Cordeiro et al. summarised reports of miRNAs in HL lymph nodes, cell lines, and microdissected Reed-Sternberg cells, listing many types, although only miR-21 was common to all three analyses [81]. They also described the potential roles for lncRNAs, such as MALAT1, FLJ42351, LINC00116, and LINC00461, and piRNAs, such as piR-651, piR-20365, and piR-20582. Paczkowska and Giefing updated and confirmed many of the miRNA findings, reporting RT-qPCR and small RNA-Seq methods that pointed to upregulation of let-7f, miR-9, miR-21, miR-23a, miR-27a, miR-155, miR-196a, along with downregulation of miR-138 and miR-150, with many other instances of up- and downregulation reported by single studies [82].

The heterologous nature of NHL is reflected by the many and different ncRNA profiles. In DLBCL, the upregulation of miR-155-5p and miR-21-5p has been consistently reported, with miR-155-5p and miR-222-3p linked to a poor prognosis [83–85]. Shi et al. reported altered expression of 21 miRNAs, which also point to links with pathophysiology [86]. For example, the downregulation of miR-26a, which targets p35 (CDK5R1), may interfere with the CDK5/STAT3 pathway, which is linked to cell proliferation and cell cycle progression [86]. Baghdadi and colleagues summarised the roles of various lncRNAs in different NHLs [78] (Table 8). Some of these lncRNAs, such as SNHG121, LINC0085, and SBFA2 in DLBCL, may also be useful in diagnosis and prognosis [84, 87].

As in B-cell lymphomas, there are numerous examples of abnormal miRNA expression in this variant: upregulation of miR-21, miR-155, miR-223 and miR-494, and downregulation of miR-15a, miR-16, miR-30b, miR-142, miR-146a, miR-148a, and miR-150 [88]. Other reports of deregulated lncRNAs include ZFAS1 overexpression (which may deregulate P53-mediated pathways), increased MALAT1 expression (linked to a poor prognosis), MTAAT associated with the progression of aggressive ALK-negative anaplastic large cell lymphoma, and MIR503HG promoting growth via the miR-503/Smuf2/TGFBR pathway [78, 88].

The dissection of the genetic pathogenesis of AML began in the 1970s, with reports of trisomy of chromosome 8 and a translocation between 9q and 22q (later shown to be the Philadelphia chromosome) [93–95]. This was subsequently extended to t(4;11)(q21;q23), t(8;21)(q22;q22), and a report on the activation of the NRAS proto-oncogene at 1p13.2, which encodes a GTPase with signal transduction pathway activity [96–98]. As the 1990s progressed, other reports appeared, such as that of inv(16) (p13;q22), which places CBFB at 16q22 (encoding a subunit of the core-binding factor) alongside MYH11 at 16p13 (encoding myosin heavy chain), a fusion that forms an in-frame mRNA coding for a protein with potential transforming activity [99]. These and other findings led to the realisation that a classification based on genetics has advantages over one based on morphology, blast counts and surface marker expression [100]. This view gained credence in the current millennium as NGS methods expanded the number of identified mutations in AML to ∼400. Additionally, with its superior sensitivity and specificity, molecular genetics identified abnormalities in other genes, including WT1 at 11p13 (encoding the Wilms tumour protein, a transcription factor), FLT3, and NPM1, alongside fusion transcripts such as MLL::MLLT3 and DEK::NUP214 [101–103]. There was also a growing appreciation that the mutational status of certain genes, such as NPM1, FLT3, NRAS, and CEBPA, is linked to prognosis [104–107].

As knowledge increased, estimates of the frequency of mutations in various genes were published, the most common being NPM1 (33%), t(15;17)(q22;q21) PML::RARA (13%) and TP53 mutation or loss (8%) [108, 109] (Table 9). Additional analysis revealed many instances of co-mutation, an example being that of NPM1, where 50% of cases also exhibited a DNMT3A mutation, while 30% displayed an FLT3-ITD mutation [108]. Similarly, the number of translocations and inversions increased, with many involving genes already described above [110–112] (Table 10). Many genes can be grouped according to their function, for example, those involved in DNA methylation (DNMTA3, IDH1/2 and TET2) or those acting as myeloid transcription factors (CEBPA, ETV6, GATA2, and RUNX1) [111].

Genes involved in translocations not described in Table 9 or elsewhere include MYH11 at 16p13.11, encoding a smooth muscle myosin, proto-oncogene DEK at 6p22.3, encoding a DNA-binder, NUP214 at 9q34.13, encoding nucleoporin 214, RBM15 at 1p13.3, coding for an RNA binding protein, MRTFA at 22q13.1–13.2, encoding a transcription factor, MECOM at 3q26.2, encoding a transcription factor, NUP98 at 11p15.4, encoding a nucleoporin, MLF1 at 3q25.32, encoding a protein with multiple roles in haemopoietic lineage commitment, NSD1 at 5q35.3, encoding a histone methyltransferase, MNX1 at 7q36.3, encoding a transcription factor that binds regulatory elements, MLLT3 at 9p21.2, encoding a molecule that increases the rate of RNA polymerase II-critical haemopoietic transcription factors [113]. The importance of this genetic analysis is demonstrated by its links with prognosis [112]:

Favourable: t(8;21)(q22;q22.1)/RUNX1::RUNX1T1, inv(16)(p13;q22), t(16;16)(p13.1;q22)/CBFB::MYH11, mutated NPM1 without FLT3-ITD, bZIP in-frame mutated CEBPA

Furthermore, knowledge of certain mutations can inform treatment decisions. IDH2 can be targeted with enasidenib, and IDH1 with ivosidenib and olutasidenib (all of which inhibit the product of these genes, isocitrate dehydrogenase). FLT3 can be targeted with midostaurin, gilteritinib, or quizartinib (each of which inhibits tyrosine kinase), while the BCL protein can be targeted with venetoclax [109, 110, 114, 115]. The National Genomic Test Directory for England for AML is extensive [52], reflecting the text described above, and many tests are also part of the directories of other UK nations [53–55]. Entries include a multi-target NGS panel for small variants in NPM1, CEBPA, RUNX1, FLT3, IDH1, IDH2, KIT, WT1, ASXL1, SRSF2, STAG2, RAD21, TP53, KRAS, NRAS, KMT2A(MLL)-PTD, PPM1D, DDX41, PHF6, and CUX1, and a multi-target NGS panel for structural variants to include the detection of t(15;17)(q24;q21) PML::RARA, t(8;21)(q22;q22) RUNX1:RUNX1T1, inv(16)(p13.1q22) CBFB::MYH11, t(9;11)(p21;q23) MLLT3::KMT2A and other 11q23 rearrangements, t(6;9)(p22;q34) DEK::NUP214, inv(3)(q21;q26) GATA2::MECOM, t(1;22)(p13;q13) RBM15::MRTFA, t(9;22)(q34;q11) BCR::ABL1, t(3;5)(q25;q34) NPM1::MLF1, t(5;11)(q35;p15.5) NUP98::NSD1, t(7;12)(q36;p13) MNX1::ETV6, inv(16)(p13.3q24.3) CBFA2T3::GLIS2 and other NUP98 rearrangements. Regarding MRD, NPM1, PML::RARA, CBFB::MYH11, RUNX1:RUNX1T1, and BCR::ABL1 are described, with a testing method being RT-qPCR.

