Recurrent histone mutations in T‐cell acute lymphoblastic leukaemia

Mutations affecting key modifiable histone type 3 (H3; Table SI) residues are frequent oncogenic events in certain solid tumours (Feinberg et al, 2016), and have also recently been implicated in a subset of acute myeloid leukaemia (AML) (Lehnertz et al, 2017). Here, we systematically reviewed the somatic mutations in >20 000 cancer specimens to identify tumours harbouring H3 mutations. In a subset of T-cell acute lymphoblastic leukaemia (T-ALL) we identified non-methionine mutations of the key modifiable H3 residues, lysine (K) 27 and 36. The starting point of our investigation was a search for H3 hotspot mutations in 1020 human cancer cell lines (Table SII). In two cell lines, both derived from T-ALL, we found lysineto-arginine mutations at H3K27 and H3K36 (Table I). One of the cell lines, LOUCY, is derived from a NOTCH1 wild-type adult T-ALL (Ben-Bassat et al, 1990). The second, CML-T1, was derived from the T-lymphoblastic blast crisis of chronic myeloid leukaemia (Kuriyama et al, 1989). Ten further T-ALL cell lines lacked coding H3 mutations (Table SIII). In solid tumours, H3K27 and H3K36 are typically mutated to methionine (Fig 1) (Feinberg et al, 2016). However, recent functional studies of H3 lysine-to-isoleucine mutations in AML demonstrate that the latter also dramatically alter global H3 methylation and acetylation patterns (Lehnertz et al, 2017). Therefore, we speculated that lysine-to-non-methionine mutations may also be drivers of a subset of T-ALL. We next searched for canonical H3 mutations in a published targeted sequencing study of 633 epigenetic regulator genes in >1000 childhood tumours encompassing 21 cancer subtypes (Huether et al, 2014). Amongst 91 T-ALL specimens, there were two cases with canonical H3 mutations: H3F3A p.K27R and H3F3A p.K36R (Table I). Both mutations were clonal, with a variant allele fraction (VAF) of 38% and 55%, respectively. Among the 37 tumours with H3K mutations, lysine-to-arginine mutations were restricted to TALL (P = 0 001502; Fisher’s exact test). We then extended our screen for H3 mutations to 18 704 tumours, encompassing >60 cancer types other than T-ALL (Tables SIV and SV). This dataset comprised 8764 internally sequenced specimens and 9940 TCGA samples re-analysed using an in-house variant calling pipeline as previously described (Martincorena et al, 2017). We identified only one neomorphic H3 mutation in an acute leukaemia specimen: a previously reported HIST1H3D p.K27M mutation in an adult AML case (TCGA-AB2927-03) (Lehnertz et al, 2017). Finally, we examined an additional T-ALL cohort by capillary sequencing of recurrently mutated modifiable residues K27, G34, and K36 across four frequently mutated H3 genes (Tables SVI and SVII). The cohort comprised 38 T-ALL cases described in detail previously (Maser et al, 2007). One specimen from a 30-year-old patient harboured a H3F3A p.K27N mutation (Figure S1). Interestingly, a H3F3A p.K27N mutation and a H3F3A p.K27T variant were previously identified in a T-ALL RNA sequencing study (n = 31) (Atak et al, 2013). Collectively, our findings indicate that H3K27 and H3K36 mutations are recurrent in T-ALL, a result we were able to reproduce across multiple different cohorts encompassing adult and paediatric cases. This finding is congruent with the fact that mutations in SETD2 and EZH2, methyltransferases that catalyse trimethylation (me3) of H3K36 and H3K27, respectively, are frequent T-ALL drivers (Belver & Ferrando, 2016). Disruptive SETD2 alterations occur in 7 8% of early T cell precursor acute lymphoblastic leukaemia (ETP-ALL), an aggressive subtype with stem cell-like features (Belver & Ferrando, 2016). Interestingly, both T-ALL specimens with H3K36R mutations originated from ETP-ALL (Table I). Notably, mutually exclusive SETD2 and H3K36/H3K34 mutations are reported in paediatric high grade glioma, where both result in reduced H3K36me3 mediated by SETD2 (Feinberg et al, 2016). It is unclear whether a similar co-mutation pattern exists in T-ALL, as H3 genes have not been included in targeted sequencing panels used by the largest T-ALL genomic studies (Belver & Ferrando, 2016). The role of H3K27 modifications in T-ALL pathogenesis is complex (Belver & Ferrando, 2016). It is plausible that mutations affecting this residue could impact the activity of several histone modifiers with established roles in T-ALL pathogenesis. Loss-of-function mutations in EZH2 or other core components of Polycomb repressive complex 2 (PRC2) are found in 42% of ETP-ALL and 25% of T-ALL overall (Belver & Ferrando, 2016). Impaired PRC2 catalytic activity in T-ALL is associated with reduced H3K27me3, stemness and poor prognosis (Belver & Ferrando, 2016). H3F3A p.K27M mutations appear to act predominantly by blocking H3K27 diand trimethylation and increasing H3K27 acetylation (Feinberg et al, 2016). Recent work demonstrates that H3K27I mutations in AML are associated with similar changes in H3 modification patterns (Lehnertz et al, 2017), suggesting that other non-methionine mutations at modifiable H3 residues may influence the activity of PRC2 and correspondence


