High incidence of MYD88 and KMT2D mutations in Chinese with chronic lymphocytic leukemia

To the Editor

Chronic lymphocytic leukemia (CLL), the most common leukemia in persons of European descent, has a marked lower age-adjusted incidence rate in East Asians¹. Asians with CLL are younger and have some distinct phenotype and genotype features. Chinese with CLL have a > 60–70 percent frequency of mutated IGHV, involvement of IGHV3–7, IGHV3–74, IGHV4–39 and stereotype subset #8 of the specific complementary-determining region stereotype compared with Europeans^{2, 3}. Several large-scale sequencing studies of persons with CLL of predominately Europeans descent (Western CLL) identified recurrent somatic driver mutations^4–6 but there are few such studies in Asians^4–8. A recent study of mutations in TP53, SF3B1, NOTCH1, MYD88 and BIRC3 in Chinese with CLL using Sanger sequencing reported a 10% frequency of MYD88 mutations compared with 2–4% in Europeans^{4, 5, 7}. This was validated in other cohorts from Hong Kong and Singapore⁸. Both studies failed to determine statistically significant genetic differences between Western and Asian CLL due to the low sensitivity of Sanger sequencing and limited numbers of genes examined ^{7, 8}.

To gain a better understanding of the mutation topography of Asians with CLL, we studied mutations in 91 genes in 126 Chinese patients with CLL using next generation sequencing (NGS) and validated the top 20 recurrent mutated genes in another 212 patients. The baseline characteristics of 303 untreated and 35 relapsed patients studied were summarized in Table S1.

The most frequently mutated genes in untreated patients were MYD88 (13%), ATM (11%), TP53 (11%), SF3B1 (9%), NOTCH1 (9%), KMT2D (8%), BIRC3 (6%), FBXW7 (5%), BRAF (3%) and POT1 (2%) (Figure 1A).

Figure 1.

Landscape of recurrent mutated genes and clinical associations. (A) Mutations in 20 recurrent mutated genes in 338 patients. Each row and column represent a mutation and a sample, respectively. Left panel indicates affected pathway(s). The bar graph indicates the frequency each pathway is involved in different cohorts. The right panel represents number of times each genetic alteration was observed. White means missing data. Red asterisks: alterations more frequently observed in samples from relapse subjects compared with newly-diagnose/untreated subjects (q < 0.01). (B) Cox multi-variable regression model of time-to-therapy (TTT) in newly-diagnosed/untreated subjects. (C) Mutation topography of Chinese cohort (n=303), Dana-Farber³ (n=538) and ICGC¹⁸ (n=452) cohorts. Red asterisks indicate p < 0.01.

We next examined whether any of the top 10 frequently mutated genes in 303 untreated patients with CLL correlated with clinical outcome. Consistent with Western CLL, we found TP53 and ATM mutations had a briefer time to first treatment (TTT; p < 0.001) and survival (p < 0.001) (Figure S1) ^{4, 6}. We found no prognostic significance in SF3B1 or FBXW7 mutations different from previous western studies (data not shown)^{4, 6}. Interestingly, KMT2D mutations tend to expect an adverse impact on TTT with a border significance (p=0.079, Figure S1A). In order to identify the independent risk factor of Chinese CLL, we grouped TP53, ATM, NOTCH1, MYD88 and KMT2D mutations together with other established predictive markers (IGHV status, age>65 years, Rai/Binet stage, β2-microglobulin) into multivariable Cox regression models. TP53 status, ATM and KMT2D mutations, Rai/Binet stage and IGHV mutational status retained prognostic significance after adjusting for other factors (Figure 1B). This supported the presence of KMT2D mutation as an independent risk factor for earlier treatment (HR=1.79 [1.1, 3.0], p = 0.025).