APL was described in detail in 1957. A key point in the molecular pathology of this cancer can be traced back to the 1980s with reports of t(15;17)(q22;q21) [116–119], subsequently shown to bring together PML at 15q24.1 and RARA at 17q21.2 to form a PML-RARA fusion protein [120]. The UK’s National Institute for Health and Care Excellence (NICE) suggests that 98% of cases of APL are due to this mutation and that 10% of AML cases are APLs. If so, then approximately 259 cases in England alone would be expected [121]. With only approximately 30 deaths [1], this represents a considerable treatment success, as the use of all-trans-retinoic acid alongside idarubicin (which interferes with topoisomerase II) can induce a complete remission in the majority of patients. Accordingly, APL has been described as “the most curable form of acute leukaemia” [121, 122]. The remaining 2% of APL cases all involve RARA, but translocated with ZBTB16, NPM1, NUMA1, STAT5B, PRKAR1A, FIP1L1, BCOR, NABP1, TBL1XR1, GTF2I, IRF2BP2 and FNDC3B, while almost half of all APL cases have an abnormal karyotype, primarily del(7q) and +8 [123].

The molecular genetics of CML have much in common with APL, in that, unlike AML, they are almost always linked to a single genetic defect. As previously mentioned [8, 9, 14, 18], this is the presence of t(9;22)(q34;q11), which results in the oncogene BCR::ABL1. This codes for a 210 kDa fusion protein (BCR-ABL1) with tyrosine kinase function that activates second messengers such as JAK2 and STAT5, which is the basis of this particular form of carcinogenesis (Figure 1) [124, 125]. This kinase proved to be inhibitable, leading to one of the first targeted drug treatments for cancer: imatinib; however, in some patients (up to 15% in some studies) there was resistance to the drug [126, 127]. Fortunately, other tyrosine kinase inhibitors (nilotinib and dasatinib, both markedly more effective than imatinib) that target other parts of the fusion protein were developed to address this issue, and others have been subsequently added to this list [125, 128–130]. Unsurprisingly, BCR::ABL1 is the lesion of choice for the detection of MRD, with Salmon and colleagues discussing methods and transcript type [131].

Reflecting the small number of genetic lesions in CML, entries for CML in the National Genomic Test Directory for England focus on BCR::ABL1 by multiplex analysis, FISH and RT-qPCR [52]. Karyotyping may be used for the detection of +8, +19, −7/7q, −5/5q, i(17q), 12p copy number, and t(9;22)(q34;q11) BCR::ABL1 (including p190 and p210 variants), with FISH used for MECOM and 11q23 (KMT2A) rearrangement. Several of these may also be part of an NGS panel [52].

This form of leukaemia is inevitably of B-cells, and, reflecting their frequently close aetiopathogenesis, the WHO guidelines describe CLL and small lymphocytic lymphoma together as mature B-cell neoplasms [25]. The familial nature of CLL, implying genetics, can be traced back to the 1960s, with discoveries of abnormal karyotypes, translocations and rearrangements in immunoglobulin genes emerging in the 1980s [132–137]. In the current millennium, reviews pointed to frequencies of del(13q14) at ∼50%, del(11q22–23) at 18%–20%, +12 at 12%–15%, del(6q21) at 10%, and del(17p13) at 5%–10% [138, 139]. These deletions are important because the absence of TP53 (a tumour suppressor located at 17p13.1), DLEU7 (located at 13q14.3, an inhibitor of transcription factors NF-kB and NFAT), and ATM (located at 11q22.3, required for damaged DNA repair) promotes leukaemogenesis [140, 141].

NGS techniques have subsequently shown loss-of-function mutations in NOTCH1 (located at 9q34.3 and coding for a membrane receptor with epidermal growth factor-like repeats) and SF3B1 (located at 2q33.1 and encoding a splicing factor) in CLL [142, 143]. In one study, mutations in NOTCH1 were present in 21.8% of cases, in SF3B1 in 12.6%, in ATM in 11.1%, in TP53 in 10.6%, and in BIRC3 in 4.2% of cases [144]. These, and other genes, such as immunoglobulin genes and BTK genes, are frequently included in lists of diagnostic and prognostic genes. For example, while del(13q14), with a prevalence of 50%–60% at diagnosis, is indicative of a prognostic risk, the presence of del(11q22-23) (present in 5%–20% of cases at diagnosis) and del(17p13) (present in 1%–13% of initial diagnoses) are both associated with a poor prognosis [145, 146]. Other mutated genes, present in 1.8%–6.3% of cases, are shown in Table 11 [143]. Several risk loci have been identified by GWAS, such as 6p21.3, the nearest genes being BAKI 3 and IRF4, both of which are close to 6q24-25, and IRF8, the latter two genes coding for interferon regulatory factors [147]. MRD in CLL can be detected by multi-colour fluorescence flow cytometry to analyse CD markers relevant to the tumour, although the preferred molecular method is to analyse immunoglobulin gene rearrangements using NGS [148, 149].

Addressing the above, the National Genomic Test Directory for England [52] refers to multi-target NGS panels for small variants in TP53, BTK, PLCG2, and BCL2; for copy number variants in TP53, ATM, DLEU2/7, RBI, and +12; and for IGH, IGK, and IGL rearrangement and hypermutation detection. FISH may be the technique of choice for abnormalities in TP53, 11q, 13q, and del(17p). If present, these may guide treatment, such as the use of idelalisib, an inhibitor of phosphoinositide 3-kinase, in patients with del(17p) or a TP53 mutation [150], and ibrutinib, an inhibitor of Bruton’s tyrosine kinase [151].

