Introduction

Since Nowell and Hungerford discovered a recurrent abnormal chromosome in patients with chronic myeloid leukaemia (CML) in 1960, haematological malignancies have been at the forefront of cancer genomics and the related quest for personalised medicine [1]. Indeed, by way of introduction, CML (a clonal myeloproliferative disorder characterised by an accumulation of mature granulocytic cells and their precursors) provides a paradigm of how genomic knowledge is used in both the diagnosis and management of haematological malignancies.

Upon advent of G banding, the abnormal chromosome was shown to be a translocation between chromosomes 9 and 22 [t(9;22)]; the Philadelphia (Ph) chromosome. Further analysis showed this resulted in a BCR-ABL1 fusion gene creating a constitutively active tyrosine kinase (TK) [2, 3]. Subsequent in vitro and in vivo studies demonstrated BCR-ABL1 is central to the pathogenesis of CML and activates an array of downstream signalling pathways resulting in uncontrolled cellular proliferation and survival [4, 5].

Detection of BCR-ABL1 forms part of the diagnostic criteria of CML while detection of certain additional cytogenetic abnormalities (major route abnormalities) is associated with a worse prognosis and can influence initial treatment choice [6•, 7]. The discovery of a consistent molecular abnormality driving the pathogenesis of the disease provided a very attractive therapeutic target and the development of tyrosine kinase inhibitors (TKI) with significant activity against the BCR-ABL1 TK followed [8, 9]. The first TKI licensed for CML, imatinib, revolutionised management of the disorder: The International Randomized Study of Interferon and STI571 (IRIS]) demonstrated 73.8% versus 8.5% of patients treated with imatinib or interferon and cytarabine respectively achieved complete cytogenetic responses, i.e. absence of t(9;22) using fluorescence in situ hybridisation (FISH) [10].

The centrality of the BCR-ABL1 fusion protein to the pathogenesis of CML means the presence of mRNA transcripts in peripheral blood leucocytes can be monitored using Quantitative Fluorescence-Polymerase Chain Reaction (QF-PCR) during treatment as a marker of measurable/minimal residual disease (MRD). This is now critical in the management of CML with International Standards allowing inter-lab comparison, and also informs the decision to stop TKI therapy [11•, 12]. In the event of primary or acquired resistance to a TKI, the BCR-ABL1 transcript can be sequenced to detect mutations predicted to reduce the efficacy of specific TKIs and therefore direct treatment changes. The BCR-ABL1 T315I mutation is associated with resistance to all licensed TKIs other than ponatinib, a third generation TKI [13, 14]. Studies are now ongoing to identify very low level BCR-ABL1 mutations prior to clinical evidence of TKI-resistance with the aim of earlier clinical intervention [15].

Therefore, CML in a single disease provides examples of how genomics is clinically used to inform diagnosis, prognosis, therapy response prediction and monitoring of disease burden. Although perhaps the most comprehensive example, CML is not alone within haematological malignancies in having genomics increasingly central to disease management: A combination of the frequent availability of both high-quality genomic material (obtainable from blood and bone marrow aspirate samples) and high throughput next generation sequencing (NGS) technologies has led to an explosion of genomic knowledge starting to filter through to clinical practice. This is illustrated by the prominence of genomics in the 2016 revision of the ‘WHO classification of tumors of hematopoietic and lymphoid tissues’ [1617••].

Although a systematic representation of the genomics of the 198 different disease entities described in the updated WHO classification is beyond the scope of this review, it will seek to discuss the more common practice-changing examples of the application of genomics to haematological malignancies as well as some potential developments on the horizon and remaining unanswered questions. Due to space constraints, discussion will be limited to somatic abnormalities. In order that the non-haematologist can orientate themselves around the haematological malignancies discussed, Fig. 1 provides a broad classification schema.

Fig. 1
figure 1

Broad classification of haematological malignancies discussed in this review

Myeloid Disorders

Acute Myeloid Leukaemia

Acute myeloid leukaemia (AML) is a clonal disorder characterised by expansion of primitive myeloid precursors (blasts) with failure of differentiation and associated bone marrow failure. Unlike CML where a single genetic abnormality defines the disease, in AML the genomic landscape is heterogeneous at both a cytogenetic and molecular level [18••, 19]. Although the diagnosis of AML is usually based on the combination of morphological and immunophenotype findings, recurrent cytogenetic and molecular abnormalities have been described with diagnostic, prognostic, monitoring and more recently predictive clinical implications. Some of these recurrent abnormalities are deemed to be sufficiently associated with a specific prognosis or management strategy that they are classified as distinct disease subtypes within the WHO guidance [17••].

