PHF6 mutations in T-cell acute lymphoblastic leukemia.

Tumor suppressor genes on the X chromosome may skew the gender distribution of specific types of cancer. T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive hematological malignancy with an increased incidence in males. In this study, we report the identification of inactivating mutations and deletions in the X-linked plant homeodomain finger 6 (PHF6) gene in 16% of pediatric and 38% of adult primary T-ALL samples. Notably, PHF6 mutations are almost exclusively found in T-ALL samples from male subjects. Mutational loss of PHF6 is importantly associated with leukemias driven by aberrant expression of the homeobox transcription factor oncogenes TLX1 and TLX3. Overall, these results identify PHF6 as a new X-linked tumor suppressor in T-ALL and point to a strong genetic interaction between PHF6 loss and aberrant expression of TLX transcription factors in the pathogenesis of this disease.

T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy in which multiple genetic defects collaborate in the transformation of T-cell progenitors4,5. Notably, T-ALL has a 3-fold higher incidence in males3, while other immature hematologic tumors such as precursor B-lineage ALL (BCP-B-ALL) are equally frequent in males and females3.

To identify a possible X-linked tumor suppressor in T-ALL, we performed an X chromosome targeted mutation analysis in tumor DNA samples from 12 male T-ALL cases. For each sample, we performed in-solution DNA capture of 7,674 regions encompassing 3,045,708 nucleotides corresponding to 5,215 X chromosome exons using the Agilent Sure Select oligonucleotide capture system6. DNA samples enriched for X chromosome exons were then analyzed by next generation sequencing using the SOLiD 3 platform. This analysis identified 66 candidate novel non-synonymous single nucleotide variants and 7 positions with high confidence calls for containing complex variants such as insertions or deletions (Fig. 1a). Dideoxynucleotide DNA sequencing of PCR products encompassing affected exons confirmed the presence of 61/66 (92%) of these single nucleotide variants and 4/7 (57%) of the more complex variants, including 2 insertions and 2 deletions (Supplementary Tables 1 and 2). Sequence analysis of paired DNA samples obtained at the time of clinical remission showed that most of these variants corresponded to previously unreported germline polymorphisms. However, and most notably, we also identified three somatically acquired changes corresponding to two non-synonymous single nucleotide substitutions (c.902A>G, p.T300A and c.990A>G, p.H330R) and a frameshift-creating insertion of 5 nucleotides (c.124_125insAGGCA, p.H43fs) in the PHF6 (plant homeodomain finger 6) gene (Fig. 1a).

Figure 1

Next generation sequencing and array CGH analysis of the X chromosome identifies PHF6 mutations in human T-ALL. (a) Overview of mutation screening approach of the human X chromosome exome in a panel of tumor DNA samples from 12 male T-ALL cases using oligonucleotide sequence capture and next generation sequencing with SOLiD3. After filtering and confirmation of high throughput sequencing data, analysis of corresponding remission DNA samples led to the identification of three somatically acquired changes in the PHF6 gene. (b) Schematic overview of the recurrent genomic deletions involving chromosomal band Xq26.3 in 8 human T-ALL samples. Specific genes located in Xq26.3 are shown. (c) Detailed view of a representative oligo array-CGH plot of leukemia DNA/control DNA ratios (blue tracing) versus the dye-swap experiment (red tracing) in a patient harboring an Xq26.3 deletion. (d) DNA quantitative PCR analysis of PHF6 copy number dose in female and male reference genomic DNAs and 2 primary samples from male T-ALL cases harboring Xq26.3 deletions.

In a complementary approach, we analyzed X chromosome array comparative genome hybridization (array-CGH) data from 246 primary T-ALL samples (179 male and 67 female) in a multi-centre setting. These analyses revealed the presence of recurrent deletions in chromosomal band Xq26 in 8 out of 246 (∼3%) T-ALL samples (Table 1). For 3 del(X)(q26) positive T-ALL cases, we performed array-CGH analysis against the corresponding remission material, which showed that these Xq26 deletions are somatically acquired leukemia-associated genetic events (Table 1). Re-analysis of all 8 del(X)(q26) positive T-ALL cases on a custom high-resolution chromosome X oligonucleotide array (Fig. 1b,c) narrowed down the common minimally deleted region to an area of 80 kb containing the PHF6 gene. Consistently, quantitative PCR analysis confirmed loss of the PHF6 locus in the del(X)(q26) positive cases (Fig. 1d). The convergent findings of our X chromosome exon mutation analysis and analysis of copy number alterations by array-CGH thus identified the PHF6 gene as a new tumor suppressor mutated and deleted in T-ALL.