Early chromosomal studies of this disease reported abnormalities such as t(21;22), t(8;14), t(9;22), t(4;11), t(11;14), t(1;3) and (t(9;22)(q34:q11)/BCR::ABL1), also known as the Philadelphia chromosome [152–156]. Others helped to differentiate the sublineages of the malignancy. In B-cell ALL, these include t(8;14)(q24;q32)/MYC::IGH, t(2;8)(p11;q24)/MYC::IGK, t(8;22)(q24;q11)/MYC::IGL, t(11;14)(q13;q32)/BCL1::IGH, and t(14;18)(q21;q32)/BCL2::IGH. In T-cell diseases these include t(8;14)(q24;q11)/MYC::TCR, t(11;14)(p15;q11)/TTG1::RHOM1, t(7:11)(q35;p13)/TTG2::RHOM2, t(11;14)(p13;q32)/TCL2::TCR, and inv(14)(q11;q32)/TCL1::IGH [157, 158].

More recent reports recognised many of these abnormalities and described additional chromosomal (hypo- and hyperploidy, iAMP21) and genetic lesions, some of which are rearrangements and indels, while others are translocations (Tables 12, 13) [25, 28, 159–165]. There are many examples of co-mutations, some of which are statistically significant and may therefore have pathogenic implications. For example, in a study of 804 ALL patients, an IKZF1 deletion at 7p12.2, which encodes a zinc finger transcription factor, was linked to a BCR::ABL1 fusion, whereas it was inversely associated with an RTV6::RUNX1 fusion, with both showing a strong probability of p < 0.001, implying pathogenic significance [159]. Elsewhere, it was reported that 35% of Ph-like B-ALLs have PAX5 alterations [160]. Notably, KMT2A has at least six translocation partners–with ELL, forming t(11;19)(q23;p13.1), with AFDN, forming t(6;11)(q27;q23), with AFF1, forming t(4;11)(q21;q23), with MLLT1, forming t(11;19)(q23;p13.3), with MLLT3, forming t(9;11)(p21;q23), and with MLLT10, forming t(10;11)(p12;q23).

Genetic lesions linked to a poor outcome include KMT2A fusions, near-haploidy (24–30 chromosomes), low hypodiploidy (31–39 chromosomes), iAMP21, TCF3::HLF, CDKN2A/B deletions, and ABL-class fusions. CNS disease at diagnosis is also associated with a poor prognosis [166, 167]. As with many other malignancies where there is a known genetic lesion, such as BCR::ABL1, this will be used for MRD. An alternative is immunoglobulin and T-cell receptor gene rearrangements, while other candidates include ETV6::RUNX1 and IKZF1 deletion [167, 168].

The National Genomic Test Directory for England regarding ALL is extensive and cannot be fully reproduced here, although many of the genes discussed above are referenced, and there is considerable duplication [52]. However, whole genome sequencing (WGS) for all variant types in the germline and tumours is described, as are global copy number changes by FISH for hyperdiploidy, high hyperdiploidy, near haploidy, and low hypodiploidy. These, and copy number changes of IKZF1, CDKN2A, CDKN2B, BTG1, EBF1, PAX5, RB1, PAR1 region (CRLF2, CSF2RA, and IL3RA), and ETV6, can also be detected by a multi-target NGS panel.

As with many other cancers, patients starting on thiopurine-based chemotherapy will be tested for TPMT and NUDT15 variants using SNP/small variant detection, as the enzyme variants coded for by these genes affect the biological activity of the drugs: “weak” acting enzyme isoforms can lead to prolonged cytotoxic effects, which may have adverse clinical consequences, while “strong” isoforms may render the drugs less effective.

MRD can be detected using QF-PCR for t(9;22)(q34:q11) BCR::ABL1, t(12;21)(p13;q22) ETV6::RUNX1, t(1;19)(q23;p13)/TCF3::PBX1, t(17;19)(q22;p13)/TCF3::HLF, t(4;11)(q21;q23)/KMT2A::AFF1, t(11;19)(q23;p13.3)/KMT2A::MLLT1, t(6;11)(q27;q23)/KMT2A::AFDN, and t(9;11)(p21;q23)/KMT2A::MLLT3, and for complex variation in IGH, IGK, IGL, TRA, TRB, TRG, and TCRD. Copy number variation (CNV) detection at a genome-wide level using FISH can be used to identify del(1p33), the location of TSL1, iAMP21 and the location of RUNX1.

The WHO guidelines place this condition within a small group of splenic B-cell lymphomas and leukaemias [25]. The principal morphological feature of this disease, the “lace-like” cytoplasmic extrusions, was described in 1958 and led to the condition being named “hairy cell leukaemia” in 1966 [169, 170]. Probing of HCL cells from 48 patients revealed that they all had a V600E (valine > glutamic acid) mutation in BRAF at 7q34, encoding B-Raf, a serine/threonine kinase, and was subsequently defined as a proto-oncogene [171]. BRAF mutations have been reported in several other cancers, and, such as the BCR-ABL1 fusion protein, the protein product is a target for inhibitors such as dabrafenib, encorafenib, ibrutinib, and vemurafenib [172, 173]. The National Genomic Test Directory for England [52] describes a multi-target NGS panel for small variants in BRAF, the V600 hotspot, immunoglobulin heavy chain (IgH) rearrangements, and small variants in MAP2K1 at 15q22.31, with mutations present in a third of cases that are unmutated for BRAF ^V600E [52, 173].

This malignancy was linked to 6% of all leukaemia deaths in England and Wales in 2023 [1], placing it third after myeloid and lymphoid leukaemias, reflecting its frequency in normal full blood counts. First described in detail in 1928 and further elaborated upon in 1975, a monoblastic variant was reported in 1980, with definitions based on morphology, cytochemistry and CD markers, leading to the French-American-British (FAB) system, in which monocytic leukaemia is designated as AML type M5 [174–176]. The WHO guidelines place it with variants of AML. Many commentators have pointed to interrelationships between the different forms of leukaemia depending on differentiation [67, 177]. He and colleagues emphasised the value of LILRB4 and LRRC25 protein products as biomarkers in the monocytic form, while others suggested the value of KMT2A (described above) and the internal tandem duplication variant of FLT3 at 13q12.2, coding for the Fms-related receptor tyrosine kinase 3, in paediatric disease. Additionally, they highlighted the importance of the cell adhesion molecule 1 encoded by CADM1 (formerly IGSF4) at 11q23.3 [178–180].

