EGR2 mutations define a new clinically aggressive subgroup of chronic lymphocytic leukemia

Abstract

Recurrent mutations within EGR2 were recently reported in advanced-stage chronic lymphocytic leukemia (CLL) patients and associated with a worse outcome. To study their prognostic impact, 2403 CLL patients were examined for mutations in the EGR2 hotspot region including a screening (n = 1283) and two validation cohorts (UK CLL4 trial patients, n = 366; CLL Research Consortium (CRC) patients, n = 490). Targeted deep-sequencing of 27 known/postulated CLL driver genes was also performed in 38 EGR2-mutated patients to assess concurrent mutations. EGR2 mutations were detected in 91/2403 (3.8%) investigated cases, and associated with younger age at diagnosis, advanced clinical stage, high CD38 expression and unmutated IGHV genes. EGR2-mutated patients frequently carried ATM lesions (42%), TP53 aberrations (18%) and NOTCH1/FBXW7 mutations (16%). EGR2 mutations independently predicted shorter time-to-first-treatment (TTFT) and overall survival (OS) in the screening cohort; they were confirmed associated with reduced TTFT and OS in the CRC cohort and independently predicted short OS from randomization in the UK CLL4 cohort. A particularly dismal outcome was observed among EGR2-mutated patients who also carried TP53 aberrations. In summary, EGR2 mutations were independently associated with an unfavorable prognosis, comparable to CLL patients carrying TP53 aberrations, suggesting that EGR2-mutated patients represent a new patient subgroup with very poor outcome.

INTRODUCTION

Chronic lymphocytic leukemia (CLL), the most common adult leukemia in the Western world,¹ is a malignancy of mature B lymphocytes that accumulate in the blood, bone marrow and other lymphoid tissues.^2,3 Although treatment has undergone profound improvements in recent years,^4–6 CLL shows a remarkable clinical variability, which is likely to be reflective of a large biological heterogeneity.^7,8 While numerous prognostic markers have been identified, the mutational status of the immunoglobulin heavy variable (IGHV) genes and certain, high-risk cytogenetic/genetic aberrations (that is, 11q, 17p deletions and TP53 mutations) have remained the strongest markers and are today applied in routine diagnostics.^9–12

More recently, whole-genome sequencing and whole-exome sequencing studies have begun to unravel the molecular landscape of CLL, revealing a limited number of frequently mutated genes (for example, ATM, NOTCH1, SF3B1, TP53)^13,14 with a long tail of genes altered in < 5% of cases (for example, CHD2,¹⁵ MED12,¹⁶ NFKBIE,¹⁷ POT1,¹⁸ RPS15,¹⁹ SETD2,²⁰ XPO1²¹). Though integration of molecular information has been proposed to improve classical risk stratification models,^22–25 a substantial proportion of patients with a dismal clinical course will not be captured by these algorithms, hence indicating a need to identify additional molecular markers of disease aggressiveness.

Recurrent missense mutations within the EGR2 (early growth response 2) gene, a versatile transcription factor involved in differentiation of hematopoietic cells,^26–28 were recently reported in ~8% of advanced-stage CLL patients and appeared to be associated with a worse outcome.²⁹ Notably, EGR2 mutations were predominantly observed within the three zinc-finger domains located in exon 2. EGR2 is activated through ERK phosphorylation upon B-cell receptor (BcR) stimulation,²⁶ and we have previously shown that EGR2-mutated CLL patients display altered expression of EGR2 down-stream target genes compared with patients wild-type for EGR2, thus pointing to a pathogenic role for EGR2 mutations through dysregulated BcR signalling.²⁹ In addition, global DNA methylation investigations linked abnormal EGR2 activity with aberrant hypomethylation of transcription factor binding sites in CLL.³⁰

In this study, we investigated the frequency, clinical and biological associations, and prognostic impact of EGR2 mutations in a large well-characterized screening cohort (n = 1283), two validation cohorts comprising untreated patients from the LRF UK CLL4 trial (n = 366) and patient samples from the CLL Research Consortium (CRC, n = 490), a Chinese CLL cohort (n = 233) and Richter’s syndrome patients (n = 31). EGR2 mutations were associated with younger age, more advanced disease and other molecular high-risk markers, and remained as an independent factor predicting poor outcome both in the screening and validation cohort. These findings suggest that EGR2 mutations define a new, poor-prognostic subgroup of the disease.