Table 1

Characteristics of 38 primary T-ALL samples showing PHF6 inactivation

							PHF6 lesion
ID	Sex	Age	WBC (×109/l)	Immuno-phenotype	Genetic subtype	NOTCH1	Type of alteration	Predicted protein	Germline/somatic
	PHF6 deletions
1	M	Ped	77	Cortical	TLX3	mut	Deletion (0.55 Mb)		NA
2	M	Ped	46	Pre-T	TLX3	wt	Deletion (0.23 Mb)		NA
3	M	Ped	31	Pre-T	TLX3	NA	Deletion (1.50 Mb)		NA
4	M	Ped	2	Pre-T	unknown	NA	Deletion (0.27 Mb)		NA
5	M	Ped	NA	Cortical	HOXA	NA	Deletion (1.90 Mb)		Somatic
6	M	Ped	NA	Cortical	unknown	NA	Deletion (0.20 Mb)		Somatic
7	M	Ped	NA	Cortical	TLX1	NA	Deletion (0.08 Mb)		Somatic
8	M	Adult	NA	Pre-T	Unkn	NA	Deletion (0.11 Mb)		NA
	PHF6 mutations
9	M	Ped	185	Cortical	TLX3	wt	Nonsense	p.G122X	NA
10	M	Ped	417	Pre-T	TLX3	mut	Nonsense	p.R116X	NA
11	F	Ped	280	Pre-T	TLX1	mut	Frameshift	p.F172fs	NA
12	M	Ped	405	Pre-T	TLX3	mut	Frameshift	p.Y303fs	NA
13	M	Ped	159	Pre-T	TLX1	mut	Nonsense	p.K235X	NA
14	M	Ped	500	Pre-T	TLX3	mut	Frameshift	p.A41fs	NA
15	M	Ped	347	Cortical	HOXA	mut	Nonsense	p.K274X	NA
16	M	Ped	129	Cortical	unknown	mut	Frameshift	p.D333fs	NA
17	M	Ped	174	Cortical	TLX3	mut	Nonsense	p.R225X	NA
18	M	Ped	27	Cortical	TLX1	wt	Nonsense	p.R116X	NA
19	M	Ped	310	Cortical	TAL1	mut	Missense	p.C283R	NA
20	M	Ped	189	Cortical	TAL1	mut	Frameshift	p.C28fs	NA
21	M	Adult	170	Cortical	TLX3	wt	Frameshift	p.H44fs	NA
22	M	Adult	21	Cortical	TLX1	mut	Frameshift	p.H43fs	Somatic
22	M	Adult	21	Cortical	TLX1	mut	Frameshift	p.H43fs	Somatic
23	M	Adult	NA	Pre-T	TLX3	mut	Frameshift	p.T98fs	Somatic
24	M	Adult	14	Pre-T	TLX3	wt	Frameshift	p.Y105fs	NA
25	M	Adult	28	Mature	TLX3	wt	Frameshift	p.S158fs	NA
26	M	Adult	NA	Cortical	TLX3	mut	Missense	p.C215Y	NA
27	M	Adult	NA	Pre-T	Unkn	mut	Missense	p.C215F	NA
28	M	Adult	31	Cortical	Unkn	mut	Missense	p.C215Y	NA
29	M	Adult	NA	Mature	TLX3	mut	Missense	p.T300A	Somatic
30	M	Adult	21	Cortical	TLX1	wt	Missense	p.A311P	NA
31	M	Adult	30	Mature	Unkn	wt	Missense	p.C280Y	NA
32	M	Adult	23	Cortical	Unkn	wt	Missense	p.H329R	Somatic
33	M	Ped	NA	NA	TAL1	NA	Nonsense	p.R257X	Somatic
34	M	Ped	NA	NA	TLX1	NA	Frameshift	p.S191fs	Somatic
35	M	Adult	NA	NA	TLX1	NA	Missense	p.C215F	Somatic
36	M	Adult	NA	NA	TLX1	NA	Nonsense	p.Y303X	NA
37	M	Adult	NA	NA	Unkn	NA	Nonsense	p.R274X	NA
38	M	Adult	NA	NA	Unkn	NA	Frameshift	p.H135fs	NA

Ped, pediatric; NA, not available; mut, mutated; wt, wild-type.