Recurrent histone mutations in T-cell acute lymphoblastic leukaemia
Mutations affecting key modifiable histone type 3 (H3 ;  Table SI) residues are frequent oncogenic events in certain solid tumours (Feinberg et al, 2016), and have also recently been implicated in a subset of acute myeloid leukaemia (AML) (Lehnertz et al, 2017). Here, we systematically reviewed the somatic mutations in >20 000 cancer specimens to identify tumours harbouring H3 mutations. In a subset of T-cell acute lymphoblastic leukaemia (T-ALL) we identified non-methionine mutations of the key modifiable H3 residues, lysine (K) 27 and 36.
The starting point of our investigation was a search for H3 hotspot mutations in 1020 human cancer cell lines (Table SII). In two cell lines, both derived from T-ALL, we found lysineto-arginine mutations at H3K27 and H3K36 (Table I). One of the cell lines, LOUCY, is derived from a NOTCH1 wild-type adult T-ALL (Ben-Bassat et al, 1990). The second, CML-T1, was derived from the T-lymphoblastic blast crisis of chronic myeloid leukaemia (Kuriyama et al, 1989). Ten further T-ALL cell lines lacked coding H3 mutations (Table SIII). In solid tumours, H3K27 and H3K36 are typically mutated to methionine (Fig 1) (Feinberg et al, 2016). However, recent functional studies of H3 lysine-to-isoleucine mutations in AML demonstrate that the latter also dramatically alter global H3 methylation and acetylation patterns (Lehnertz et al, 2017). Therefore, we speculated that lysine-to-non-methionine mutations may also be drivers of a subset of T-ALL.
We next searched for canonical H3 mutations in a published targeted sequencing study of 633 epigenetic regulator genes in >1000 childhood tumours encompassing 21 cancer subtypes (Huether et al, 2014). Amongst 91 T-ALL specimens, there were two cases with canonical H3 mutations: H3F3A p.K27R and H3F3A p.K36R (Table I). Both mutations were clonal, with a variant allele fraction (VAF) of 38% and 55%, respectively. Among the 37 tumours with H3K mutations, lysine-to-arginine mutations were restricted to T-ALL (P = 0Á001502; Fisher's exact test).
We then extended our screen for H3 mutations to 18 704 tumours, encompassing >60 cancer types other than T-ALL (Tables SIV and SV). This dataset comprised 8764 internally sequenced specimens and 9940 TCGA samples re-analysed using an in-house variant calling pipeline as previously described (Martincorena et al, 2017). We identified only one neomorphic H3 mutation in an acute leukaemia specimen: a previously reported HIST1H3D p.K27M mutation in an adult AML case (TCGA-AB2927-03) (Lehnertz et al, 2017).
Finally, we examined an additional T-ALL cohort by capillary sequencing of recurrently mutated modifiable residues K27, G34, and K36 across four frequently mutated H3 genes (Tables SVI and SVII). The cohort comprised 38 T-ALL cases described in detail previously (Maser et al, 2007). One specimen from a 30-year-old patient harboured a H3F3A p.K27N mutation ( Figure S1). Interestingly, a H3F3A p.K27N   mutation and a H3F3A p.K27T variant were previously identified in a T-ALL RNA sequencing study (n = 31) (Atak et al, 2013). Collectively, our findings indicate that H3K27 and H3K36 mutations are recurrent in T-ALL, a result we were able to reproduce across multiple different cohorts encompassing adult and paediatric cases. This finding is congruent with the fact that mutations in SETD2 and EZH2, methyltransferases that catalyse trimethylation (me3) of H3K36 and H3K27, respectively, are frequent T-ALL drivers (Belver & Ferrando, 2016). Disruptive SETD2 alterations occur in 7Á8% of early T cell precursor acute lymphoblastic leukaemia (ETP-ALL), an aggressive subtype with stem cell-like features (Belver & Ferrando, 2016). Interestingly, both T-ALL specimens with H3K36R mutations originated from ETP-ALL (Table I). Notably, mutually exclusive SETD2 and H3K36/H3K34 mutations are reported in paediatric high grade glioma, where both result in reduced H3K36me3 mediated by SETD2 (Feinberg et al, 2016). It is unclear whether a similar co-mutation pattern exists in T-ALL, as H3 genes have not been included in targeted sequencing panels used by the largest T-ALL genomic studies (Belver & Ferrando, 2016).