More than 20 years of cytogenetic analysis (metaphase karyotyping) correlated with outcomes within clinical trials have allowed the prognostic significance of recurrent cytogenetic abnormalities to be determined [20]. Cytogenetic changes are separated into favourable, intermediate and poor risk groups [21•, 22]. Approximately 40% of patients have a normal karyotype cytogenetically (deemed intermediate risk). Favourable changes include t(15;17) PML-RARA, t(8;21) RUNX1-RUNX1T1 and inv(16)/t(16;16) CBFB-MYH11 while unfavourable abnormalities include inv(3)/t(3;3) GATA2-MECOM, t(6;9) DEK-NUP214, certain MLL rearrangements, del5, del5q, del7, del7q, del17, 17p abnormalities and complex karyotype (≥ 3 abnormalities). All other patients are considered to be in the intermediate risk group.

The presence of t(15;17) correlates with the distinct morphological entity acute promyelocytic leukaemia (APML), characterised by an excess of promyelocytes and associated with a coagulopathy clinically. Identification of this abnormality warrants treatment using all trans retinoic acid (ATRA) either in combination with chemotherapy, or arsenic trioxide [23]. AML with t(8;21) or inv(16)/t(16;16), collectively known as core binding factor (CBF) leukaemia, appear to derive benefit from gemtuzumab ozogamicin in combination with chemotherapy, particularly in comparison with AML with unfavourable cytogenetics. Studies are now exploring whether intensification of the accompanying chemotherapy backbone improves survival [24, 25]. Assuming a patient is otherwise suitable, in the event of unfavourable cytogenetics, a consolidative allogeneic haematopoietic stem cell transplant (HSCT) would normally be considered in first complete remission [21, 22, 26].

In both APML and CBF-leukaemia, the disease-defining gene fusions are used for QF-PCR-based MRD monitoring. In APML early detection of incipient relapse is of particular importance owing to the frequently associated coagulopathy which can result in life-threatening haemorrhage [27]. In CBF-leukaemia, molecular response after induction is predictive of the risk of relapse; remission monitoring can allow further treatment prior to a full-blown morphological relapse and prolonged neutropenic period with the associated risk of infection [28].

Individual gene mutations, particularly in the context of a normal karyotype, can also affect prognosis and therefore management strategy, with some defining specific (provisional) subtypes of AML within the WHO classification [17]. An insertion within exon 12 of NPM1 destroying the nucleolar localisation signal at the C terminus or biallelic CEBPA mutations are associated with a better prognosis, and in the absence of other poor risk features, these patients would not normally have an allogeneic HSCT in first remission [29,30,31]. NPM1 mutant transcripts can also be used for MRD monitoring with the absence of the transcript in peripheral blood after the second cycle of induction chemotherapy being associated with a lower risk of relapse [32••].

In contrast the presence of a FLT3-ITD (removes autoinhibition of the tyrosine kinase) particularly with a high mutant: wild-type allele ratio is associated with a poor prognosis, as is the presence of mutations in ASXL1, RUNX1 or TP53, all of which are recent additions into international guidelines containing screening recommendations [21, 33,34,35,36,37]. FLT3, being a tyrosine kinase, is an attractive target for TKIs, and a recent phase 3 trial of the multi-targeted kinase inhibitor midostaurin has shown improved overall survival (OS) and event free survival when given in combination with chemotherapy. This benefit was also seen in the context of activating mtuations within the tyrosine kinase domain (FLT3-TKD mutants), an abnormality whose effect on prognosis is less clear [38•, 39].