PHF6 encodes a plant homeodomain (PHD) factor containing four nuclear localization signals and two imperfect PHD zinc finger domains7 with a proposed role in the control of gene expression7. Notably, inactivating mutations in PHF6 cause the Börjeson-Forssman-Lehmann syndrome (BFLS; MIM#301900), a relatively uncommon type of X-linked familial syndromic mental retardation which has not been associated with increased incidence of T-ALL7-9. Quantitative RT-PCR analysis demonstrated ubiquitous expression of PHF6 transcripts in human tissues, with the highest levels of expression in thymus, ovary and thyroid and moderate levels of expression in spleen, testes and adipose tissue (Supplementary Fig. 1). Consistent with these results, PHF6 was readily detected by immunohistochemistry in mouse thymus (Supplementary Fig. 1). Finally, quantitative RT-PCR analysis of human thymocyte populations at different stages of development showed variable levels of PHF6 expression, with marked upregulation of PHF6 transcripts in CD4/CD8 double positive cells (Supplementary Fig. 1).

Mutation analysis of PHF6 in an extended panel of pediatric and adult T-ALL primary samples identified truncating or missense mutations in PHF6 in 38% (16/42) of adult and ∼16% (14/89) of pediatric T-ALL samples (Fig. 2a and Table 1). In all available cases (7/30), analysis of matched buccal and/or bone marrow remission genomic DNA confirmed the somatic origin of PHF6 mutations (4/21 frameshift mutations and 3/9 missense mutations) (Fig. 2b and Table 1). Finally, no mutations in PHF6 were identified in DNA samples from BCP-B-ALLs (n = 62), suggesting that mutational loss of PHF6 in lymphoid tumors could be restricted to T-ALL.

Figure 2

PHF6 mutations and expression in T-ALL lymphoblasts. (a) Structure of the PHF6 protein including four nuclear localization signals and two imperfect PHD zinc finger domains. Overview of all PHF6 mutations identified in primary T-ALL samples and T-ALL cell lines. Filled circles represent nonsense and frameshift mutations, whereas missense mutations are depicted as open circles. Circles filled in gray indicate mutations identified in female T-ALL cases. (b) Representative DNA sequencing chromatograms of paired diagnosis and remission genomic T-ALL DNA samples showing a somatic mutation in exon 7 of PHF6. (c) Western blot analysis of T-ALL cell lines revealed complete loss of PHF6 protein expression in the PHF6 mutated T-ALL cell lines. (d) PHF6 immunostaining in the Jurkat and HPB-ALL, wild-type and mutant T-ALL cell lines, respectively. (e) Western blot analysis of PHF6 and gamma-H2AX expression in HEK293T cells upon PHF6 shRNA knockdown. Actin levels are shown as loading control.

Nonsense and frame-shift mutations accounted for 70% (21/30) of all PHF6 mutations identified in our series and were evenly distributed throughout the gene. Missense mutations accounted for the remaining 30% (9/30) of PHF6 lesions and recurrently involved codon C215 and the second zinc finger domain of the protein (Fig. 2a). DNA sequence analysis of PHF6 in a panel of 15 well characterized T-ALL cell lines (Supplementary Table 3) showed the presence of truncating mutations in the PHF6 gene in the DND41, HPB-ALL and T-ALL1 cell lines. Western blot analysis and immunohistochemical staining of PHF6 demonstrated robust expression and nuclear localization of PHF6 in PHF6-wild-type tumors and complete loss of PHF6 protein in T-ALL cell lines harboring mutations in PHF6 (Fig. 2c,d).

PHD finger-containing proteins have been implicated in numerous cellular functions, including transcriptional regulation and in some instances as specialized reader modules that recognize the methylation status of histone lysine residues10. In addition, PHF6 has been reported to be phosphorylated during mitosis11 and by the ATM and ATR kinases upon DNA damage12, which suggests a dynamic regulation of PHF6 during cell cycle and DNA repair. Consistent with this notion, shRNA knockdown of PHF6 resulted in increased levels of phosphorylated H2AX (gamma-H2AX), a posttranslational modification associated with the presence of DNA double strand breaks13 (Fig. 2e).