As the name suggests, this variant, first described in detail by Osgood in 1968, occupies the space between myelocytic and monocytic leukaemia, defined by the FAB system as M4 [181, 182]. Bower and colleagues were among the first to report genetic changes in MML (now KMT2A) and t(6;11)(q27;q23), while Levine et al. described the mutant JAK2 ^V617F in 9 out of 116 (7.8%) cases of chronic MML [183, 184]. As array-CGH and NGS have become more widely used, the list of genetic abnormalities has been expanded, with several more cytogenetic abnormalities and mutations reported (Table 14) [185, 186], alongside mutations in numerous genes in the spliceosome component pathway such as SF3B1, SRSF2, U2AF1 (U2AF35), ZRSR2, SF3A1, PRPF40B, U2AF2 (U2AF65) and SF1 [187, 188]. Genes with a frequency of ≤10% include NRAS, DNAH2, NPM1, IDH2, PTPN11, CSMD1, PTCH1, CDH23, and JAK2, while in a study of 69 patients, 15 subjects (21.7%) had 6 mutations, 8 subjects (11.6%) had 8, and three sets of 5 patients had 5, 9 or 10 mutations. Unsurprisingly, the total number of mutations had a strong impact on the outcome. ASXL1, NRAS, TET2, SRSF2, SETBP1, and RUNX1 status can be used to create a risk stratification model to predict outcomes that include transformation to AML, myelodysplasia, and death [188–191]. The National Genomic Test Directory for England does not refer to MML, but it does have a section on juvenile MML, to be discussed in a subsequent section, which cites many of the genes described above.

In the 1980s, it was recognised that a small number of lymphocytes had phenotypes that were larger and more granular than “standard” lymphocytes, with many of these (∼85%) subsequently defined as NK (= natural killer) cells, which express CD56 but not CD3, the T-cell marker [192, 193]. Chromosomal abnormalities reported in the late 1990s included + X, +8, del(6q21-23), del(13q), del(17p) and del(17q), along with rearrangements in 11q23 [194]. Current WHO guidelines have a section on mature T-cell and NK-cell neoplasms [25], and within this list are NK-large granulocytic lymphocytic leukaemia and NK/T-cell lymphoma. The latter is characterised by mutations in TP53, DDX3X, STAT3, JAK3, MGA, BCOR, ECSIT, and MCL1, alongside deletions in chromosomes 6, 8 and 14 [195]. Aggressive NK cell leukaemia may harbour abnormalities in STAT3 or STAT5B (present in approximately half of cases), which code for molecules of the JAK/STAT pathway, while array CGH has reported loss of 7p15.1-q22.3 and 17p13.1, with gains of 1q23.1-q23.2 and 1q31.3-q44 in NK leukaemia compared to extra-nodal NK lymphoma [196]. Others have used WGS to report mutations in TP53 (present in 34% of cases), TET2 (present in 28% of cases), CREBBP, and MLL2 (both present in 21% of cases) in aggressive NK cell leukaemia [197]. The National Genomic Test Directory for England points to multi-target NGS panels for detecting small variants of STAT3 and STAT5B in NK cell/gamma-delta T-cell lymphoma, and for investigating large granular lymphocyte leukaemia [52].

Chronic neutrophilic leukaemia (CNL), in which the blood film is primarily dominated by mature neutrophils and often a small number of metamyelocytes, was described over a hundred years ago as hyperleukocytosis [198]. Diagnosis first relies on the elimination of differential diagnoses such as a leukaemoid reaction from CML or atypical CML, the latter of which is achieved by failure to find BCR::ABL1 and rearrangement of PDGFRA, PDRGRB, or FGFR1 [67, 199–201]. As with all very rare conditions, confidence in the diagnosis is limited by small numbers. A study of 9 patients with CNL found one individual to have JAK ^V617F and eight to have a CSF3R mutation; of these, five also had a SETBP1 mutation [200], while a larger study of 39 patients found mutations in ASXL1 (in 77%), CSF3R (62%), SRSF2 (41%), SETBP1 (40%), TET2 (20%), and U2AF1 (14%), with NRAS, PTPN11, JAK2, CBL, ABL1, GNB1, USAF2, and several others found at a frequency of <10%, with abnormal cytogenetics in over 30% of patients [201].

There are several differential diagnoses for eosinophilic leukaemia, such as variants of CML, ALL, myelodysplasia, and hypereosinophilic syndrome. Although there are no eosinophil-specific genetic lesions, rearrangements involving PDGFRA, PDGFRB, FGFR1, or PCM1::JAK2 may be present [67, 202, 203]. Investigation of chronic basophilic leukaemia focuses on the MYB::GATA1 fusion arising from t(X;6)(p11;q23), although there are reports of the Philadelphia chromosome t(9;22)(q34;q11), t(3;6)(q21;p21), and t(16;21)(p11;q22) [204]. The 2022 WHO guidelines refer to numerous forms of mastocytosis, with KIT variants (such as D816V) possibly linked to a systemic form, but there may also be a role for TET2, SRSF2, ASXL1, RUNX1, and JAK2 [67, 205].

The potential pathological roles of miRNAs were first reported over 20 years ago, with one of the earliest studies showing that del(13q14), which is present in over 50% of CLLs, is the location of miR15 and miR16. This suggested a route to transformation via BCL2-directed apoptosis [206, 207]. Calin and colleagues subsequently reported that a 13-miRNA signature was linked to the time from the initial CLL diagnosis to the start of treatment [208], opening the door to its use in diagnosis and management [209]. More recently, several miRNAs have been linked to prognosis in CLL (Table 15) [210, 211].

There are numerous reports of altered miRNAs in ALL. For example, Alvarez-Zuniga and colleagues reported that plasma miR-511, miR-34a, miR-22, miR-26a, miR-221, and miR-223 all exhibit good sensitivity and specificity for B-cell progenitor ALL, although others found miR-92a and miR-638 to be less discriminatory, as did another research group with respect to miR-21, miR-24, miR-26, miR-133b, and miR-148a in peripheral blood mononuclear cells [212]. Mendiola-Soto et al. summarised miRNAs in ALL (Table 16), some of which, such as upregulated miR-137 and miR-510, and downregulated miR-100 and miR-151, are differentially expressed in T-ALL versus B-ALL [213], while others hypothesised roles for miRNAs in haemopoiesis and leukaemogenesis [213–215]. MiR-22, miR-24, miR-150, miR-148a, miR-155, miR-633, and others may play a role in HCL by potentially activating the MAP-JNK pathway and possibly forming a unique signature [216].

There is also a substantial body of literature on lncRNAs. Baghdadi and colleagues summarised the role of these molecules in lymphopoiesis, with possible roles in lineage differentiation [78]. Altered lncRNAs in B-cell ALL include LINC0098 (whose target is miR-330-5p) and ZEB1-AS1 (targeting the IL1/STAT3 pathway), which act as oncogenes, and LINC00221 (targeting miR-152-3P) and CASC15 (targeting SOX4), which act as tumour suppressors [217].