METHODS

Patients

Peripheral blood samples from 1283 CLL patients with tumor content ⩾ 60% (median 94%) were collected from collaborating institutions in the Czech Republic, France, Germany, Greece, Italy, the Netherlands, Sweden, and the United States and comprized the screening cohort. All CLL cases were diagnosed according to the iwCLL guidelines and displayed a typical CLL phenotype.³¹ Over 76% of samples were collected before treatment and within a median of 7 months from time of diagnosis (EGR2-mutated cases, median 2 months). Clinical and biological characteristics of the screening cohort are summarized in Table 1; this cohort had a lower median age at diagnosis and a higher proportion of IGHV-unmutated cases compared with ‘general’ CLL cohorts, likely reflecting that several of the participating institutions are referral centers. In addition, 366 CLL patients, entered into the multicenter trial UK LRF CLL4 (a randomized first-line comparison of chlorambucil, fludarabine and fludarabine plus cyclophosphamide) served as the first validation cohort (Supplementary Table S1). Details of the CLL4 treatment protocol have been previously reported.³² Four hundred and ninety patients collected within the CRC served as the second validation cohort (Supplementary Table S2) with 81% samples obtained before treatment start. Finally, 233 cases from a Chinese CLL patient cohort (Supplementary Table S3), as well as 31 patients with Richter’s syndrome were also screened for EGR2 mutations. Written consent was obtained in accordance with the Declaration of Helsinki and with ethical approval obtained from the local ethics committees.

Table 1.

Comparison of clinical and biological characteristics between EGR2 mutated and wild-type CLL patients within the screening cohort (n = 1283)

	EGR2 wild-type n = 1233	EGR2 mutated n=50	P-value
Age
Median (years)	62.1	57.4	0.0042
<55 years—no. (%)	278 (25%)	17 (35%)
>71 years—no. (%)	194 (17%)	5 (10%)
No information	105	1
Sex
Female—no. (%)	393 (34%)	19 (40%)	0.3490
Male—no. (%)	769 (66%)	28 (60%)
No information	71	3
Binet stage
A—no. (%)	674 (70%)	17 (44%)	0.0005
B/C—no. (%)	290 (30%)	22 (56%)
No information	269	11
Need of treatment
Yes—no. (%)	726 (68%)	42 (91%)	0.0010
No—no. (%)	334 (32%)	4 (9%)
No information	173	4
CD38+
High (>30%)—no. (%)	161 (27%)	20 (67%)	< 0.0001
Low (≤30%)—no. (%)	438 (73%)	10 (33%)
No information	634	20
IGHV
Mutated—(<98% identity)—no. (%)	460 (39%)	9 (19%)	0.0041
Unmutated (≥98% identity)—no. (%)	710 (61%)	39 (81%)
No information	63	2
del(13)(q14)
Absent—no. (%)	665 (64%)	35 (81%)	0.0194
Present—no. (%)	374 (36%)	8 (19%)
No information	194	7
del(11)(q22)
Absent—no. (%)	852 (82%)	29 (67%)	0.0161
Present—no. (%)	187 (18%)	14 (33%)
No information	194	7
+ 12
Absent—no. (%)	927 (89%)	36 (84%)	0.2587
Present—no. (%)	112 (11%)	7 (16%)
No information	194	7
TP53abn
Absent—no. (%)	993 (86%)	41 (84%)	0.6394
Present—no. (%)	161 (14%)	8 (16%)
No information	79	1
NOTCH1 mutation
Absent—no. (%)	812 (91%)	42 (91%)	0.8773
Present—no. (%)	84 (9%)	4 (9%)
No information	337	4
SF3B1 mutation
Absent—no. (%)	804 (90%)	38 (84%)	0.2703
Present—no. (%)	93 (10%)	7 (16%)
No information	336	5
Sampled before treatment
Yes—no. (%)	665 (76%)	35 (80%)	0.5480
No—no. (%)	215 (24%)	9 (20%)
No information	353	6

Abbreviation: TP53abn, TP53 aberrations (that is, mutations and/or deletions). Recurrent genomic aberrations were classified according to the Döhner classification.⁹ A two-sided student’s t-test was used to assess differences in age at diagnosis, while a Chi-square test was applied to evaluate all other variables in EGR2 wild-type versus mutated cases.

