autosomal dominant early-onset Alzheimer disease. Mol Psychiatry. 2012 Apr 3; PubMed.

    . SORL1 rare variants: a major risk factor for familial early-onset Alzheimer's disease. Mol Psychiatry. 2016 Jun;21(6):831-6. Epub 2015 Aug 25 PubMed.

    . Contribution to Alzheimer's disease risk of rare variants in TREM2, SORL1, and ABCA7 in 1779 cases and 1273 controls. Neurobiol Aging. 2017 Nov;59:220.e1-220.e9. Epub 2017 Jul 14 PubMed.

    . A comprehensive study of the genetic impact of rare variants in SORL1 in European early-onset Alzheimer's disease. Acta Neuropathol. 2016 Aug;132(2):213-24. Epub 2016 Mar 30 PubMed.

    . Characterization of pathogenic SORL1 genetic variants for association with Alzheimer's disease: a clinical interpretation strategy. Eur J Hum Genet. 2017 Aug;25(8):973-981. Epub 2017 May 24 PubMed.

    . The neuronal sortilin-related receptor SORL1 is genetically associated with Alzheimer disease. Nat Genet. 2007 Feb;39(2):168-77. PubMed.

    . Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet. 2013 Dec;45(12):1452-8. Epub 2013 Oct 27 PubMed.

    . Elucidating molecular phenotypes caused by the SORL1 Alzheimer's disease genetic risk factor using human induced pluripotent stem cells. Cell Stem Cell. 2015 Apr 2;16(4):373-85. Epub 2015 Mar 12 PubMed.

  2. When the first rare protein-damaging SORL1 variants were identified by our group, the families were selected with the aim to identify a fourth autosomal-dominant gene. Probands presented with early onset AD (EOAD, onset before 65 years) and a positive family history of EOAD in first- and second-degree relatives (Pottier et al., 2012). However, segregation data was lacking to conclude about pathogenicity in the context of a putative monogenic inheritance.

    Since then, we and others have reported such variants in other families with diverse patterns of family history, including presence of other early onset cases, late-onset cases, and no family history of AD. Recently, we reported the case of a patient with biallelic SORL1 protein-truncating variants and a family history of both early and late-onset AD in the parents (Le Guennec et al., 2018). In addition, protein-truncating variants have been identified in population databases such as ExAC, although extremely rare (estimated number of carriers from allele counts: 41 out of ~60,000) and exceptionally in controls from case-control studies. The spectrum associated with carrier status of two or one protein-truncating SORL1 variants hence extends from familial early or late-onset AD, sporadic early or late-onset AD, to the status of control.

    We previously showed that rare SORL1 protein-truncating and missense, predicted-damaging variants, most of which are singleton variants, significantly increase the risk of EOAD with exome-wide significance in the ADES-FR study (p<10-6) (Nicolas et al., 2016; Bellenguez et al., 2017). They have a significantly greater effect on EOAD risk than on late-onset risk. In our latest work, 1.29 percent of EOAD cases (0.62 percent of all AD cases) carried a singleton protein-truncating variant, compared to one out of 1,273 controls (OR for EOAD: 16.22, 95 percent CI=[23.08-inf], p=3.4e-4).

    In this paper, Raghavan et al. took advantage of one of the largest existing case-control exome data sets and reached exome-wide significance for truncating variants. The proportion of protein-truncating variant carriers (0.27 percent) remains lower than in the ADES-FR study, probably because the ADSP case-control data set excludes cases with an age of onset  younger than 60 years. The extreme rarity of protein-truncating variants in controls in our case-control studies suggested a strong effect. In the Raghavan et al. data, the OR was estimated to be 36 (95 percent CI 5.8 – 1493.0), which is consistent with our previous results.

    Anyway, it is currently impossible to make a conclusion about the penetrance of SORL1 damaging variants without segregation data. Such data remain too limited in the literature to conclude full penetrance of some SORL1 variants. Even if we can provide enough evidence of Mendelian inheritance for some of the most damaging rare SORL1s, they remain less penetrant than most of the APP, PSEN1, and PSEN2 pathogenic variants. Up to now, SORL1 rare damaging coding variants are considered a strong risk factor.

    The word “pathogenic” should therefore not be used, nor genetic counseling proposed to families, until compelling co-segregation evidence has been reported in families.  

    Of note, the odds ratios of some TREM2 rare variants were even higher than SORL1’s in our latest case-control study (Bellenguez et al., 2017), suggesting that the same questions hold true for TREM2, although a Mendelian role is not under debate up to now.