However, even the limited selection of genomic abnormalities discussed above often does not occur in isolation, and it is necessary to try and group combinations of mutations into prognostic hierarchies. One such example is the work of Papaemmanuil et al. who used data obtained from cytogenetics, and a 111 gene-targeted sequencing panel to define 14 different genomic subgroups of AML based on segregation of genetic abnormalities [18••]. Reassuringly this demonstrated many of the previously reported subtypes (see above) but in addition defined two new poor prognosis (mutations in chromatin and/or RNA-splicing regulators, and TP53 mutations and/or chromosomal aneuploidies) and one new good prognosis (IDH2 R172 mutant) categories. Other work investigating the ontogeny of AML suggests mutations in chromatin and/or RNA-splicing regulators are likely to be associated with a precedent myelodysplastic syndrome (MDS) which is not always documented clinically but is associated with a poorer prognosis [40].

More extensive sequencing approaches have revealed new potential drug targets: early trials of IDH1 and IDH2 inhibitors have returned promising results in the relapsed/refractory population where treatment options are usually limited [41, 42]. They also raise the potential of molecular MRD for all; however, studies have shown that the persistence of some mutations (those most frequently associated with clonal haematopoiesis of indeterminate prognosis [CHIP], i.e. ASXL1, DNMT3A, TET2) does not correlate with an increased risk of relapse, whilst others persist at variant allele frequencies which suggest they are present in morphologically normal cells [43•]. One final caution relating to choice of sequencing technology which has implications beyond AML is how to equate results from a high sensitivity technique such as whole genome sequencing (WGS) to a low sensitivity assay such as conventional metaphase karyotyping: what constitutes a complex karyotype with WGS? Further evaluation is required within clinical trials to generate data of equivalent robustness to that which underpins accepted management-informing genomic knowledge in AML.

Myelodysplastic Syndromes

Myelodysplastic syndromes (MDS) are a clinically heterogenous group of haematological malignancies characterised by ineffective haematopoiesis, peripheral blood cytopenias and a predisposition to AML. Recurrent cytogenetic abnormalities are reported in approximately 50% of patients with MDS and are associated with differential prognoses which can be incorporated into multifactorial prognostic scores [44,45,46].The Revised International Prognostic Scoring System (IPSS-R) includes a refined cytogenetic classification containing additions such as discriminating between a complex karyotype with 3 or > 3 abnormalities (poor and very poor risk respectively) and accommodating dual cytogenetic abnormalities [46].

Addition of FISH and high resolution single nucleotide polymorphism (SNP) array analyses can provide additional structural abnormality data allowing further reclassification. However, it should be noted that the IPSS-R used only metaphase karyotyping, and it is unclear whether the same clinical correlations will be apparent using higher resolution technology [47, 48]. The most frequently occurring cytogenetic abnormality of del(5q) is found in ~ 15% of patients with MDS [49]. This is associated with a macrocytic anaemia often resulting in red blood cell (RBC) transfusion dependence. Treatment with the immunomodulatory agent lenalidomide can restore RBC-transfusion independence in ~ 50% of patients [50]. Approximately 20% of patients with MDS with del(5q) also harbour TP53 mutations either at diagnosis or which emerge during treatment and are associated with an increased risk of progression to AML [51, 52].

With the ready availability of NGS, attention has now turned to understanding molecular abnormalities in MDS. Progressively broader sequencing studies have demonstrated a molecularly heterogeneous genomic landscape with frequent mutations in genes involved in RNA splicing and epigenetic modifications (i.e. DNA methylation and histone modifications) [53, 54•, 55]. Aside from there being a correlation between increasing numbers of driver mutations and worsening leukaemia-free survival, such studies have shown a strong genotype:phenotype correlation between somatically acquired mutations in the spliceosome constituent gene SF3B1 and the presence of ring sideroblasts: Approximately 80% of patients with refractory anaemia with ring sideroblasts (RARS) and refractory cytopenia with multilineage dysplasia (RCMD) have an SF3B1 mutation [56]. SF3B1-mutated MDS typically has a more indolent course and better clinical outcome compared with other MDS subtypes. The presence of an SF3B1 mutation also correlates with a response to the TGFβ superfamily inhibitor luspatercept in early phase clinical trials [57].