Lymphoid
The role of H3K27 modifications in T-ALL pathogenesis is complex (Belver & Ferrando, 2016). It is plausible that mutations affecting this residue could impact the activity of several histone modifiers with established roles in T-ALL pathogenesis. Loss-of-function mutations in EZH2 or other core components of Polycomb repressive complex 2 (PRC2) are found in 42% of ETP-ALL and 25% of T-ALL overall (Belver & Ferrando, 2016). Impaired PRC2 catalytic activity in T-ALL is associated with reduced H3K27me3, stemness and poor prognosis (Belver & Ferrando, 2016). H3F3A p.K27M mutations appear to act predominantly by blocking H3K27 di-and trimethylation and increasing H3K27 acetylation (Feinberg et al, 2016). Recent work demonstrates that H3K27I mutations in AML are associated with similar changes in H3 modification patterns (Lehnertz et al, 2017), suggesting that other non-methionine mutations at modifiable H3 residues may influence the activity of PRC2 and other histone modifying enzymes. The lysine-specific demethylases JMJD3 and UTX are further important regulators of H3K27me3 distribution in T-ALL (Belver & Ferrando, 2016), and it is conceivable that these enzymes may also be affected by H3K27 or H3K36 mutations.
A feature of H3 mutations in solid cancers is their exquisite tumour type specificity (Fig 1) (Feinberg et al, 2016). In this context, it is notable that 5/5 H3 mutations in T-ALL identified by this survey are lysine-to-non-methionine mutations, and 4/5 are lysine-to-arginine mutations. Out of the >20 000 tumour specimens screened for H3 variants, only two other samples harboured H3 lysine-to-arginine mutations, both at low VAF and in tumours with relatively high coding mutation burdens (TCGA-BT-A20Q-01 and TCGA-AN-A0FW-01). Hence, it is possible that lysine-to-arginine mutations confer particular selective advantage in the context of T cell leukaemogenesis.
In summary,~3% of T-ALL harbour non-methionine variants in H3 genes at key modifiable lysine residues. Given the role of dysregulated H3K27/H3K36 modification in T-ALL pathogenesis and the established prognostic significance of mutations in lysine-specific histone modifiers (Belver & Ferrando, 2016), this finding warrants further investigation of the prevalence, clinical and functional significance of H3 mutations in T-ALL. In light of the recent discovery of oncogenic H3K37 mutations in AML (Lehnertz et al, 2017), our findings suggest a broader role for histone mutations in acute leukaemias and clearly justify incorporation of H3 genes into haematological cancer sequencing panels.
The copyright line for this article was changed on 9 July 2019 after original online publication.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Histone 3 mutation in T-ALL validation cohort. Bortezomib plus EPOCH is effective as frontline treatment in patients with plasmablastic lymphoma Plasmablastic lymphoma (PBL) is a rare and aggressive CD20-negative lymphoma associated with poor outcomes. Multiple studies have shown median survival times of 12-18 months (Castillo et al, 2012;Schommers et al, 2013;Morscio et al, 2014). Several case reports and small case series have suggested an increased response rate in patients treated with bortezomib alone or in combination, especially the combination of bortezomib and dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide and doxorubicin (V-EPOCH) (Castillo et al, 2015a;Fedele et al, 2016). Due to its rarity, prospective studies exclusively in PBL patients are unlikely to be performed. We evaluated the potential therapeutic value of V-EPOCH in patients with PBL.
We retrospectively reviewed medical records at participating institutions of all patients with a diagnosis of PBL who received frontline V-EPOCH. The lymphomas were required to have plasmablastic morphology, lack CD20 expression and express one plasmacytic marker (CD38, CD138, MUM1/ IRF4). This study was approved by institutional review boards at each of the participating centres. This report