One of the earliest comprehensive studies on miRNAs in AML described an expression signature (including miR-128a and −128b, let7b, and miR-223) that could be used to discriminate it from ALL, while another studied 26 species, finding miR-126, miR-130a, miR-93, miR-125a, and miR-146 to be downregulated, while high expression of miR-191 and miR-199a were predictors of poor survival [218, 219]. In addition, other studies reported 17 upregulated and 16 downregulated variants, and that cases with t(15;17) had a unique miRNA signature in 14q32 that included miR-127, miR-154, miR-299, miR-323, miR-368, and miR-370 [220]. More recently, Bhattacharya and Gutti summarised the roles of miRNAs such as miR-124, miR-126, miR-223, and miR-193b. Meanwhile, Fletcher and colleagues described 17 miRNAs with the potential to serve as therapeutic targets, such as mimics of miR-29b and miR-181a, which have been shown to be effective in cell biology and in animal models. Finally, Liu et al. reported that high levels of miR-362-5p and low levels of miR-34a are linked to a poor prognosis [221–223].

There is also a large body of literature on lncRNAs in AML, and, as with other ncRNAs, many interact with other ncRNAs [221]. Examples of this regulation include NEAT1, which is under-expressed in AML and targets miR-23a-3p, with possible consequences for increased myeloid cell proliferation and for apoptosis. Other examples include SATB1-AS1, which binds to and so acts as a neutralising sponge for the miR-580, and MALAT1, which sponges the miR-328-3p, itself acting on the cell cycle regulator CCND2 [223, 224]. There are many examples of lncRNAs with a role in drug resistance, such as DANCR, which, by repressing miR-874-3p, raises ATGL16 protein levels (a key component of the autophagy mechanism) and confers resistance to cytarabine, while others, such as KCNQ1OT1, which is upregulated in AML, act on miR-296-5p and may have a role in c-Myc expression and in suppressing apoptosis [224–227]. Once data reached a critical mass, meta-analyses became possible, one of which, pooling four studies, showed that elevated miR-155 was linked to AML, with an odds ratio (95% confidence interval) of 1.68 (1.41–2.00) [228].

One of the first reports of an ncRNA in CML came from Venturi and colleagues, with evidence of the importance of the miR-17–92 polycistron at 13q31-32. This polycistron includes miR-17-5p, miR-17-3p, miR-18a, miR-19a, miR-20a, miR-19b, and miR-92–1 [228]. Subsequent reports have described the downregulation of miR-10a, miR-150 (which normally targets MYB), miR-328, and miR-181A, with the upregulation of miR-130A [229–231]. However, the key link to pathophysiology is that numerous miRNAs that would normally target BCR::ABL1 are downregulated, with examples being miR-29b, miR-30a, miR-23a, and miR-342-5p, while the upregulation of others influences alternative pathways, such as miR-29a-3p, which decreases apoptosis, and miR-126-3p, which increases therapy resistance [232].

As in other blood cancers, lncRNAs may participate in the disease process. Examples include the downregulation of BGL3, normally acting as a tumour suppressor by targeting several miRNAs and thereby altering the function of PTEN, and the upregulation of lncRNA H19, which targets c-MYC when upregulated, resulting in disease progression and a poor prognosis. As with miRNAs, lncRNAs, such as HOTAIR and HULC, also play a role in resistance to therapy with tyrosine kinase inhibitors [232, 233].

A third type of ncRNA is the closed, circular RNA molecule known as circRNA, which is of interest in blood cancers [221, 223]. Many circRNAs may regulate miRNAs and consequently downstream genes, such as the circ-HIPK2-miR-124a-CEBPA axis, circ_0009910, which targets miR-20a-5p and predicts an adverse prognosis, and circBA9.3, a fusion product of BCR::ABL1. When circBA9.3 is upregulated, it increases tyrosine kinase activity and consequently resistance to therapy [221, 223, 234, 235].

A link between WM and IgM-MGUS has been noted in that the L256P mutation of MYD88 (located at 3p22.2 and coding for a transduction adaptor) may be present in approximately 90% of the former, approximately 50% of cases of the latter, 10% of cases of marginal zone lymphoma, and 4% of cases of CLL, but is absent from cases of IgG-MGUS, pointing to its potential use in diagnosis [238]. Other studies have described certain cases of WM as “smouldering,” which is generally taken to mean that they develop slowly, and have reported that the frequency of a panel of abnormalities [+14, del(6q23-25), +12, and +18q11-23] increased progressively from 18% of cases of IgM-MGUS cases to 20% of smouldering WM cases and to 73% of symptomatic WM cases. This suggests a multi-step transformation of clonal B cells that already harbour the phenotypic and molecular features of a malignant WM clone [239]. Similarly, others reported the L256P mutation in MYD88 to be present in 27%, 80% and in 85% of cases of IgM-MGUS, smouldering WM, and WM, respectively [240].

The observation that 27% of WM patients had a CXCR4 mutation (located at 2q22.1, which codes for the receptor for the chemokine stromal cell-derived factor 1, also known as CXCL12) was followed by its identification in MGUS, while another study used NGS to identify KMT2D mutations (located at 12q13.12, which codes for lysine methyltransferase 2D) in 24% of WM cases and in 5% of IgM-MGUS cases [241–243]. The combination of high frequencies of cells with both MYD88 and CXCR4, compared to low levels of both, gives a hazard ratio (95% CI) of 3.5 (1-4-9.3) for progression to symptomatic WM [244]. The National Genomic Test Directory for England [52] refers to a multi-target NGS panel for detecting small variants in MYD88 and CXCR4 in the investigation of WM and MGUS.

Investigation of the molecular pathology of this condition has been, and continues to be, informed by that of myeloma, which dominates the literature. A key observation in one study was that although the incidence of translocations at 14q32 (the site of the IgH locus) was similar, +13 was present in 40% of patients with a myeloma or with PCL, but in only 21% of MGUS cases, suggesting a transformation route from MGUS to myeloma, a finding subsequently confirmed in several studies [236, 245, 246].

Other authors have noted that t(14;20)(q32;q12)/(IGH::MAFB) is present in 1.5% of myeloma cases but in 5% of MGUS cases [247], that deletions of TP53, although common in myeloma, are absent in MGUS [248], and that chromosome 13 abnormalities are strongly associated with t(4;14)(p16;q32)/(FGFR::IGH), which is present in 10.3% of MM cases and in 9.6% of MGUS cases [249]. This has implications for oncogenesis, as 4p16.2 is the location of FGFR3, coding for fibroblast growth factor receptor 3, with mutations in this gene having been implicated in several cancers [250]. Multiple cytogenetic abnormalities are common: for example, in one series, t(14q32) and del(13q14) were present in 2% of MGUS cases but in 18% of MM cases [251].