Sex determination in humans is controlled by differential representation of the X and Y chromosomes, with presence of an XY pair in the male genome and two copies of chromosome X in females. The presence of numerous genes in the non-autosomal region of the X chromosome could result in a genetic imbalance between male and female cells, which is compensated by random chromosomal inactivation of one copy of chromosome X in female cells14. However, allelic expression analysis has shown that some genes can escape X chromosome inactivation in certain tissues1,2,15. To test the possibility that PHF6 could escape X-inactivation in T-ALL cells, we performed allelic expression analysis of a silent SNP (rs17317724) located in the 3′ untranslated region of PHF6 in lymphoblasts from 3 informative female T-ALL cases. In each of these samples, PHF6 was monoallelilically expressed, suggesting that biallelic expression of PHF6 is not commonly found in T-ALL (Fig. 3a). Most notably, we found that PHF6 mutations are almost exclusively found in male T-ALL cases. PHF6 mutations were present in 29/92 (32%) males and only in 1/39 (∼2.5%) females (P < 0.001; Fig. 3b and Supplementary Table 4). Moreover, all 8 PHF6 deletions identified by array-CGH analysis were found in male T-ALL cases, and each of the three cell lines with mutations in PHF6 were derived from male T-ALL cases.

Figure 3

PHF6 expression in T-ALL lymphoblasts. (a) Sequence analysis of paired genomic DNA and cDNA samples shows monoallelic expression of PHF6 SNP rs17317724 in lymphoblasts from a wild-type PHF6 female T-ALL case. (b) Differential distribution of PHF6 mutations in T-ALL samples from male and female cases. (c) Immunohistochemical analysis of PHF6 expression in wild type and mutant T-ALL lymphoblasts.

Immunohistochemical analysis of PHF6 expression in wild-type primary T-ALL samples showed positive PHF6 immunostaining (n = 5; 3 males and 2 females), while cases with PHF6 truncating mutations (n = 4) (Fig. 3c) or a point mutation in C215 (p.C215F) were negative for PHF6 protein expression (Fig. 3c). In contrast, primary T-ALL cells harboring a PHF6 point mutation in the PHD2 domain (p.T300A) were positive for PHF6 protein expression (Fig. 3c). Overall, these results suggest that truncating mutations and point mutations in C215 impair PHF6 expression, while amino acid substitutions in the PHD2 domain of PHF6 may selectively impair the tumor suppressor function of this protein.

Leukemic transformation of immature thymocytes is the result of a multistep process involving numerous genetic abnormalities, which can be associated with different clinical features, including age and prognosis. Notably, PHF6 mutations were significantly more prevalent in adult (16/42; 38%) than in pediatric T-ALLs (14/89; 16%) (P = 0.005; Fig. 4a). Detailed genetic information was available for T-ALL cases treated in DCOG clinical trials (n = 65) (Supplementary Table 5). In this cohort, PHF6 mutations were significantly associated with the aberrant expression of TLX1 and TLX3 (P < 0.005; Fig. 4b and Supplementary Table 5), two related oncogenes activated by chromosomal translocations in T-ALL16-18. No significant associations were observed between PHF6 mutations and mutations in NOTCH1, FBXW7 or PTEN in either pediatric (n = 65) or adult (n = 34) T-ALL cohorts (Supplementary Tables 5 and 6). Overall survival in PHF6 wild-type pediatric T-ALL cases treated on DCOG protocols19 was 65% (33/51) vs. 71% (10/14) for PHF6-mutated cases (log-rank P = 0.71) (Fig. 4c). Overall survival in PHF6 wild-type adult T-ALL leukemias treated in the ECOG2993 clinical trial was 36% (7/12) vs. 58% (8/22) for PHF6-mutated samples (log-rank P = 0.24) (Fig. 4d).

Figure 4

Clinical and biological characteristics associated with PHF6 mutations in T-ALL. (a) Frequencies of PHF6 mutations in pediatric and adult T-ALL samples. (b) Differential distribution of PHF6 mutations in TLX1/TLX3 positive and negative T-ALL samples. (c) Kaplan-Meier curve of overall survival in pediatric T-ALL patients from DCOG trials ALL7, ALL8 and ALL9 with and without PHF6 mutations. (d) Kaplan-Meier survival curve in adult T-ALL patients with and without mutations in PHF6 treated in ECOG clinical trial ECOG2993.

Overall, these results identify PHF6 as a new X-linked tumor suppressor gene and strongly suggest a specific interaction between the oncogenic programs activated by aberrant expression of TLX transcription factors and the mutational loss of PHF6 in the pathogenesis of T-ALL.