Work is ongoing to more systematically determine which gene mutations are associated with a favourable and unfavourable prognosis and how this can be combined with cytogenetics data into a comprehensive genetic risk model: One such model from the Mayo clinic identifies SF3B1 versus TP53, RUNX1, U2AF1, ASXL1, EZH2 and SRSF2 mutations as favourable and unfavourable respectively on univariate analysis [58]. The impact of mutations on commonly employed treatments is also being evaluated. Mutations in TP53, JAK2 and the RAS pathway are associated with inferior survival post allogeneic HSCT (the only curative treatment for MDS) compared with patients without these abnormalities, although in the case of RAS pathway mutations use of an intensive conditioning regiment may ameliorate this risk [59•]. Identification of mutations associated with a poor prognosis when treated with conventional agents including chemotherapy has led to investigations of novel targeted agents. Examples include the use of the BCL2 inhibitor venetoclax in TP53-mutated disease and IDH1/2 inhibitors in IDH-mutated MDS [60].

In addition to providing prognostic information, the detection of a clonal cytogenetic or molecular marker can be useful diagnostically, particularly if there are potential alternative diagnoses or no excess of blasts morphologically. However, care must be taken in the interpretation of any molecular variant as whole exome sequencing (WES) studies of large numbers of individuals unaffected by haematological malignancies has shown age-related acquisition of mutations in genes associated with haematological malignancies: CHIP. The presence of such mutations is associated with an increased risk of subsequent haematological malignancy development and cardiovascular disease [61•, 62]. Furthermore, the presence of such mutations, particularly in PPM1D, pre-autologous HSCT, is associated with an increased risk of secondary MDS/AML post-autograft [63•].

Myeloproliferative Neoplasms

Myeloproliferative neoplasms (MPN) are characterised by the overproduction of mature blood cells and are divided into Ph+ and Ph− diseases. The Ph+ disease entity is CML, the genetics of which is discussed in the introduction. Ph− MPN are a far more heterogeneous group of disorders which can be separated into classic and non-classic groups. Classic Ph− MPN include polycythaemia vera (PV), essential thrombocythaemia (ET) and primary myelofibrosis (PMF) associated with increased erythroid activity, increased platelet numbers and bone marrow fibrosis respectively. The first driver mutation discovered in Ph− MPN was JAK2 V617F with large scale sequencing studies showing its presence in 95% PV and 50–60% ET and PMF patients [64, 65]. In the majority of JAK2 V617F mutation-negative patients with PV an insertion/deletion mutation can be found in JAK2 exon 12 [66].

A minority of patients with ET (3%) and PMF (5%) were found to have mutations of the thrombopoietin receptor MPL [65, 67]. In 2013, WES studies of patients with MPN which hitherto did not have a demonstrable driver mutation demonstrated most had an insertion/deletion mutation in CALR exon 9 resulting in loss of the protein’s endoplasmic reticulum retention motif [68, 69]. The driver mutations of classic Ph− MPN all ultimately activate the cytokine receptor/JAK2/STAT pathway and its downstream effectors a fact exploited by the use of JAK1/2 inhibitors in the treatment of particularly PMF [65, 70]. The essential role of these driver mutations in generating the MPN phenotype has led to their inclusion in disease-specific diagnostic criteria. In their absence, the presence of a variant in a gene recurrently mutated in myeloid disorders can be used as evidence of clonal haematopoiesis [17].

The presence of specific mutations (driver and non-driver) in MPNs is being used to refine prognosis: A high-molecular risk category of PMF has been defined by mutations in ASXL1, EZH2, IDH1/2 and SRSF2 and is associated with reduced survival and increased rate of transformation to AML [71]. Both high risk cytogenetic and molecular abnormalities are being combined with other laboratory and clinical parameters in PMF to produce enhanced, prognostic scoring systems such as MIPSS70+ Version 2.0 [72•]. Similarly, clinical, laboratory, cytogenetic and molecular mutation data has been combined to create a personalised outcome prediction tool for the different MPNs that may help identify those patients at unexpected high risk of an early death or transformation to AML who would potentially benefit from clinical trials of novel therapies [73••].

Among the non-classic Ph− MPNs, the identification of specific mutations can also have both diagnostic and therapeutic implications: Systemic mastocytosis (SM), characterised by organ dysfunction in association with the accumulation of abnormal mast cells, is almost always associated with an activating KIT D816 mutation which can be targeted by kinase inhibitors (e.g. midostaurin) [74, 75]. Chronic neutrophilic leukaemia is often associated with a mutation of CSF3R the nature of which (membrane proximal or cytoplasmic tail-truncating) influence whether the disease responds to TK or JAK inhibitors [76]. Similarly, myeloid neoplasms associated with eosinophilia are often associated with translocations of PDGFRA, PDGFRB, FGFR1 or PCM1-JAK2 with the identity of the translocation partners determining differential response to TK and JAK inhibitors [77].