Analysis of EGR2 mutations

The EGR2 mutational hotspot region, covering the three zinc-finger domains located in exon 2, was screened as follows: Sanger sequencing was performed to analyze 1048 CLL and 31 Richter’s syndrome patients; 622 cases were investigated using targeted next-generation sequencing (NGS), 171 patients were assessed by both techniques, and in 490 patients MassARRAY iPLEX assay was applied. In addition, for 41 patients in the screening cohort EGR2 mutation status was derived from whole-exome sequencing data.¹⁹ Bidirectional Sanger sequencing was performed according to standard protocols (primers available on request).³³ For targeted NGS, a 500 bp amplicon was bead-purified and library preparation was performed using the Nextera XT (Illumina, CA, USA) kit. Libraries were sequenced on the MiSeq instrument using v2 sequencing chemistry (Illumina). Applying standard settings, sequences were mapped using the alignment tool BWA (v.0.7.12).³⁴ Variant calling was carried out using VarScan 2 (v.2.3.7)³⁵ with a minimum variant allele frequency (VAF) of 0.5% and variants were annotated using Annovar.³⁶ Samples from 366 patients enrolled in the UK LRF CLL4 trial were investigated by targeted NGS using a custom design TruSeq gene panel (Illumina) that included the entire coding region of EGR2.²⁰ Approximately 490 CRC samples were analyzed using the MassARRAY iPLEX assay (Agena, CA, USA) for recurrent mutations at EGR2 amino acid positions 356, 384 and 411/12; all mutated samples were validated by targeted NGS of a 210 bp amplicon using MiSeq (Illumina). Samples with >5% VAF were considered mutated.

Analysis of concurrent mutations by targeted deep-sequencing

Thirty-eight EGR2-mutated patients (including 5 patients with VAF <5%) were analyzed using Haloplex technology (Agilent Technologies, CA, USA) according to the manufacturer’s protocol. Probes targeting all coding exons or hotspot regions of 27 known CLL driver genes and/or genes previously reported in EGR2-mutated CLL^{19,29,37–44} were designed using Agilent’s SureDesign service (https://earray.chem.agilent.com/suredesign/home.htm; Supplementary Table S4). Cluster generation and 125 cycle paired-end sequencing of the pooled library over one lane of the HiSeq 2500 instrument using v4 sequencing chemistry was performed (Illumina). Illumina sequencing adapters were removed using TrimGalore (v.0.3.7) and trimmed reads were aligned to the hg19 human reference genome (February 2009 assembly) using BWA (v.0.7.12). Variants were detected using VarScan2 with a VAF cutoff of 5% and a minimum 30 reads covering the variant was required. Non-synonymous single nucleotide variants and insertions/deletions (indels) that were not present in the 1000 genomes database were included for downstream analyses.

Statistical analysis

Mutational frequencies were assessed using two-sided, descriptive statistics. Overall survival (OS) was calculated from date of diagnosis or time from randomization (UK LRF CLL4) until last follow-up or death, while time-to-first-treatment (TTFT) was calculated from date of diagnosis until initial treatment. Kaplan–Meier analysis was performed to construct survival curves and the Cox–Mantel log rank test was used to determine differences between groups. Cox regression analysis was applied to compare the prognostic significance of EGR2 mutations in relation to other prognostic markers. A significance level of p < 0.05 was applied and all statistical analyses were performed using Statistica Software 13.0 (Dell Inc., OK, USA).

