TREM2 variants that cause early dementia and increase Alzheimer's disease risk affect gene splicing

Abstract

Loss-of-function variants in the triggering receptor expressed on myeloid cells 2 (TREM2) are responsible for a spectrum of neurodegenerative disorders. In the homozygous state, they cause severe pathologies with early onset dementia, such as Nasu-Hakola disease and behavioural variants of frontotemporal dementia (FTD), whereas heterozygous variants increase the risk of late-onset Alzheimer's disease (AD) and FTD. For over half of TREM2 variants found in families with recessive early onset dementia, the defect occurs at the transcript level via premature termination codons or aberrant splicing. The remaining variants are missense alterations thought to affect the protein; however, the underlying pathogenic mechanism is less clear. In this work, we tested whether these disease-associated TREM2 variants contribute to the pathology via altered splicing. Variants scored by SpliceAI algorithm were tested by a full-size TREM2 splicing reporter assay in different cell lines. The effect of variants was quantified by qRT-/RT-PCR and western blots. Nanostring nCounter was used to measure TREM2 RNA in the brains of NHD patients who carried spliceogenic variants. Exon skipping events were analysed from brain RNA-Seq datasets available through the Accelerating Medicines Partnership for Alzheimer's Disease Consortium. We found that for some Nasu-Hakola disease and early onset FTD-causing variants, splicing defects were the primary cause (D134G) or likely contributor to pathogenicity (V126G and K186N). Similar but milder effects on splicing of exons 2 and 3 were demonstrated for A130V, L133L and R136W enriched in patients with dementia. Moreover, the two most frequent missense variants associated with AD/FTD risk in European and African ancestries (R62H, 1% in Caucasians and T96K, 12% in Africans) had splicing defects via excessive skipping of exon 2 and overproduction of a potentially antagonistic TREM2 protein isoform. The effect of R62H on exon 2 skipping was confirmed in three independent brain RNA-Seq datasets. Our findings revealed an unanticipated complexity of pathogenic variation in TREM2, in which effects on post-transcriptional gene regulation and protein function often coexist. This necessitates the inclusion of computational and experimental analyses of splicing and mRNA processing for a better understanding of genetic variation in disease.

Keywords: alternative splicing; brain; exon skipping; gene dosage; microglia; neurodegeneration.

Figure 1

A reporter assay for analysis of TREM2 alternative splicing. (A) TREM2 transcript repertoire in human brain. FL = a full-length transcript (ENST00000373113) that encodes a transmembrane receptor. Δe4 (ENST00000338469) and Alt e4 (ENST00000373122) = isoforms with skipped or alternatively started exon 4, respectively, both encoding soluble polypeptides that lack the transmembrane domain. Δe2 = membrane-bound isoform that lacks IgV domain encoded by exon 2. Δe3 = isoform in which exon 3 skipping introduces a premature termination codon (PTC). Other minor transcripts result from skipping of multiple exons. Coding and non-coding regions are drawn in blue and grey, respectively. *Isoforms for which protein expression was not experimentally confirmed. (B) A reporter construct bearing genomic TREM2 sequence. (C) Transient splicing reporter assay with genomic TREM2 sequence shows a pattern of isoforms similar to that of the endogenous TREM2 transcript. Endogenous TREM2 transcript from frontal cortex of two subjects, HB1 and HB2, THP1 monocytes and HMC3 microglia was converted to cDNA using poly dT primer. THP1, HMC3 and BV2 cells were nucleofected with a common variant of TREM2 (CV); reporter-specific cDNA was synthesized with a reverse primer positioned on the construct downstream of TREM2 sequence. Gel-resolved RT-PCR products were amplified with primers positioned in exons 1 and 4 (e1–e4, left) and in exons 3 and 5 (e3–e5, right).