A further group of myeloid disorders straddle the MDS and MPN diagnostic categories, i.e. MDS/MPN. The most common of these is chronic myelomonocytic leukaemia (CMML), characterised by a persistent blood monocytosis with accompanying morphological dysplasia. Although conventional karyotyping is often normal, targeted sequencing usually reveals mutations in one or more of TET2, SRSF2 or ASXL1 [17]. The clinical phenotype and disease behaviour of CMML is heterogeneous, correlating with the variation in acquired genetic abnormalities: Progression towards AML is seen in association with the acquisition of new RAS mutations whereas a more dysplastic phenotype can occur with co-occurrence of traditional MDS mutations, e.g. of U2AF1 or SF3B1 [78]. In some MDS/MPN category disorders, the combined phenotype appears to reflect co-occurring mutations: In MDS/MPN with ring sideroblasts and thrombocytosis, both SF3B1 and JAK2 mutations are frequently seen [79].

Lymphoid Disorders

Acute Lymphoblastic Leukaemia

Acute lymphoblastic leukaemia (ALL) is a clonal disorder characterised by expansion of primitive lymphoid precursors (blasts) and associated bone marrow failure. The diagnosis of ALL is usually made using morphological and immunophenotype information and further categorisation reflects antigen receptor rearrangement, i.e. B cell receptor (BCR) in B-ALL and T cell receptor (TCR) in T-ALL. The genomics of both subtypes are heterogeneous with more being known about that of B-ALL owing to its higher incidence. Although a broad range of both structural and small variant abnormalities have been described in ALL, currently structural aberrations are more frequently used for diagnostic, prognostic and predictive purposes in both paediatric and adult disease [80, 81].

As seen with AML, decades of cytogenetic and FISH analysis have allowed identification of recurrent abnormalities with strong enough prognostic or management implications for them to be considered distinct disease subtypes within the WHO guidance [17, 82,83,84]. In B-ALL these include the good prognosis hyperdiploidy (> 50 chromosomes) and t(12;21) ETV6-RUNX1, and poorer prognosis hypodiploidy (< 44 chromosomes), t(9;22) BCR-ABL1, MLL-rearrangement and iAMP21 (intrachromosomal amplification of chromosome 21) [80]. Identification of poorer prognosis cytogenetic abnormalities allows upfront treatment intensification or, in the case of detection of the BCR-ABL1 translocation, addition of a TKI which can improve long-term outcomes [85, 86].

A recent refinement to the genomic classification of B-ALL followed gene expression profiling which revealed a group of patients with no risk-defining cytogenetic abnormalities had a similar expression profile to patients with a BCR-ABL1 translocation, i.e. Ph-like/BCR-ABL1-like ALL [87]. Further investigations have shown most patients with this profile have kinase activating changes, e.g. translocations of ABL1, ABL2, CRLF2, CSF1R, EPOR, JAK2, NTRK3, PDGFRB, PTK2B, TSLP or TYK2 or mutations of FLT3, IL7R or SH2B3. These have differential responses to available TKIs which can be used in combination with conventional chemotherapy [88•]. It is of note that the prevalence of the different subtypes of B-ALL varies with age; MLL-rearranged disease is seen < 12 months, hyperdiploidy, and the ETV6-RUNX1 translocation are more frequent in younger children while Ph+ and Ph-like ALL has a higher prevalence in adolescents and young adults [80]. The age-related distribution of genomic abnormalities is reflected in OS rates in different age-groups.

Sequencing studies in T-ALL have shown that although heterogeneous, genomic abnormalities largely correlate with the stage of maturation arrest of the blasts. Activation of the NOTCH1-signalling pathway, including activating mutations of NOTCH1 itself, is common to many cases of T-ALL. Frequent translocations of transcription factors, e.g. TAL1, TXL1, TXL3 or loss of cell cycle control loci such as CDKN2A are also seen. Although some mutations are potentially targetable, practical implementation is problematic due to the need for evaluation within clinical trials despite the paucity of patients with any particular combination of mutations and the risk of secondary resistance; issues not unique to T-ALL [89].