RESULTS

EGR2 mutations and their association with patient characteristics The overall prevalence of EGR2 mutations in this study was 3.8% (91/2403 patients). In detail, EGR2 mutations were detected in 50/1283 CLL patients (3.9%) of the screening cohort, in 12/366 patients (3.3%) of the UK CLL4 trial cohort, and in 18/490 of the CRC cohort (3.7%; Figure 1a; Supplementary Table S5). Further screening revealed that 9/233 (3.9%) patients in the Chinese cohort, and 2/31 (6.5%) Richter’s syndrome patients carried an EGR2 mutation (Figure 1a;Supplementary Table S5).

Figure 1.

Frequency, localization and dynamics of EGR2 mutations. (a) Frequency of EGR2 mutations in investigated cohorts. (b) Localization of mutations identified in EGR2. (c) Co-existing mutations in 38 EGR2-mutated CLL patients. (d) Clonal dynamics of six EGR2-mutated CLL patients. The first 5 patients received FCR therapy between the first (blue bars) and second (green bars) time point sample, while the last patient remained untreated at the second time point (7 years between the samples).

All mutations represented heterozygous missense mutations except for a recurrent in-frame 3-bp insertion identified in 3 cases. The majority of EGR2 mutations (85/91, 93%) were localized to the DNA-binding sites of the 3 zinc-finger domains, and predominantly affected codons E356, H384 and D411 (Figure 1b). These domains are highly conserved between orthologues in different species (Supplementary Figure S1). The somatic nature of EGR2 mutations affecting codons E356, H384, D411, and E412 has been previously confirmed.^13,14,19,29 With the exception of R318Q, all identified amino acid substitutions were predicted damaging with a SIFT score <0.05.⁴⁵

For 49 EGR2-mutated samples from the screening cohort and all patients from the two validation cohorts, information on allelic frequency was available from deep-sequencing and the median VAFs were 38.9%, 36.3% (UK CLL4 trial) and 39% (CRC cohort), respectively (range, 5.6–62%). In addition, 33 patients were found to exhibit low-frequency EGR2 variants at the aforementioned hotspot codons with a VAF ranging from 0.5 to 5%. However, only 14/33 (42%) of these variants could be verified in an independent experiment (Supplementary Table S6). These low-frequency EGR2 mutations were excluded from subsequent survival analyses.

We next evaluated the correlation between clinico-biological characteristics and EGR2 mutation status in all 1283 patients included in the screening cohort (Table 1). Compared with wild-type cases, EGR2-mutated patients were significantly younger (57 vs 62 years; P = 0.0042), more often presented with an advanced Binet stage at diagnosis (Binet B/C 56 vs 30%; P = 0.0005), carried unmutated IGHV genes (81 vs 61%; P = 0.0041) and del(11)(q22) (33 vs 18%; P = 0.0161), as well as expressed high levels of CD38 (67 vs 27%; P < 0.0001).

Co-existing mutations and clonal dynamics of EGR2-mutated CLL Through targeted enrichment, we investigated 27 known CLL driver genes and/or genes previously reported in EGR2-mutated patients^{19,29,37–44} in 38 EGR2-mutated CLL patients (including 5 cases with VAF < 5%). Overall, a mean coverage of 4100 reads per targeted region was achieved with at least 500 reads in 95% and 1000 reads in 90% per nucleotide base of the targeted regions (Supplementary Table S7), hence allowing reliable detection of co-existing mutations. A total of 92 nonsynonymous alterations were detected (Supplementary Table S8); 15 cases harbored one, 8 showed two, 5 patients had three and a single patient presented with four additional gene mutations (Figure 1c), while no additional mutations were detected in the remaining 9 EGR2-mutated cases. Mutations occurred most frequently in ATM (12/38, 31.6%), TP53 (7/38, 18.4%) and SF3B1 (4/38, 10.5%). Of note, 4/12 ATM-mutated cases had two ATM mutations and another 5 cases showed del(11q), resulting in a high frequency of multiple ATM aberrations in EGR2-mutated CLL (9/12, 75%). Alterations affecting the NOTCH signalling pathway were found in 7 patients (18.4%; NOTCH1 (n = 3), FBXW7 (n = 3) and SPEN (n = 1); Figure 1c). Similar findings were observed when comparing identified mutation frequencies with published whole-exome sequencing data from 964 EGR2-wildtype patients,^13,14 showing a significant enrichment of ATM (31.6 vs 10.9%, P < 0.001, two-sided Fisher’s exact), TP53 (18.4 vs 5.4%, P = 0.005, two-sided Fisher’s exact) and FBXW7 (7.9 vs 1.8%, P = 0.036, two-sided Fisher’s exact) mutations.

