The R47H mutation in the TREM2 gene triples a person’s risk of Alzheimer’s disease, and for a while researchers thought they had discovered why. Using CRISPR to introduce the mutation in mice, two independent groups reported that it dramatically diminished TREM2 expression. A third model, described September 10 in Molecular Neurodegeneration, confirmed this and uncovered the mechanism: The R47H mutation bungles splicing of the mouse gene, leading to a stunted transcript destined for proteolysis. The rub? This only happens in the mouse version of the gene. The researchers, led by Christian Haass at the German Center for Neurodegenerative Diseases in Munich, reported that in people with the R47H mutation, TREM2 splicing and expression were completely normal.

  • R47H mutation in TREM2 dampens expression of the mouse gene.
  • Suppression linked to alternative splice site.
  • CRISPR-induced silent mutations exacerbate the effect.

The results point to an inconvenient truth: mouse TREM2 R47H models cannot provide reliable insight into how the mutation alters TREM2 function in human carriers, the researchers contend. “This is just bad luck,” Haass told Alzforum. Mouse models that use a humanized version of the gene will be needed to study the biology of the mutation, he suggested.

In a joint comment to Alzforum, the researchers who had generated two other R47H-TREM2 models weren’t so quick to write them off completely. They pointed out that homozygous animals do express a significant amount of the correctly spliced gene, meaning that they may yet yield some information about how the mutation affects TREM2 function.

In the six years since researchers linked the R47H-TREM2 variant to AD, mounting evidence from cell culture models has indicated that the mutation stifles the microglial receptor’s interactions with ligands, such as phospholipids, ApoE, and Aβ (Nov 2012 news; Oct 2015 news; Kober et al., 2016). TREM2-deficient mouse models suggest the receptor is crucial for activating microglia and containing Aβ plaques (Feb 2015 news; May 2016 newsJun 2017 newsJul 2018 conference news). 

To learn how R47H affects TREM2 function in vivo, several groups have developed mouse models expressing the variant. Researchers led by Marco Colonna at Washington University in St. Louis used a bacterial artificial chromosome (BAC) to express the human gene, with or without the mutation, in mice lacking their own TREM2. The R47H-TREM2 bound poorly to ligands, and in mice crossed to 5xFAD mice, the mutated receptor failed to rally microglia to surround and contain plaques (TREM2, humanized (R47H) X 5XFAD mouse model; Jan 2018 news). However, the BAC model expressed eight copies of the human transgene, making it difficult to distinguish between effects of R47H and effects of overexpression.

To avoid the latter, four groups used CRISPR-Cas9 to introduce R47H directly into endogenous mouse TREM2. They were led by Bruce Lamb and Gary Landreth at Indiana University School of Medicine in Indianapolis (Trem2 R47H KI (Lamb/Landreth) mouse modelCheng-Hathaway et al., 2018); the MODEL-AD consortium at the Jackson Laboratory in Bar Harbor, Maine (Trem2 R47H KI (JAX)); Shilpa Shambashivan and colleagues at Amgen (Cheng et al., 2018); and now Haass’ group.

In addition to the G-to-A point mutation required to generate R47H, the Munich group introduced three silent mutations: GA-to-TC mutations to create a restriction site to ease genotyping; and a G-to-A mutation to boost CRISPR gene editing efficiency. First author Xianyuan Xiang found that heterozygous R47H-TREM2 mice had about 40 percent less TREM2 mRNA in their brains than wild-type, while homozygous R47H-TREM2 mice had about 70 percent less. In the heterozygotes, only the mutant copy was reduced. Similarly, microglia from the mutant mice expressed less TREM2 RNA and made less TREM2 protein than cells from wild-type mice. Xiang came to similar conclusions when she measured TREM2 expression in the R47H-TREM2 mice generated at Jackson Labs.

Analysis of TREM2 RNA transcripts pointed to splicing gone awry. Using RT-PCR, Xiang found that the R47H mutation, which lies in the second exon, introduced a new splice site. This resulted in removal of 119 base pairs at the 5' end of exon 2, introducing a premature stop codon that created a stunted transcript. The researchers speculated that this short transcript would make fodder for nonsense-mediated decay, a cellular surveillance mechanism that gobbles up transcripts with premature stop codons. In support of this idea, RT-PCR amplified a only a small amount of the short TREM2 transcript, even in the R47H TREM2 homozygotes. The researchers did detect traces of the correctly spliced transcript as well, suggesting some residual normal splicing still occurred. They also analyzed splicing patterns in the R47H-TREM2 mice generated at Jackson Lab, and found both transcripts in those mice as well, although the Jackson Lab mice appeared to have a greater proportion of the correctly spliced transcript than the Haass mice did.

In a joint comment to Alzforum, Landreth, Lamb, and Michael Sasner of Jackson Labs confirmed that the splicing defects occurred in their two R47H-TREM2 models (see full comment below). Amgen did not respond to requests for comment about its model.

Itching to see whether R47H TREM2 expression in people also fell prey to splicing antics, Xiang and colleagues expressed human and mouse TREM2 genes in human embryonic kidney (HEK) 293 cells. The R47H mutation triggered the same splicing defects in mouse TREM2 as it had in the CRISPR mice. Additional silent mutations introduced to make CRISPR more efficient exacerbated the problem. In stark contrast, neither R47H nor the silent mutations affected splicing of the human TREM2 gene. Moreover, TREM2 was spliced normally in induced microglia derived from R47H-TREM2 mutation carriers, and also in a postmortem brain sample from an AD patient with the R47H TREM2 mutation. Xiang also found that human TREM2 was spliced normally in Colonna’s BAC model.