Similar to myeloid malignancies, genomic abnormalities can be used for MRD monitoring. BCR-ABL1 transcripts are monitored in Ph+ ALL (as per CML), and there is the potential to monitor other chromosomal translocations if appropriately sensitive QF-PCR assays exist. However, most MRD performed exploits the clonal nature of the BCR or TCR rearrangement seen in ALL [90, 91]. The unique VDJ and VJ rearrangements of the heavy and light chains respectively from the BCR (i.e. immunoglobulin [IG] genes)/TCR can be measured using QF-PCR post-treatment to detect MRD. The absence of MRD during treatment has been shown to correlate with a better prognosis and increasingly MRD negativity is being used to de-intensify treatment to try to minimise side effects [92••, 93•].

Lymphoma

Lymphomas are clonal disorders arising from lymphocytes and constitute a heterogenous group of malignancies found in the lymph nodes, spleen, bone marrow, blood, and extra-nodal locations. They are categorised into Hodgkin and non-Hodgkin types with the latter further divided by lymphocyte origin (T, B or NK cell). Their clinical course ranges from indolent to aggressive, and presentation varies including lymphadenopathy, splenomegaly and bone marrow failure. Genomic evaluation can be hampered by lack of fresh material as many do not involve blood or marrow. Technological improvements in the analysis of formalin-fixed paraffin-embedded material and more comprehensive collection of fresh tissue have allowed progress untangling the genomics of these tumours. The following are all non-Hodgkin lymphomas.

Diffuse large B cell lymphoma (DLBCL), the commonest adult lymphoma, is an aggressive subtype usually treated with combination chemo-immunotherapy (e.g. R-CHOP) to which the response can vary. FISH evaluation for MYC, BCL2 and/or BCL6 rearrangements highlights a conventionally poor risk group (double/triple hit) who tend to inferior outcomes with this approach [94]. To date, there is no consensus as to appropriate treatment intensification/modification for this group, not least because of inconsistencies in outcomes in small retrospective case series. Emerging data demonstrating the importance of the translocation partner identity may explain some of this variation and also plan for future trials: poor prognosis is mostly limited to patients with double/triple hit with a MYC-IG gene partner translocation [95•].

Gene expression profiling (GEP) demonstrated two distinct subtypes of DLBCL with profiles similar to activated—or germinal centre—B cells (ABC v GCB) [96]. This division was deemed significant as patients with ABC-type often appear to have a worse outcome with standard chemo-immunotherapy treatment [97]. DNA sequencing showed that although DLBCL is genomically heterogeneous, there is some correlation between GEP and mutation type, i.e. ABC and GCB types have activating mutations in signalling pathways (e.g. MYD88, CD79B) and epigenetic modifiers (e.g. EZH2) respectively [98, 99].

Further studies have shown associations between specific mutations and site of disease: primary CNS and testicular DLBCL are associated with MYD88 L265P mutations while primary mediastinal B cell lymphoma has an over-representation of PTPN1 mutations [100, 101]. More recently sequencing studies have attempted to use multiple genomic abnormalities to classify DLBCL: Work by Staudt using an iterative approach generated an algorithm classifying ~ 50% of cases into four distinct genomic categories which had implications for both the predicted origin of the DLBCL and utility of potential targeted therapies, e.g. Bruton’s tyrosine kinase (BTK) inhibitors [102]. A similar approach by Shipp has generated a more comprehensive classifier able to assign over 95% of cases into one of five separate genomic categories. This includes the identification of a lower-risk ABC-type and poorer prognostic GCB-type which may explain some of the inconsistencies seen in different data series in patient outcomes [103••].

Owing to its B cell origin, DLBCLs have a clonal BCR (similar to B-ALL) which has the potential for use in non-invasive MRD detection. Unlike leukaemias, DLBCL cells are seldom found in the blood and marrow, meaning it is necessary to detect circulating tumour DNA shed into the plasma. Studies have shown tumour-associated VDJ/VJ rearrangements can be detected using QF-PCR of plasma DNA with reemergence of MRD being associated with subsequent relapse. Clinically it is relevant that MRD positivity predates radiologically detectable disease by a median 3–5 months [104].