To gain insights into the clonal dynamics of EGR2-mutated CLL patients, we studied the VAF derived from whole-exome sequencing data before fludarabine, cyclophosphamide, and rituximab treatment and at relapse in five patients with available samples from both time points. A sixth patient was sampled one year after diagnosis and again 7 years later; however remaining untreated at both time points. All samples at both times had been negatively selected for CD5⁺/CD19⁺ cells to ensure a high tumor content (>95%). In all six patients, the clone harboring an EGR2 mutation expanded during the clinical course and became the dominant clone at relapse or at follow-up (Figure 1d).

Clinical impact of EGR2 mutations

In the screening cohort (1178 cases with available clinical data), the median follow-up time for patients who remained alive was 87.5 months (interquartile range, 49.1 to 138.3 months). Patients with mutated EGR2 (VAF >5%) had a significantly worse TTFT (median, 7.8 vs 38.5 months; HR 1.86, 95% CI 1.35–2.57, P < 0.001; Figure 2a) and OS as compared with EGR2-wild-type patients (median, 74.7 vs 127.2 months; HR 2.03, 95% Ci 1.41–2.92, P<0.001; Figure 2b). No survival difference was observed neither between patients with TP53abn and EGR2 mutations (P = 0.900; Figures 2c and d), nor between EGR2-mutated patients with or without concomitant ATM lesions (P=0.665; Supplementary Figure S2). Notably, within the subgroup of 691 U-CLL patients, EGR2-mutated cases (n = 39) showed a significantly inferior OS than wild-type cases (P = 0.009;Supplementary Figure S3). Similarly, among the 164 patients with TP53abn, a high-risk group defined by a concomitant EGR2 mutation with a shorter OS was identified (P = 0.023; Supplementary Figure S4).

Figure 2.

Clinical impact of EGR2 mutations in 1178 CLL patients from the screening cohort. (a) Time-to-first-treatment and (b) overall survival in the screening cohort according to EGR2 mutation status. (c) Time-to-first-treatment and (d) overall survival in the screening cohort according to the established hierarchy for genomic aberrations⁹ and EGR2 mutation status. Patients with TP53abn and concomitant EGR2 mutation are grouped into the TP53abn group (TTFT, n = 7; OS, n = 8).

In multivariate analysis, EGR2 mutations remained an independent negative prognostic marker both for TTFT (HR 1.44, 95% CI 1.01–2.06;P = 0.047) and OS (HR 1.72, 95% CI 1.12–2.65, P = 0.014; Table 2), when including EGR2 mutation status, age, gender, Binet stage, IGHV mutational status, del(11q) and TP53abn in the model. This independent effect on shorter TTFT and OS remained significant also when including additional molecular markers to the model such as NOTCH1 and SF3B1 mutations (Supplementary Table S9).

Table 2.

Multivariate Cox proportional hazard analysis of time-to-first-treatment (TTFT, cases, n = 898; events, n = 624) and overall survival (OS, cases, n = 863; events, n = 371) in the screening cohort

Variable	TTFT			OS
	Hazard ratio	95% Confidence interval	P-value	Hazard ratio	95% Confidence interval	P-value
EGR2 mutation status	1.44	1.01–2.06	0.047	1.72	1.12–2.65	0.014
Age	1.04	0.89–1.22	0.611	2.20	1.78–2.72	<0.001
Gender	1.11	0.94–1.31	0.230	1.32	1.05–1.65	0.016
Binet stage	NA	NA	NA	2.26	1.82–2.81	<0.001
IGHV mutation status	4.18	3.41–5.12	<0.001	3.28	2.54–4.25	<0.001
del(11q)(q22)	1.05	0.87–1.27	0.620	0.84	0.64–1.09	0.190
TP53abn	1.31	1.06–1.62	0.013	1.66	1.26–2.18	<0.001

Abbreviations: NA, not applicable; TP53abn, TP53 aberrations (that is, mutations and or deletions).