However, not all studies support this: a case-control (243/1285) GWAS analysis of MGUS in a German population reported 10 risk loci on 8 chromosomes, but none of these were significant (p < 0.05) in a parallel study of 294 cases and 272 controls in a Czech population, a finding emphasising the need for caution [252]. The C allele SNP in ULK4, at 3p22,1, which codes for a serine/threonine kinase, has an OR (95% CI) of 1.32 (1.02–1.72) for MGUS and 1.39 (1.04–1.86) for myeloma [253]. Sun and colleagues used microarray analysis to probe plasma cell mRNA from 334 MGUS patients, 40 of whom progressed to MM, and found that the downregulation of IGLV1-44 at 22q11.22, IGKC at 2p11.2, IGHA1 at 14q32.33, PTPN1 at 20q13.13, and ECHDC2 at 1p32.3 was linked to progression to MM [254].

One of the first studies reported abnormalities in 251 patients, the most common being a trisomy with an IgH translocation in 43.9%, t(11;14)(q13:q32) in 16.2%, and t(4;14)(p16;q32) in 10.3% [255]. The significance of t(11;14) is that it brings together CCND1 at 11q13.3 with IgH at 14q32.22, often leading to high CCND1 expression. A concurrent review pointed to del13q as a common feature of all stages of the myeloma pathway, with primary genetic events in MGUS being IGH translocations, hyperdiploidy, and cyclin D dysregulation, and secondary events in SMM and MM being mutations in NRAS (in 24% of cases), KRAS (27% of cases) and BRAF (4% of cases), with inactivation of TP53, PTEN, and RB1, and with mutations in the NFκB pathway [256]. Many of these were confirmed, and new data were provided by researchers such as Busteros et al. in a study of 214 patients. These authors reported (in approximate order of frequency) hyperdiploidy (55% of patients), del(13q) (45% of patients), +1q (27% of patients), del(16q) (20% of patients), del(6q), del(14q) (both accounting for ∼13% of patients), del(22q) (12% of patients), del(8p) (10% of patients), del(1p) (8% of patients) del(20p), amp(8q24), and del(17p) (all accounting for ∼5–6% of patients), amp(2p), and del 4q (both accounting for ∼4% of patients), and others. Therefore, this is yet another example of the importance of del(17p), in that the tumour suppressor TP53 is located at 17p13.1. The most common translocations were t(11;14)(∼12%), t(4;14)(∼10%), and t(14;20)(∼2%), while mutations in KRAS (∼13% of patients), NRAS (∼6% of patients), and BRAF, TP53, ATM, DIS3, and FAM46C (all accounting for ∼2% of patients) were the most common gene abnormalities. Biallelic inactivation was present in 6% of patients, primarily involving TP53, RB1, CDKN2C, ZNF292, DIS3, and FAM46C [257].

Numerous genetic abnormalities have been linked to disease progression, such as those in MYC, BRAF, FAM46C, NRAS, t(4;14), t(6;14), and deletions in 1p, 14q, 16q, and 17p, while abnormal cyclins have also been postulated as a common driver. In one study, KRAS mutations were associated with a hazard ratio (95% CI) of 3.5 (1.5–8.1) for shorter time to progression. In another study, t(4;14) resulted in a median time to disease progression of 28 months compared to 55 months if t(11;14) was present [255–258]. Despite the importance of single gene/chromosomal abnormalities, there is a desire to combine genes and other factors to develop scoring systems for the risk of disease progression.

Khan et al. presented data from 105 patients on the potential of a four-gene panel to predict SMM progression to MM, these being (in order of predictive power) RRM2 (located at 2p25.1, coding for ribonucleotide reductase regulatory subunit M2), DTL (located at 1q32.3, coding for a ubiquitin protein ligase homolog), TMEM48/NDC1 (located at 1p32.3, coding for a transmembrane nucleoporin) and ASPM (located at 1q313, coding for an abnormal spindle protein homolog) [259]. Using these as a scoring system, a cut-off point was identified for a subset of 14 patients with an 85.7% probability of requiring therapy, compared to the remaining 91 patients whose probability was 17.8%. Botta and colleagues focused on inflammation, suggesting the value of an 8-gene signature (IL8, IL10, IL17A, CCL3, CCL5, VEGFA, EBI3 and NOS2), which could identify MGUS/SMM/MM with 84% accuracy [260]. Other scoring systems, some of which include non-gene features (primarily biochemical), refer to cytogenetics, such as t(4;14), t(14;16), +1q (some 1q21), del 17p (some more precisely 17p13), monosomy 13, and del 13q, in addition to genes with mutations in TP53, ATM, KRAS, NRAS, and MYC [258, 260–262].

The earliest reports of cytogenetic abnormalities in MM were of extra bands in 14q, i.e., +14q [263–265], followed by abnormalities at 17p, t(11;14), t(4;14), and +11q13 [249, 266, 267] and c-MYC, N-RAS, K-RAS, and TP53 [268–271]. As mentioned above, t(4;14)(p16.3;q32) provides the basis for at least one transforming event, as it brings together the IGH locus and the FGFR3 proto-oncogene, as does t(11;14)(q13:q32) with BCL1/CCND1 at 11q13.3, coding for the transcription regulator cyclin D1 (each present in ∼25% of cases), and t(4;14)(p16;q32) with the WHSC1/MMSET/NSD2 locus at 4p16 [236, 272–274].

The presence (in 45% of cases) of hyperdiploidy, most often of chromosomes 3, 5, 7, 9, 11, 15 and 17, although −13q is considered to be an important pathophysiology event, while del(17p), del(1p) and +1q21 may be secondary events. The majority of the remaining 55% of cases are characterised by reciprocal translocations between IGH and a number of oncogenes, such as FGFR3/MMSET (t(4;14)), CCND3 (t(6;14)), CCND1 (t(11;14)), MAF (t(14;16)) and MAFB (t(14;20)), [275]. Awada and colleagues summarised the frequencies of KRAS and NRAS (each present in ∼20% of cases), FAM46C and DIS3 (present in ∼11% of cases each), TP53 (present in 8% of cases), and BRAF (present in 6% of cases), with TRAF3, LTB, and ATM present in <5% of patients [276]. Several of these were also described by Maura et al., who reported the potential roles of the histone-coders HIST1H1B/H1-5, HIST1H1D/H1-3, HIST1H1E/H1-4, and HIST1H2BK/HSBC12, all at 6p22, with FUBP1 at 1p31.1 and MAX at 14q23.3, both of which code for MYC-related factors [277].

