Most functional studies of TREM2 have so far been performed with either total loss-of-function models or by studying the TREM2 T66M mutation [
7,
8,
11‐
13,
19,
20,
44,
45], which is associated with an FTD-like syndrome. Homozygous mice expressing the Trem2 T66M variant phenocopy a number of functional deficits also observed upon total loss of the Trem2 encoding gene [
11]. These include delayed resolution of inflammation upon lipopolysaccharide stimulation, reduced phagocytic activity, reduced microglial activation during physiological ageing and neurological insults, reduced cerebral blood flow, reduced cerebral brain glucose metabolism, impaired chemotaxis and clustering of microglia around amyloid plaques [
11]. However, much less is known about the functional impact of the TREM2 R47H variant, which is associated with a high risk for AD similar to that caused by the ApoE ε4 allele [
3,
4,
46]. In vitro studies suggested reduced binding of Aβ oligomers, ApoE, and phosphatidylserine due to structural alterations in TREM2 [
8,
21‐
24]. Furthermore maturation of the R47H variant within the secretory pathway may also be delayed [
25]. Expression of human TREM2 R47H in Trem2 knockout mice failed to rescue the knockout phenotypes again supporting the notion that TREM2 is protective and that TREM2 variants associated with neurodegenerative diseases may cause a loss-of-function [
14]. Trem2 R47H knock-in mice expressing this variant under physiological conditions resulted in phenotypes that were compatible with a loss-of function [
26]. Cheng-Hathaway et al. reported that Trem2 R47H heterozygous mice showed reduced Trem2 expression in microglia close to amyloid plaques, reduced microglial proliferation, reduced dense core plaques and increased neuritic dystrophy [
26]. These phenotypes were associated with reduced Trem2 mRNA expression [
26]. Similarly, Sudom et al. reported reduced Trem2 protein expression in brain as well as reduced sTrem2 in plasma from R47H knock-in mice [
24]. All together, this suggests that the AD-associated Trem2 R47H variant is also associated with a loss-of-function. However, while the T66M mutation causes a loss-of-function due to retention of the misfolded mutant protein within the endoplasmatic reticulum [
11,
25], the R47H mutation appears to cause haploinsufficiency due to reduced mRNA levels [
26]. We generated a similar mouse model using the CRISPR/Cas9 technology. Consistent with published findings, our mouse model also showed reduced Trem2 mRNA and protein levels. Furthermore, similar findings were made using an independent mouse generated by Jackson Laboratories. We therefore searched for a joined cellular mechanism, which may be involved in the significant reduction of Trem2 mRNA and protein in both mouse models. Surprisingly, we found that the introduction of the R47H variant by itself but even more so the introduction of additional silent mutations caused aberrant splicing. Direct sequencing of the aberrant splicing product revealed that it lacks 119 base pairs. Close inspection of the sequence of the alternative splice product revealed that a cryptic splice acceptor site within the exon 2 was activated. This leads to the elimination of parts of exon 2 resulting in a splicing product containing a premature stop codon. It is well known that such aberrant mRNAs are rapidly degraded by nonsense-mediated mRNA decay [
47]. Note that the TaqMan probes used for detecting Trem2 mRNA were either bound to exon 3/4 or exon 4/5 boundary (Fig.
1d) thus they detect both the full length functional Trem2 mRNA and the aberrantly spliced shorter variant. Therefore, the apparent degree of reduction for Trem2 mRNA and protein is not equal (Fig.
1d and
f). However, we cannot fully exclude the possibility that the R47H variant may also affect mouse Trem2 protein stability.
The mRNA reductions caused by aberrant splicing, which were consistent in two independent mouse models, led us to investigate if aberrant exon 2 splicing in humans also leads to TREM2 haploinsufficiency. We found that TREM2 mRNA levels and splicing patterns were both normal in iPSC-derived human microglia-like cells and in patient brains with the TREM2 R47H variant. Furthermore, cellular splicing assays using minigene constructs demonstrate that the R47H variant induced abnormal splicing only occurs in mice but not in humans. Thus, both the R47H variant as well as additionally introduced silent variants cause a mouse-specific reduction of Trem2 mRNA. Therefore, these findings cannot be translated to humans calling for novel humanized R47H mouse models.