In contrast to DLBCL, follicular lymphoma (FL) is a low-grade B cell lymphoma associated with a t(14;18) IGH-BCL2 translocation in ~ 90% cases [105]. Extensive sequencing reveals it too has a heterogeneous genomic landscape both between separate patients and different locations of disease in an individual patient (evidence of subclones) [106, 107]. A proportion of patients (30–40%) undergo transformation from FL to DLBCL, and sequencing of longitudinal samples demonstrates most transformed-FL arise from divergent evolution from a common precursor rather than direct linear evolution from the FL [108]. Attempts have been made to incorporate genomic information from NGS panels into clinical prognostic scoring systems (e.g. M7FLIPI) although, as yet, these are not widely used in clinical practice [109]. Such panels also have the potential to identify targetable mutations, e.g. activating EZH2 mutations [110].

Although with the exception of FISH testing, most of the data analysis described for DLBCL, and FL remains in the research/clinical trial sphere, genetic testing is increasingly being applied to routine lymphoma diagnostics/management: Mantle cell lymphoma (MCL), an intermediate grade malignancy, is associated with t(11;14) IGH-CCND1 in ~ 95% of cases (with the remainder likely associated with translocations of CCND2 or CCND3) which forms part of the diagnostic algorithm [111, 112]. Demonstration of a TP53 mutation in MCL suggests patient will have a poor response to intensive chemo-immunotherapy followed by consolidative autologous HSCT opening the way for alternative innovative treatment approaches [113].

Waldenstrom’s macroglobulinaemia (WM), a low-grade lymphoma associated with secretion of a monoclonal IgM paraprotein, is associated in > 90% cases with a MYD88 L265P mutation which can be used to differentiate the malignancy from its mimics [114]. Trial data shows that patients with MYD88-mutated WM have a better clinical response to the BTK-inhibitor ibrutinib than MYD88 wild-type WM, while co-occurrence of a truncating mutation in CXCR4 exon 2 attenuates that response [115]. The presence of the BRAF V600E mutation can be used to assist in the diagnosis of hairy cell leukaemia and can also be targeted by specific inhibitors (e.g. vemurafenib) in patients not responding to conventional management [116, 117]. Similarly, although not pathognomonic, the presence of a MAP2K1 mutation can assist in the diagnosis of hairy cell leukaemia variant [118].

Mutational analysis is also starting to be of clinical utility in the rarer T cell lymphomas given these can often be challenging histologically to differentiate from reactive T cells in the context of infection or an autoimmune reaction. Demonstration of mutations such as RHOA, TET2 or STAT3 can support a malignant diagnosis although it should be noted there is some promiscuity of mutations across subtypes [119, 120].

Chronic Lymphocytic Leukaemia

Chronic lymphocytic leukaemia (CLL) is a clonal disorder with a heterogeneous clinical course associated with progressive accumulation of mature B lymphocytes in the lymph nodes, spleen and bone marrow which can result in bone marrow failure. Similar to other haematological malignancies, the earliest prognostic abnormalities detected were cytogenetic: del(17p)/del(11q) and del(13q) were associated with shorter and longer OS rates respectively [121]. More recently, a complex karyotype (defined as ≥ 3 cytogenetic abnormalities in a single clone) has been shown to be associated with a poor prognosis with those patients with unbalanced chromosomal rearrangements having the worst prognosis [122]. However, technical difficulties with the culture of CLL cells means failure rates for conventional karyotyping are high and alternative technologies such as WGS may be required.

Owing to the vast difference in clinical outcomes of apparently similar diseases, great effort was put into defining the disease genetically: It was discovered that CLL can be divided into two similar-sized cohorts on the basis of whether the immunoglobulin heavy chain variable region sequence (IgHV) has undergone somatic hypermutation (physiological part of antibody affinity maturation in the lymph node) [123]. Patients with IgHV hypermutation have a better OS compared to those with unmutated IgHV and long-term follow-up suggests that some of these patients may even be cured with first-line chemo-immunotherapy (FCR) [124, 125]. The consistency of the effect of IgHV mutation status on prognosis has led to its incorporation in the CLL international prognostic index [126].