In the UK CLL4 trial cohort (n = 366), the median follow-up time for patients who remained alive was 145 months (interquartile range, 127.7–157 months). Among the 12 EGR2-mutated cases treated within the CLL4 trial, 7 patients were randomly assigned to the fludarabine plus cyclophosphamide arm, 5 patients received chlorambucil and none fludarabine treatment. In univariate analysis, EGR2 mutations were significantly associated with a reduced median OS from time of randomization of 24.6 versus 75.7 months for mutated vs. wild-type patients, respectively (HR 1.71, 95% CI 1.30–2.25, P = 0.004; Figures 3a and b). Multivariate analysis confirmed EGR2 mutation status as an independent risk factor for OS (HR 1.95, 95% CI 1.02–3.73, P = 0.043;Table 3).

Figure 3.

Clinical impact of EGR2 mutations in 366 patients from the UK LRF CLL4 trial and 486 patients from the CRC cohort. (a) Overall survival in the UK CLL4 validation cohort according to EGR2 mutation status. (b) Overall survival in the UK CLL4 validation cohort according to the established hierarchy for genomic aberrations⁹ and EGR2 mutation status. Patients with TP53abn and concomitant EGR2 mutation are grouped into the TP53abn group (n = 1). (c) Time-to-first-treatment and (d) overall survival in the CRC validation cohort according to EGR2 mutation status.

Table 3.

Multivariate Cox proportional hazard analysis of overall survival in the UK LRF CLL4 patients (cases, n = 297; events, n = 231)

Variable	Hazard ratio	95% Confidence interval	P-value
EGR2 mutation status	1.95	1.02–3.73	0.043
Age	1.72	1.32–2.24	<0.001
Gender	1.44	1.04–1.99	0.029
Binet stage	1.21	0.86–1.71	0.271
IGHV mutation status	2.22	1.64–2.99	<0.001
del(11q)(q22)	1.46	1.07–1.99	0.018
TP53abn	4.85	2.90–8.11	<0.001

Abbreviation: TP53abn, TP53 aberrations (that is, mutations and/or deletions).

In the CRC cohort (486 cases with available clinical data), the median follow-up time for patients who remained alive was 64.9 months (interquartile range, 35.6–103.7 months). In univariate analysis, EGR2 mutations were significantly associated with a shorter TTFT (median, 35.8 vs 55.7 months;HR 2.08, 95% CI 1.21–3.57, P = 0.007;Figure 3c) and reduced median OS of 98.3 versus 152.9 months for mutated vs wild-type patients, respectively (HR: 2.22, 95% CI 1.03–4.77, P = 0.036; Figure 3d). Lack of cytogenetic/molecular data in a substantial proportion of CRC patients precluded testing of the above multivariate model. Nevertheless, multivariate analysis including age, gender, IGHV and EGR2 mutation status, confirmed EGR2 as an independent risk factor for TTFT (HR 1.92, 95% CI 1.11–3.32, P = 0.020), while only a borderline significance was seen for OS (HR 1.90, 95% CI 0.88–4.12, P = 0.10; Supplementary Table S10), probably because of relatively few events and shorter median follow-up time in this cohort.

DISCUSSION

Missense mutations within the EGR2 gene were recently reported in progressive and/or relapsing CLL patients and hence indicated to be associated with a worse clinical outcome.²⁹ Here, by investigating large well-characterized cohorts, we not only confirm and significantly extend this observation, but also reveal that EGR2-mutated CLL patients display distinctive clinicobiological features and a rapidly progressive disease course. Indeed, survival analysis in our screening cohort demonstrated a significant, negative prognostic impact of EGR2 mutations with markedly short TTFT and OS, similar to patients with TP53abn, that remained as an independent negative factor in multivariate analysis. While EGR2 mutations were confirmed as a high-risk marker of short TTFT and OS in the CRC cohort, they were also shown to independently predict short OS from time of randomization in the UK CLL4 trial cohort. Importantly, the negative prognostic impact of EGR2 mutations was also evident in aggressive subgroups, such as U-CLL and patients with TP53 aberrations, which displayed a particularly short OS. Taken together, our data support that EGR2 mutations define a new subgroup of patients with a particularly dismal outcome.