Several groups have used NGS to recognise candidate genes involved in the development and progression of myeloma. Zhan and colleagues probed CD138⁺ plasma cells from 74 patients with MM, using Affymetrix gene chips to identify 50 downregulated genes (including SDF1, TNFRSF7, RNASE6, APOC1, DEFA1 and LYZ) and 70 upregulated genes (including CDKN1A, EIF3S9, GMPS, H1F2, LAMC1 and PTPRK) [278]. Shaughnessy et al. further demonstrated the strengths of NGS by reporting 70 genes linked to MM, the majority of which were found on chromosome 1 [279]. However, Greenberg and colleagues published a list of 22 genes associated with MM (BAX, CASP9, CD4, CYP1A1, DNAH11, DNTB, HGF, HPSE, IL-1RN, IL6, IL1A, IL1B, IRS1, ITGA6, KLK3, LAG3, RIPK1, SERPINE1, TRAF3, ULK4, VCAM1, and XRCC4) [280]. Notably, none of these 22 genes are shared with the 70 genes described by Shaughnessy et al., illustrating the difficulty of interpreting certain analyses and, in sum, casting doubt on the particular methods and the comparability of the subjects from whom the samples were obtained.

The development of overt bone disease has been linked to the over-expression of genes such as TNFRSF11A at 18q21.33, which codes for the receptor activator of NFκB (RANK) and its ligand RANKL, coded for by TNFSF11 at 13q14.11, and to TNFRSF11B, at 8q24.12, coding for osteoprotegerin, members of the signalling NOTCH and WNT families. Meanwhile, mutations in IL6 and its product, with its receptor, are known to have effects on the bone marrow microenvironment [236] and have a place in the PI3K/AKT/mTOR second messenger pathway [281]. Stein and colleagues compared 182 untreated patients with myeloma, 329 patients undergoing treatment, and 294 patients at or near relapse, finding mutations in RAS-RAF in 31.3%, 35.6%, and 46.6%, respectively, NRAS in 14.3%, 16.1%, and 24.5%, and TP53 in 6.6%, 9.7%, and 17% respectively (all p < 0.01). These data point to potential roles for different genes as drivers of disease development, but, notably, there were no trends in BRAF ^V600E , TRAF3, FGFR3, RB1, CDKN2C, DNMT3A, ATM/ATR, TET2, or BIRC3 [282].

A research group from China sequenced 400 plasma cell genes from 50 MM patients, reporting that 76% had a TP53 mutation, 18% had an NRAS mutation, and 14% had a BRAF mutation, while low levels of BCL6 at 3q27.3, coding for a transcription repressor (a mutation also common in B cell lymphomas), BIRC3 at 11q22.2, coding for an inhibitor of apoptosis, HLA-DQA1 at 6p21.32, and VCAN at 5q14.2-14.3, coding for a proteoglycan, were linked to a poor prognosis [283]. Uckun and Qazi focused on mRNA for the ERBB isotypes (coding receptor kinases) in 787 patients, reporting no difference in levels of ERBB1 at 7p11.2, coding for the epidermal growth factor receptor, ERBB2 at 17q12, coding for HER2, or ERBB3 at 12q13.2, coding for HER3, according to disease stage. However, patients with the highest tertile of ERBB2 message had the most adverse outcomes with a hazard ratio (95% CI) of 2.34 (1.30–4.22), although age, serum beta-2-microglobulin, and albumin also affected outcomes [284].

Wallington-Beddoe and Mynott suggested that cytogenetic abnormalities such as trisomies (of odd-numbered chromosomes, present in 40%–50% of patients), t(11;14) (involving CCND1, present in 15%) and t(6;14) (involving CCND3, present in 5%) all indicate a favourable prognosis. However, they also reported that +1q (the location of CKS1B, present in 35%–40% of cases), del(1p) (FAM46C, CDKN2C and FAF1, present in 30% of cases), and abnormalities in MYC at 8q24 (present in 15%–20% of cases) all indicate a poor prognosis, with −13 (affecting RB1 and present in 45%–50% of patients) indicating an intermediate prognosis, and that the prognosis if t(14;14) is present (affecting FGFR3 and MMSET, present in 15% of patients), is poor to intermediate, although Heider et al. suggested that this translocation indicates a high risk, as do t(14;16) (affecting MAF, present in 3%–5% of cases) and t(14;20)(affecting MAFB, present in ∼1% of cases) [285, 286]. Black and Glavey considered the presence of t(11;14) and t(6;14) to be a standard risk factor for poor overall survival, whereas the presence of t(4;14), t(14;16), t(14;20), del(17p), +1q, and −13 [275] are considered high-risk factors.

Perhaps unsurprisingly, an increased number of circulating plasma cells indicates a poor prognosis in MGUS and MM [287]. Perroud and colleagues used an NGS panel comprising CCND1, DIS3, EGR1, FAM46C (TENT5C), FGFR3, PRDM1, TP53, and TRAF3, along with seven hotspots in BRAF, IDH1, IDH2, IRF4, KRAS, and NRAS in 87 patients with newly-diagnosed MM and 11 patients with relapsed/refractory MM, finding that the mutational load was generally higher in relapsed disease. Despite the very small sample size, TP53 was the most prevalent mutation, occurring in 13% of patients with newly diagnosed MM but in 81% of patients with relapsed/refractory MM [288]. NICE guidelines NG35 “Myeloma: Diagnosis and Management” suggest using FISH on CD138-selected bone marrow plasma cells for the detection of t(4;14), t(14;16), 1q gain, del(1p) and del(17p)(TP53 deletion), t(14;20), and the standard-risk abnormalities t(11;14) and hyperdiploidy [289].

The National Genomic Test Directory for England has a section on “Plasma cell dyscrasia” [52]. It describes the use of an NGS multi-target panel for small variant detection in KRAS, NRAS, BRAF, TP53, DIS3, TENT5C, and IRF4; for structural variant rearrangement detection in IGH::FGFR3, IGH::CCND3, IGH::CCND1, IGH::MAF, IGH::MAFB, and MYC; and for copy number variation detection with respect to hyperdiploidy, del(1p), +1q, and del(17p). Other analyses, using FISH/RT-PCR, include t(4;14) (IGH::FGFR3), t(6;14) (IGH::CCND3), t(11;14)(q13;q32) (IGH::CCND1), t(14;16) (IGH::MAF), t(14;20) (IGH::MAFB), and for IGH and MYC rearrangement, hyperdiploidy copy number, and del(1p), +(1q), and del(17p) (location of TP53) copy numbers, all by FISH.