Figure 2

c.401A>G (p.D134G) increases exon 3 skipping similar to that of the c.482+2T>C splicing mutant. (A) Left: positions of Nasu-Hakola disease (NHD) variants in TREM2 sequence. Right: Sequences of exon 2/3 splice junction in full-length transcript (FL) and soluble TREM2 (sTREM2) were aligned with sequence of exon 2/4 splice junction in isoform with skipped exon 3 (Δe3). (B) Representative gel and quantification of the reporter assay in two cell lines nucleofected with various constructs. cDNAs were amplified with primers positioned in exons 1 and 4. Identity of Δe3 and Δe2e3 bands was confirmed by Sanger sequencing (Supplementary material, File S2). Level of isoforms with skipped exon 3 (Δe3 + Δe2e3) was calculated as % of sum intensity of all indicated bands. (C) qRT-PCR analysis of the reporter assay shown in B. Isoforms with included/skipped exon 3 were quantified by TaqMan assays to exon 3/4 and exon 1/2 splice junctions. Expression of isoforms was normalized to their level in CV samples. Neo expressed from the reporter plasmid was used for qRT-PCR data normalization. Left: Nearly complete absence of isoforms that include exon 3 due to splicing mutations. Right: Shows that mutant reporter constructs express a reduced level of isoforms with skipped exon 3. Shown are means ± standard deviation (SD) from at least four experiments. **P < 0.01, ***P < 0.001, ****P < 0.0001; two-way ANOVA, Tukey’s test. (D) Representative western blot showing that TREM2 protein is nearly absent in THP1 TREM2 KO cells nucleofected with c.482+2T>C and c.401A>G reporters. Membrane was probed with antibodies (Abs) to the C-terminal part of TREM2. GAPDH was used as loading control. NTC = non-transfected control; CTF = C-terminal fragment. (E) Reduced TREM2 transcript in brains of NHD patients with splice-altering variants. Hippocampal RNA from NHD patients with TREM2 c.401A>G (p.D134G, red) and c.482+2T>C (orange), from two patients with the TYROBP c.141delG (green) and from 17 non-demented controls (CNT, blue) were assayed by NanoString nCounter. *P < 0.05, **P < 0.01, ***P < 0.001, one-way ANOVA, Tukey’s test. ns = not significant.

Figure 3

L133L and R136W increase exon 3 skipping. (A) Positions of variants in exon 3. (B) Representative gels and quantifications of the reporter assay in which cell lines were nucleofected with various constructs. cDNA was amplified with primers positioned in exons 1 and 4. FL (full length) = isoform with conventional exons 1, 2, 3 and 4; Δe2 = isoform with skipped exon 2; Δe3, Δe2e3 = isoforms with skipped exon 3. Isoforms with skipped exon 3 (e3 skipped) are sum of intensities of Δe3 and Δe2e3 bands. Shown are means ± standard deviation (SD) of two experiments per cell line. **P < 0.01; ***P < 0.001; ****P < 0.0001, one-way ANOVA, Tukey’s test. (C) qRT-PCR quantification of Δe3 and Δe2e3 isoforms. Expression of isoforms was normalized to their level in common variant (CV) samples. Neo expressed from the reporter plasmid was used for qRT-PCR data normalization. Shown are means ± SD of two experiments per cell line. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns = not significant; two-way ANOVA, Dunnett’s test.

Figure 4

Mis-splicing of exon 4 caused by K186N decreases functional TREM2 isoform. (A) Representative RT-PCR gels and quantification of the reporter assay in which cell lines were nucleofected with common variant (CV) and K186N constructs. cDNA was amplified with primers positioned in exons 3 and 5. FL (full length) = isoform containing conventional exon 4; Δe4 (ENST00000338469) = isoform with skipped exon 4; Δe3e4 = isoform with skipped exons 3 and 4. Alt e4 (ENST00000373122) = isoform with alternatively started exon 4. Minor = minor bands corresponding by size to splicing products due to usage of predicted cryptic acceptor sites within exon 4. Shown are means ± standard deviation (SD) of two experiments per cell line. *P < 0.05; ***P < 0.001; two-way ANOVA, Holm-Sidak’s test. (B) Positions of cryptic acceptor sites in exon 4 according to Alternative Splice Site Predictor algorithm. Intronic and exonic sequences are shown in small and capital letters, respectively. Conventional and predicted cryptic acceptor sites are in boxes. A nucleotide changed by K186N is highlighted in red. (C) qRT-PCR quantification of TREM2 isoforms. Expression of isoforms was normalized to their level in CV samples. Neo expressed from the reporter plasmid was used for qRT-PCR data normalization. Left: Quantification of FL and Δe4 expression in THP1 cells nucleofected with CV and K186N TREM2 reporter constructs. Isoforms with included/skipped exon 4 were quantified by TaqMan assays to exon 4/5 and exon 3/5 splice junctions. Shown are means ± SD of four experiments. *P < 0.05, Brown-Forsythe and Welch ANOVA tests. Right: SYBR Green qRT-PCR quantification of Alt e4 expression in HMC3 cells transfected with CV and K186N TREM2 reporter constructs. Shown are means ± SD of two experiments. ns = not significant, two-tailed t-test.