To compare mutation frequencies between Chinese and western CLL, we focused on International Cancer Genome Consortium (ICGC) and Dana-Farber datasets. The most striking difference was the high frequencies of MYD88 and KMT2D mutations in Chinese CLL (Figure 1C). MYD88 mutations were detected in 38 of 303 newly-diagnosed patients (13%) but were not found in 35 patients with relapsed CLL (Figure 1A, Figure S1B). 33 of 38 (87%) of MYD88 mutations were clonal (variable allele frequency [VAF] > 80%) and 45% were localized to L265P (Figure S2A). Additional hotspot mutations were also detected (MYD88^V217F; MYD88^S219C; Figure S2A). MYD88 mutations are common in Waldenström macroglobulinemia (WM) and other B-cell cancers. To compare the mutation spectrum of different B-cell cancers, we summarized mutational sites of 46 MYD88 mutant out of 1345 CLL patients reported in 4 western CLL cohorts as well as 38 CLL in our cohort, and compared with 94 DLBCL, and 80 WM cases with mutated MYD88 (Figure S2B). A L265P hotspot mutation was more common in patients with WM (95%) compared with CLL (56%) and DLBCL (63%). MYD88^V217F mutations were more common in CLL compared with DLBCL and WM (20% vs. 3% vs. 3%; p < 0.001). S243N mutations were more frequent in DLBCL compared with CLL and WM (14% vs. 1% vs. none; p< 0.001; Figure S2B).

MYD88 mutant CLL had a significant male predominance (p = 0.002) with IGHV mutation (p < 0.001; Table S2), a great extent of monoclonal gammopathy (3 IgG and 2 IgM, p = 0.008) and a higher frequency of hepatitis B-virus (HBV) infection (HBV surface antigen positive, 15% vs. 4%, p = 0.029; anti-HBV core antibody positive, 56% vs. 34%, p= 0.015). Greater exposure to HBV and other environmental factors might explain the increased frequency of MYD88 mutations in Chinese versus Western CLL (Table S2). MYD88 mutations were predominantly in patients with IGHV VH3–7 and VH3–23 use (Figure S2C). VH3–7 use was common in V217F and S219C mutation (6/16), whereas VH3–23 use was common in the MYD88 L265P mutant (6/17). Patients with MYD88 mutation had a longer TTT compared with those without mutation (m-TTT 5.5 vs. 2.9 years; p = 0.027; Figure S2D). However, MYD88 mutation lost its prognostic significance of TTT if only IGHV-mutated patients are considered (m-TTT 5.5 vs. 4.1 years, p=0.433) (Figure S2D).

KMT2D (also known as MLL2) encodes a highly conserved histone lysine methyltransferase of the SET1 family and is responsible for histone H3-lysine 4 (H3K4) trimethylation⁹. KMT2D is recurrently mutated and functions as an early abnormality in about 30% of persons with DLBCL and about 80% of those with follicular lymphoma (FL). KMT2D mutations in these cancers are predicted to generate truncated proteins lacking part or all of the C-terminal domains^10–12. KMT2D mutations are rarely reported in Western CLL with a frequency < 1%^{4, 5}. We detected 30 KMT2D mutations in 24 untreated and 3 relapse patients with CLL (Figure 1A). We validated 10 mutations in KMT2D with Sanger sequencing and confirmed KMT2D mutations had corresponding mRNA transcripts by polymerase chain reaction (PCR; Figure S3). 21/30 (70%) were loss-of-function mutations including nonsense (N = 11) and splice site (N = 4) mutations or frameshift indels (N = 6) (Figure S4A). KMT2D mutations significantly co-occur with trisomy 12 and mutations in FBXW7 (p<0.001; Figure 2A). 19 of 27 KMT2D mutations had IGHV mutation with over-representation of variable regions VH4–34 and VH4–39 (p = 0.001; p = 0.034) and enrichment in stereotyped subset #8 (p = 0.012; Figure 2B). Subset #8 is a subset of IgG-switched cases expressing restricted IGHV4–34/IGKV2–30 BCRs accounting for < 1% of Western CLL and about 5% of Chinese CLL, a > 5-fold difference^{2, 13}. In Western CLL, subset #8 was significantly enriched in patients with trisomy 12 and NOTCH1 mutations with a higher risk for Richter transformation^{14, 15}. It remains elusive the potential impact of this stereotype with BCR signaling and their clinical association in Chinese CLL. Nonetheless, the biases of IGHV usage in the KMT2D mutant samples provide strong rationale to identify unique molecular characteristics in the subset of Chinese CLL patients with these mutations.