Disruption of TP53 either via deletion, mutation or both has also been shown to be a poor prognostic factor in CLL associated with primary refractoriness to chemo-immunotherapy, even if the aberration is in a small subclone [127, 128]. In such cases, small molecular inhibitors targeting BTK, PI3K or BCL2 are preferred [129]. Deep sequencing using NGS panels can detect subclonal mutations, which may ultimately confer resistance to these inhibitors, many months before relapse, e.g. mutations in BTK and PLCG2 can result in resistance to BTK inhibitors [130]. Ongoing sequencing studies are continuing to identify driver mutations affecting a range of cellular processes which have the potential to direct therapy [131].

Plasma Cell Myeloma

Plasma cell myeloma is defined by the accumulation of clonal plasma cells in the bone marrow resulting in end organ damage such as bone marrow failure, destructive bone lesions, hypercalcaemia or renal impairment. It too is associated with both a variable clinical outcome and genomic heterogeneity. Metaphase FISH performed in the context of clinical trials has revealed a number of poor prognosis aberrations defining a cytogenetically high-risk group of patients. Although most trials agree del(17p), t(4;14) and t(14;16) are associated with a poor outcome, different trials have reached varying conclusions about the significance of t(14;20), non-hyperdiploidy, del(1p) and gain(1q), and it is likely that some trial therapies may mitigate some of the poor prognosis otherwise seen [132, 133]. Identification of high-risk patients allows for upfront intensification of treatment and enrolment in clinical trials evaluating novel therapies [134]. It should be noted that the critical gene target of many of these structural abnormalities remains unknown [135].

Although some recurrently mutated genes have been described in myeloma (e.g. BRAF, NRAS, KRAS), unlike in other malignancies, these are not typically targeted therapeutically. Reasons for this include mutations being outside typical hotspots (i.e. ~ 50% BRAF mutations are not at codon 600), being sub-clonal or the presence of two or more clonal mutations in the same pathway (i.e. MAPK pathway) meaning responses to a single inhibitor are likely to be limited [135]. The subclonality of myeloma is notable for both small variants and structural abnormalities and shows evidence of both divergent and linear evolution. Its presence raises questions about the utility of targeted approaches in this disease [136].

Similar to other B cell–derived malignancies the presence of unique VDJ/VJ rearrangements in the BCR provides a target for MRD. Studies have shown that MRD negativity determined using NGS with a sensitivity of 1-in-106 is associated with better progression free survival raising the possibility that this can be used to adapt treatment intensity as seen in ALL trials [137].

Changing Technologies

Tables 1 and 2 (myeloid and lymphoid disorders respectively) summarise the testing described above together with its clinical utility and the common technologies used in its delivery. These tables demonstrate that many disease subtypes, e.g. acute myeloid leukaemia undergo multi-modality testing with each assay looking for a different genomic abnormality. With increasing breadth of the genome being clinically actionable, there is a move towards more comprehensive testing such as that delivered by WGS. Within England, building on the 100,000 Genomes Project, WGS will be made available through the National Health Service for all acute leukaemias (as specified in the National Genomic Test Directory for Cancer) [138, 139].

Table 1 Clinical utility of genomic abnormalities in different myeloid conditions
Table 2 Clinical utility of genomic abnormalities in different lymphoid conditions

Upon robust demonstration of equivalence to existing techniques, and with availability of clinically relevant turn-around times, it is envisaged WGS will be able to replace a number of individual assays with a single test; particularly advantageous if there is limited sample or no viable cells (necessary for karyotyping). It should be noted, however, that the present turn-around times and sensitivity of WGS mean it is unsuitable for replacing very rapid (e.g. FISH for t(15;17) in suspected APML) or high sensitivity testing (e.g. MRD). Nevertheless the technique has the advantage of covering the full breadth of the genome including germline data and subject to appropriate patient consent, WGS coupled with high quality clinical data will provide a research resource that can be interrogated for novel genomic information (including the non-coding regions) to inform prognosis and suggest possible new targets for therapeutic approaches.

Conclusion

Building on early cytogenetic and FISH data and now utilising NGS technologies, genomic information remains central to the diagnosis, risk stratification and treatment choice in haematological malignancies. Increasingly multiple aberrations are being combined to produce new classifications of disease and more accurate prognostic information. The challenge now is how this information can be derived in a timely fashion after diagnosis and used to deploy novel targeted therapies in development, including in clinical trials, to improve outcomes. Equally the relevance of the persistence of any particular abnormality post-treatment and as to whether this represents true MRD remains to be fully determined.