    References:

    . High frequency of potentially pathogenic SORL1 mutations in autosomal dominant early-onset Alzheimer disease. Mol Psychiatry. 2012 Apr 3; PubMed.

    . Biallelic Loss of Function of SORL1 in an Early Onset Alzheimer's Disease Patient. J Alzheimers Dis. 2018;62(2):821-831. PubMed.

    . SORL1 rare variants: a major risk factor for familial early-onset Alzheimer's disease. Mol Psychiatry. 2016 Jun;21(6):831-6. Epub 2015 Aug 25 PubMed.

    . Contribution to Alzheimer's disease risk of rare variants in TREM2, SORL1, and ABCA7 in 1779 cases and 1273 controls. Neurobiol Aging. 2017 Nov;59:220.e1-220.e9. Epub 2017 Jul 14 PubMed.

  3. This work clearly confirms the work of Holstege and colleagues showing that loss-of-function variants in SORL1 are associated with a large increase in risk of AD. These mutations confer very high risk that can be considered close to penetrant.

    The authors continue the trend of such findings. The ABCA7 gene, too, is a GWAS hit for AD, for which loss-of-function variants have a larger effect on risk. In general this type of relationship is likely to be found in other genes among the GWAS hits.

    It’s interesting that the yield of novel genes was so low.

  4. Family segregation studies and linkage analysis would be needed to confirm if SORL1 is equivalent to APP or PS as an autosomal-dominant gene. However, this is tremendous work and clearly implicates loss of SORL1 function as a very strong risk factor for AD.

    This could lead to compelling insights into the biology of SORL1, not only in terms of its well-characterized interactions with APP but potentially other roles in endocytosis and vesicular trafficking in the brain. For example, both truncation variants and non-synonymous coding variants in SORL1 have been associated with AD, raising the question if these missense variants impair an essential function of SORL1?

    Since SORL1 variants are implicated in both early and late-onset AD, it would also be interesting to understand genetic interactions between SORL1 and other vesicular trafficking genes associated with increased AD risk.

    In general, this study provides strong evidence that SORL1 and related pathways are viable candidates for investigation into therapeutic development.

  5. The findings are interesting and add significance to previous studies on SORL1 implicated in AD. Some of those were cited and some were not, including our paper (Thonberg et al., 2017), which identifies three variants in SORL1 segregating with disease in familial AD with a mix of early onset and late-onset disease in the families.

    It would be interesting to know if the late-onset AD cases with deleterious SORL1 variants in the Mayeux paper have a positive family history and if other family members developed disease before 65 years, indicating an age-at-onset variability for SORL1 variants similar to our findings rather than, or in addition to, SORL1 being implicated in “sporadic late-onset AD.”

    In any event, it is an exciting paper, which together with previous studies clearly motivates further studies on SORL1 as an important player in AD pathogenesis.

    References:

    . Identification and description of three families with familial Alzheimer disease that segregate variants in the SORL1 gene. Acta Neuropathol Commun. 2017 Jun 9;5(1):43. PubMed.

  6. The paper by Richard Mayeux and colleagues is indeed very interesting. It adds evidence for SORL1 actually being causative for AD, and strengthens the role of SORL1 in APP/Aβ metabolism.

    However, caution must still be taken when it comes to using this finding in clinical settings. Our advice is not to use it at all for risk assessment or similar when facing patients. At the moment, we cannot place it alongside APP and presenilin genes, as it is still questionable whether it is a true cause of autosomal inheritance.

    We hope the research field will be able to add data in the very near future.

  7. It is exciting to see that this huge effort underlines previous evidence that predicted loss-of-function mutations in SORL1 are strong risk factors for Alzheimer's dementia, even in patients with later disease onset.

    Considering the lack of reports of these mutations from cognitively healthy individuals, these mutations are likely to have a high penetrance, and several studies have documented increased familial clustering of AD among carriers of SORL1 loss-of-function mutations that would be in line with this (e.g. Pottier et al., 2012; Nicolas et al., 2015; Verheijen et al., 2016; Vardarajan et al., 2015).

    The current paper does not provide insight in potential familial inheritance; it will be important to investigate this further to get a better appreciation of possible patterns of inheritance as well as modifiers of penetrance. The wide spread in onset age among carriers is suggestive of the latter, and more in-depth understanding of these characteristics will be required before we could consider predictive testing in a clinical setting. It also re-emphasizes the need to design assays to experimentally determine which of the many SORL1 missense mutations that are being identified in sequencing studies will lead to loss of SORL1 functionality with similarly strong pathogenic effects.