Figure 2.

KMT2D mutations and their functional impacts on CLL cells. (A) Pairwise association of recurrent mutations, cytogenetic abnormalities and IGHV mutation state (n = 292). (B) Comparison of IGHV mutation state, IGVH repertoire and stereotyped subsets between KMT2D-mutated (n = 27) and wild-type (n = 311) subjects. * p < 0.05; ** p < 0.01. (C) Cells from KMT2D-mutated (n = 9) and wild-type (n = 6) samples were treated with the indicated doses of decitabine, chidamide with and without ibrutinib for 72 h and percentage cell viability was assessed compared with a dimethyl sulfoxide (DMSO) control using the Cell-Titer Glo assay..

Because KMT2D is a histone lysine-specific methyltransferase⁹, we studied whether loss of KMT2D function in CLL alters H3K4 methylation. We used an immune fluorescence assay to measure H3K4me3 in 3 KMT2D-mutated and 3 KMT2D wild-type CLL cells. KMT2D mutant cells had significantly less fluorescence intensity of H3K4me3 compared with wild-type CLL cells (Figure S4B). These data confirmed these mutations confer loss-of-function. Next, we explored whether there was a relationship between KMT2D mutation and sensitivity to chromatin modifying drugs. Cells from patients with mutated (N = 9) or wild-type KMT2D (N = 6) were treated with decitabine or chidamide alone or combined with ibrutinib in the CellTiter-Glo assay. KMT2D-mutated cells were more sensitive to decitabine (IC₅₀, Mean±SEM, 22.2 ± 3.9 vs. 73.6 ± 25.4 μM; p = 0.029; Figure S4C), reduced sensitivity to ibrutinib (IC₅₀ 5.3 ± 0.9 vs. 2.9 ± 0.9 μM; p = 0.095) and similar sensitivity to chidamide (IC₅₀ 0.6 ± 0.11 vs. 6.9 ± 6.1 μM; p = 0.218) compared with KMT2D wild-type cells. Ibrutinib and decitabine were synergistic (Figure 2C; p < 0.05) suggesting decitabine increases ibrutinib sensitivity in KMT2D-mutated CLL cells. We calculated a combination index (CI) and confirmed ibrutinib and decitabine have strong synergistic cytotoxicity (Figure S4D). Ibrutinib and chidamide also demonstrated synergy in KMT2D-mutated cells at a lower dosage (Figure S4D). Patients with KMT2D mutations or wild-type were similar in CLL-IPI stage, IGHV mutation state, frequency of del(17p)/TP53 mutation and complex karyotype (Table S3). Thus, the poor prognostic significance of KMT2D mutation is independent of known prognostic factors (Figure 1B).

In conclusion, we observed high frequency mutations of MYD88 and KMT2D in Chinese with CLL compared with persons of predominately European descent. MYD88-mutated CLL had a favorable outcome with male predominance, mutated IGHV status, and higher HBV exposure. KMT2D-mutated CLL showed impaired H3K4 methylation activity and increased sensitivity to decitabine in vitro. The addition of decitabine to ibrutinib may yield synergistic treatment for KMT2D-mutated CLL patients. KMT2D-mutated CLL patients were independently associated with progression adjusting for IGHV mutational status. The data help to explain some of the phenotype and genotype differences between Chinese and Western CLL.