Figure 5

Pathogenic missense variants in and near the donor splice site of exon 2 increase its skipping and reduce TREM2 dosage. (A) Positions of variants in TREM2 sequence. (B) Representative RT-PCR gel and quantification of splicing products of the reporter assay in which HMC3 cells were nucleofected with various constructs. cDNA was amplified with primers positioned in exons 1 and 4. FL (full length) = isoform with conventional exons 1, 2, 3 and 4; Δe2, Δe2e3 = isoforms with skipped exon 2. Shown are means ± standard deviation (SD) of two experiments. ***P < 0.001; ****P < 0.0001; one-way ANOVA, Dunnett’s test. (C) qRT-PCR quantification of Δe2 isoform in splicing reporter experiments performed in three cell lines. Δe2 expression was normalized to its level in common variant (CV) samples. Neo expressed from the reporter plasmid was used for qRT-PCR data normalization. Shown are means ± SD of two experiments per cell line. **P < 0.01; ***P < 0.001; ****P < 0.0001; two-way ANOVA, Dunnett’s test. (D) Representative western blot of protein extracts of HMC3 cells nucleofected with various constructs. TREM2 protein was detected with antibodies (Abs) specific to its C-terminal part. GAPDH was used as loading control. NTC = non-transfected control; FL = full length protein; CTF = C-terminal fragment. Bar graph (middle): level of FL and Δe2 protein isoforms was normalized to GAPDH signal. Percentages above bars indicate Δe2 fraction. Bar graph on the right: percentage of FL and CTF bands. Shown are means ± SD of at least two experiments. **P < 0.01; ***P < 0.001; ****P < 0.0001; ns = not significant, one-way ANOVA, Dunnett’s test.

Figure 6

AD− and FTD-associated variants in TREM2 codons R62 and T96 increase exon 2 skipping. (A) Representative gel and quantification of the reporter assay in which HMC3 cells were nucleofected with various constructs. cDNA was amplified with primers positioned in exons 1 and 4. FL (full length) = isoform with conventional exons 1, 2, 3 and 4; Δe2, Δe2e3 = isoforms with skipped exon 2. (B) Representative western blot of protein extracts of HMC3 cells nucleofected with various constructs. TREM2 protein was detected with antibodies (Abs) specific to its C-terminal part. GAPDH was used as loading control. FL = full length protein; CTF = C-terminal fragment. Bar graph: Level of protein isoforms was normalized to GAPDH signal and calculated as % of sum intensity of FL, Δe2 and CTF bands. Shown are means ± standard deviation (SD) of two experiments. **P < 0.01; ***P < 0.001; ****P < 0.0001; ns = not significant; one-way ANOVA, Dunnett’s test. (C) qRT-PCR quantification of Δe2 isoform from experiment in B. Δe2 expression was normalized to its level in common variant (CV) samples. Neo expressed from the reporter plasmid was used for qRT-PCR data normalization. Shown are means ± SD of two experiments. ns = not significant; **P < 0.01; ***P < 0.001; ****P < 0.0001, one-way ANOVA, Dunnett’s test. AD = Alzheimer’s disease; FTD = frontotemporal dementia.

Figure 7

TREM2 exon 2 skipping is increased in the brain of R62H carriers. Expression of Δe2 isoform in brains of R62H carriers and non-carriers (R62) was assessed from RNA-Seq data of three independent cohort studies. Percentage of Δe2 was calculated for each sample in which sum of reads through the e1/2 and e1/3 junctions was >11 reads. Dashed lines depict the median. **P < 0.01, *P < 0.05, Mann-Whitney test. MSBB = Mount Sinai Brain Bank; ROSMAP = Religious Orders Study and Memory and Aging Project.