This comprehensive analysis identified EGR2 mutations in 3.8% of 2403 investigated patients with similar mutation frequencies in the different cohorts analyzed (range, 3.3–3.9%), although different techniques for mutation screening were used. This relatively low mutation rate reflects the known genetic heterogeneity in CLL, with only a handful of genes mutated in 10–20% of cases, but with a long list of gene mutations occurring in less than 5% of CLL cases.^13,14 Notably, no difference was observed between European/American and Chinese CLL patients with respect to their EGR2 mutation frequencies as suggested for other known CLL drivers such as SF3B1.⁴⁶ EGR2 mutations were less often found in subsets carrying stereotyped BcR (4/326 = 1.2%, P = 0.027) in contrast to SF3B1, NOTCH1, TP53 and NFKBIE mutations that recently were reported to be enriched in specific stereotyped subsets.^17,47–49

EGR2 mutations were often detected with a high mutant allele burden (median VAF 38.9%; Supplementary Table S11), indicating that these aberrations occur early in CLL development.²⁹ From our targeted NGS panel, we noted that patients with EGR2 mutations frequently displayed concurrent mutations in DNA damage response, including, ATM and TP53 and in NOTCH signalling pathway, that is, NOTCH1 and FBXW7, indicating that aberrations within these pathways are important contributors to the evolution of the aggressive phenotype in EGR2-mutated patients. In fact, the majority of EGR2-mutated patients (75%) showed multiple ATM aberrations, which is considerably higher than reported in EGR2 wild-type, ATM-mutated CLL patients (30–40%).^50–52 Bi-allelic ATM inactivation is known to be associated with a shorter TTFT and OS.⁵⁰ Collectively, our data indicate an accumulation of several distinct poor-prognostic markers in EGR2-mutated CLL.

Although most EGR2 mutations were deemed to be clonal, we observed a minor proportion of patients (n = 14) with a low EGR2 mutation burden (<5% VAF, confirmed by independent experiments). Survival analysis revealed a trend for shorter OS in these low-burden cases compared with wild-type patients (P = 0.11; Supplementary Figure S5). Furthermore, serial sampling in five treated CLL cases revealed an expansion of the EGR2-mutated subclone over the clinical course and at relapse (Figure 1d; Supplementary Figure S6). Interestingly, one additional case that remained untreated also showed an expansion of the EGR2-mutated clone. This preferential selection of molecularly defined subclones under pressure of chemotherapy is similar to other poor-prognostic markers such as TP53 in CLL⁵³ and BCOR in myelodysplastic syndromes.⁵⁴ Larger studies are now warranted to further analyze the potential clinical impact of low-frequency EGR2 mutations as was recently shown for TP53 and NOTCH1.^53,55

Similar to the pivotal studies,^29,30 EGR2 mutations were clustered in DNA binding domains pointing to a pathogenic role for the hotspots in codons E356, H384 and D411. The functional role of mutated EGR2 in leukemogenesis is however still poorly understood. We recently showed that mutations in the EGR2 DNA binding domain affect cell cycle behavior and lead to altered transcriptional activity and dysregulated BcR signalling.²⁹ Oakes et al. also identified altered EGR2 activity as a mediator for hypomethylation of distinct transcription factor binding sites.³⁰ Whether these effects are restricted to mutations localized in the three sites within the EGR2 DNA binding domain remains to be addressed. Nevertheless, considering the potential role of mutated EGR2 in altering BcR signalling, it will be particularly relevant to study the efficacy of BcR inhibitors in this patient subgroup.

In summary, our novel data highlight EGR2 mutations as an adverse prognostic biomarker for CLL. EGR2 appears to identify a subgroup of CLL patients with a particularly dismal outcome similar to patients with TP53abn. On confirmation of our current data in other cohorts, in particular in patients treated with novel agents (for example, ibrutinib and venetoclax), EGR2 mutation analysis should be considered for inclusion in the current work-up of CLL to identify high-risk patients.