    Of note, the paper used very stringent inclusion criteria for variants, hence the lack of association for other genes may in part be attributable to selecting against less rare loss-of-function mutations, as observed in, e.g., ABCA7.

    References:

    . High frequency of potentially pathogenic SORL1 mutations in autosomal dominant early-onset Alzheimer disease. Mol Psychiatry. 2012 Apr 3; PubMed.

    . SORL1 rare variants: a major risk factor for familial early-onset Alzheimer's disease. Mol Psychiatry. 2016 Jun;21(6):831-6. Epub 2015 Aug 25 PubMed.

    . A comprehensive study of the genetic impact of rare variants in SORL1 in European early-onset Alzheimer's disease. Acta Neuropathol. 2016 Aug;132(2):213-24. Epub 2016 Mar 30 PubMed.

    . Coding mutations in SORL1 and Alzheimer disease. Ann Neurol. 2015 Feb;77(2):215-27. PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. AD Genetics—Problems and Promise
  2. Stem Cells Reveal Mechanism Behind Alzheimer’s Risk Factor
  3. New Genetic Insights Into AD: SORL1 and Natural Selection
  4. The Search for the Missing AD Heritability Turns Up New Rare Variants
  5. Sorrento: Sorting Out Shedding of Ectodomains
  6. Sorting Out SorLA—What Role in APP Processing, AD?
  7. SORLA Serves Up Aβ for Destruction
  8. Scanning Top 20 GWAS Hits for Amyloidosis: Slim Pickings

Paper Citations

  1. . The neuronal sortilin-related receptor SORL1 is genetically associated with Alzheimer disease. Nat Genet. 2007 Feb;39(2):168-77. PubMed.
  2. . The association between genetic variants in SORL1 and Alzheimer disease in an urban, multiethnic, community-based cohort. Arch Neurol. 2007 Apr;64(4):501-6. PubMed.
  3. . Loss of apolipoprotein E receptor LR11 in Alzheimer disease. Arch Neurol. 2004 Aug;61(8):1200-5. PubMed.
  4. . LR11/SorLA expression is reduced in sporadic Alzheimer disease but not in familial Alzheimer disease. J Neuropathol Exp Neurol. 2006 Sep;65(9):866-72. PubMed.
  5. . Neuronal LR11/sorLA expression is reduced in mild cognitive impairment. Ann Neurol. 2007 Dec;62(6):640-7. PubMed.
  6. . Finding the missing heritability of complex diseases. Nature. 2009 Oct 8;461(7265):747-53. PubMed.
  7. . A comprehensive study of the genetic impact of rare variants in SORL1 in European early-onset Alzheimer's disease. Acta Neuropathol. 2016 Aug;132(2):213-24. Epub 2016 Mar 30 PubMed.
  8. . Identification and description of three families with familial Alzheimer disease that segregate variants in the SORL1 gene. Acta Neuropathol Commun. 2017 Jun 9;5(1):43. PubMed.
  9. . SORL1 Variants in Familial Alzheimer's Disease. J Alzheimers Dis. 2018;61(4):1275-1281. PubMed.
  10. . Biallelic Loss of Function of SORL1 in an Early Onset Alzheimer's Disease Patient. J Alzheimers Dis. 2018;62(2):821-831. PubMed.
  11. . SORL1 mutations in early- and late-onset Alzheimer disease. Neurol Genet. 2016 Dec;2(6):e116. Epub 2016 Oct 26 PubMed.
  12. . Coding mutations in SORL1 and Alzheimer disease. Ann Neurol. 2015 Feb;77(2):215-27. PubMed.
  13. . Rare Genetic Variant in SORL1 May Increase Penetrance of Alzheimer's Disease in a Family with Several Generations of APOE-ɛ4 Homozygosity. J Alzheimers Dis. 2017;56(1):63-74. PubMed.
  14. . Characterization of pathogenic SORL1 genetic variants for association with Alzheimer's disease: a clinical interpretation strategy. Eur J Hum Genet. 2017 Aug;25(8):973-981. Epub 2017 May 24 PubMed.

Other Citations

  1. Healthy Exome

External Citations

  1. Exome Aggregation Consortium

Further Reading

Primary Papers

  1. . Whole-exome sequencing in 20,197 persons for rare variants in Alzheimer's disease. Ann Clin Transl Neurol. 2018 Jul;5(7):832-842. Epub 2018 May 24 PubMed.