Haass told Alzforum that to his mind, these findings render all the R47H-TREM2 CRISPR models unsuitable for studying human TREM2 biology. Rather than modeling functional effects of the R47H variant, the mice more closely model TREM2 deficiency, he said. Moving forward, researchers will need to employ different strategies, such as generating knock-in mice that express human TREM2 variants under control of the endogenous mouse promoter, he said.

Lamb, Landreth, and the MODEL-AD consortium scientists acknowledge that better models are needed. Even so, they think the R47H-TREM2 CRISPR models have some use. They noted that although the majority of the mutant TREM2 transcript is mis-spliced, their mice still produce some full-length mutant transcript. They are comparing whole-brain transcriptomes of TREM2-deficient mice to those of R47H CRISPR mice; this should provide some insight into specific effects of the R47H allele, they wrote.

The MODEL-AD consortium is considering strategies to generate TREM2 R47H models that splice the transcript normally, and seeks ideas from the scientific community on how to make that happen.—Jessica Shugart


  1. Xiang et al. report altered splicing of mouse TREM2 that is dependent upon the presence of the R47H mutation that was introduced into mice by three different groups (Christian Haas and colleagues, Bruce Lamb/Gary Landreth, and the MODEL-AD Consortium). Using whole-brain transcriptomics, we have confirmed an alteration in splicing in both the Lamb/Landreth and MODEL-AD R47H mice at multiple ages up to 12 months that involves a novel splice site and removal of 119 base pairs of the coding sequence of TREM2. Frances Edwards from the Dementia Research Institute, London, who kindly shared her analysis of the MODEL-AD R47H mice, has reported the same splicing. The alternative isoform is generated at 10 percent of the levels of full-length TREM2, likely due to nonsense-mediated decay. Given that all three R47H TREM2 mouse strains contain slightly different silent mutations induced as part of CRISPR targeting, it seems likely that the alternative splicing is due primarily to the R47H allele.

    Xiang and colleagues report that the human R47H TREM2 variant is not spliced in this fashion, and we also have not observed altered splicing of human TREM2 via transcriptomic analysis of brain RNA from an R47H carrier. Importantly, however, in both the Lamb/Landreth and MODEL-AD homozygous R47H knock-in mouse strains it appears that full-length Trem2 transcripts are still expressed at ~50 percent of wild-type levels based upon RNA-seq analysis. Therefore, it is surprising and intriguing that Xiang et al. observe almost no detectable protein in microglia homozygous for R47H TREM2.

    We are currently performing analyses of brain protein levels of TREM2 to determine if brain TREM2 protein expression is similarly reduced in the Lamb/Landreth and MODEL-AD R47H models. Furthermore, we are also performing whole-brain transcriptomic analysis comparing heterozygous and homozygous Trem2 KO mice to heterozygous and homozygous R47H knock-in mice, which should provide further insight into the specific effects of the R47H allele.  For now, we argue that the conclusions that “functional data derived from Trem2 R47H knock-in mice cannot be translated to humans” put forward in this manuscript may be a bit premature because full-length transcripts containing the R47H mutation are still generated in these models, albeit at reduced levels. Notably, all of the strategies to generate TREM2 R47H mice have unique strengths and weaknesses, which makes it difficult to exactly match expression levels of wild-type and R47H-containing TREM2 and/or examine the potential impact on expression of other TREM and TREM-like genes within the locus. The MODEL-AD consortium is currently considering alternative strategies to generate TREM2 R47H models that do not exhibit the effects on mouse TREM2 splicing, and we look forward to obtaining input from the scientific community as we move forward with these efforts.

    By Bruce Lamb, representing the IU/JAX MODEL-AD Center.

Make a Comment

To make a comment you must login or register.


Mutations Citations

  1. TREM2 R47H

News Citations

  1. Enter the New Alzheimer’s Gene: TREM2 Variant Triples Risk
  2. Alzheimer’s Risk Genes Interact in Immune Cells
  3. TREM2 Buoys Microglial Disaster Relief Efforts in AD and Stroke
  4. Barrier Function: TREM2 Helps Microglia to Compact Amyloid Plaques
  5. Hot DAM: Specific Microglia Engulf Plaques
  6. TREM2: Diehard Microglial Supporter, Consequences Be DAMed
  7. New Mouse Models Reveal Unexpected Property of TREM2

Research Models Citations

  1. TREM2, humanized (R47H) X 5XFAD
  2. Trem2 R47H KI (Lamb/Landreth)
  3. Trem2 R47H KI (JAX)
  4. Trem2 R47H KI (Haass)

Paper Citations

  1. . Neurodegenerative disease mutations in TREM2 reveal a functional surface and distinct loss-of-function mechanisms. Elife. 2016 Dec 20;5 PubMed.
  2. . The Trem2 R47H variant confers loss-of-function-like phenotypes in Alzheimer's disease. Mol Neurodegener. 2018 Jun 1;13(1):29. PubMed.
  3. . TREM2-activating antibodies abrogate the negative pleiotropic effects of the Alzheimer's disease variant Trem2 R47H on murine myeloid cell function. J Biol Chem. 2018 Aug 10;293(32):12620-12633. Epub 2018 Mar 29 PubMed.

Further Reading

Primary Papers

  1. . The Trem2 R47H Alzheimer's risk variant impairs splicing and reduces Trem2 mRNA and protein in mice but not in humans. Mol Neurodegener. 2018 Sep 6;13(1):49. PubMed.