PCL may be classified as primary (pPCL), where it arises without prior evidence of a MM, or secondary (sPCL), where it is known to arise from pre-existing MM. Mosca et al. used FISH on pathological samples from 23 patients with pPCL, finding that87% of cases harboured an IGH translocation, with the most common being t(11;14) in 40% of cases and t(14;16) in 30.5% of cases, with abnormalities observed in 1p (38%), 1q (48%), 6q (29%), 8p (42%), 13q(74%), 14q (71%), 16q (53%), and 17p (35%) [306]. They also reported a biallelic deletion in 8p21.2, the location of PPP2R2A, coding for a protein phosphatase subunit, which the authors considered to belong to a family of putative tumour suppressors. Gowin and colleagues summarised the cytogenetics as pPCL being commonly hypodiploid and the IgH translocation t(11;14) the most prevalent, with sPCL more likely to be hyperdiploid (having possibly evolved from an MM clone) and with more diverse IgH translocations, such as t(11;14), t(4:14) and t(14;16), and so on [307].

Tiedemann and colleagues compared 39 cases of sPCL with 49 of pPCL and 439 cases of MM, finding the frequency of hypodiploidy to be 42%, 60% and 40%, respectively, and hyperdiploidy to be 17%, 0% and 60%, respectively. Despite the small number of cases, the difference in the presence of t(11;14) in 71% in pPCL cases and in 23% in sPCL cases was statistically significant [308]. The cytogenetics of pPCL and MM have been reported by three research groups, the most consistent results being the increased frequencies of t(11;14), t(14;16), and del(17p), the latter two showing the greatest increases (Table 19). This supports the hypothesis of a transformation driver at one or both of these sites [309–311]. Gundesen et al. [309] and Lionetti et al. [311] also reported amp-1q/1q gain in 32%/48% of pPCL cases and 40%/51% of MM cases, respectively. The latter research group also reported t(14;20) frequencies of 5% and 3%, respectively. Avet-Loiseau et al. [310] also reported that t(11;14) and del(17p) were linked to a poor prognosis in pPCL, while Tiedeman found that MYC translocations were linked to poorer survival in pPCL [308]. Chang reported outcomes in 14 patients with pPCL (with an overall survival of 45 months) and 26 with sPCL (with an overall survival of 19 months), which may be a false negative (p = 0.09) due to the small sample size [304].

Todoerti and colleagues used the power of Affymetrix NGS to probe total RNA from 21 pPCL cases and 55 MM cases, identifying a 503-gene panel that could distinguish between the two types of cancer, and a further 27-gene panel with potential clinical relevance. Of these 27, 10 (PECAM1 (at 17q23.3), MKX (10p12.1), CALCRL (2q32.1), C3orf14 (3p14.2), ALDH1L2 (12q23.3), WARS (14q32.2), SLC15A2 (3q13.33), RNU5D (1p34.1), CTH (1p31.1), and (1p32.1)) were linked positively to survival, with 17 (FAM111B (at 11q12.1), MCTP1 (5q15), C10orf10 (10q11.21), FNBP1 (9q34.11), EFEMP1 (2p16.1), FAIM3 (1q32.1), CPEB4 (5q35.2), EDN1 (6p24.1), PVALB (22q12.3), LY86 (6p25.1), LAPTM5 (1p35.2), PARP15 (3q21.1), PLEKHF2 (8q22.1), PDK4 (7q21.3), TNFAIP3 (6q23.3), FAM105A (5p15.2), and TCN2 (22q12.2) linked negatively to survival [312].

A further study analysed the expression levels of genes across the spectrum of plasma cell dyscrasia, including normal plasma cells, MGUS, SMM, MM, and PCL. A clear downward sequential trend was found across the five groups for BTN3A2, C13oft15, CD36, FOS, TSTL1, and MGC29506, and a clear upward expression trend was found for BUD31, C16orf42, PSMD6, CDC42SE1, RHOA, CORO1A, EIF1B, GNB2, MRPS18A, MVP, and MYL12B. EZH2 expression was linear between normal controls, MGUS and SMM, then increased in a stepwise manner for MM and PCL, while ILK expression increased only in PCL. Compared to expression in normal controls, PRG3 expression was lower in MGUS, SMM and MM, but reduced in PCL. These data provide fascinating insights into the role of certain genes in the natural history of plasma cell dyscrasias, from health to frank leukaemia. The National Genomic Test Directory for England has a section on plasma cell dyscrasias, but does not refer to PCL or either of its subtypes [52].

One of the earliest and most thorough investigations of these molecules, by Pichiorri and colleagues, used mRNA and miRNA microarray chips and RT-PCR to probe 49 MM-derived cell lines and samples from 6 normal donors, 6 patients with MGUS and 16 patients with MM [313]. The principal findings were 5.8-fold–15.6-fold increases in the expression of miR-21, miR181a, miR-93, miR-106b, miR-25, and miR-106a, and a 0.15-fold decrease in the expression of miR-328 in MGUS. Similarly, in MM, there were 4.6-fold–288.9-fold increases (depending on whether they originated from plasma cells or from transformed cell lines) in the expression of miR-25, miR-32, miR-20a, miR-93, miR-106b, miR-106a, miR-181a, miR-21, miR-19a, miR-19b, miR-181b, and miR-92a, with a 0.48-fold reduction in the expression of miR-328. The authors speculate that these abnormalities may contribute to malignant transformation by factors such as promoting plasma cell survival and blocking apoptosis. Some miRNAs, such as miR-21 and miR-32, are differently expressed in MGUS and MM, with others, such as miR-221 and miR-222, exhibiting similar levels in both conditions [313, 314]. Chi et al. built on these cross-sectional studies, probing samples from 33 MM cases, 5 MGUS cases and 9 controls, reporting that 109 miRNAs were upregulated and 20 were downregulated (on average twofold) in MM relative to controls [315].

Jones and colleagues described a microarray/TaqMan RT-PCR method for detecting serum miRNAs, using it to suggest that miR-720, miR-1308 and miR-1246 have potential as diagnostic biomarkers in MM and that combinations such as miR-720 and miR-1308, and miR-1246 and miR-1308 can distinguish between different groups of cases and controls [292]. This theme was extended by Kubiczkova et al., who used TaqMan/qPCR methods to report altered miRNAs, such as miR-34a and let-7e, in MGUS and MM, but also to show that levels of miR-744 and let-7e are linked to overall survival and time to progression [316] (Table 20). Similarly, Li et al. concluded that serum miR-134-5p, miR-107, and miR-15a-5p are also potential biomarkers in MGUS and MM [317]. Lionetti et al. [311] reported 423 upregulated and 41 downregulated miRNAs in pPCL compared to MM. Of these, relative (high/low) expression of miR-22, miR-146a, miR-92a, and miR-330-3p was found to be linked to progression-free survival (PFS).

The power of NGS was once again demonstrated by Ronchetti and colleagues, who used an Affymetrix GeneChip array and qRT-PCR to probe CD138⁺ plasma cells from 268 individuals representing the spectrum of plasma cell dyscrasias [318]. The expression of 21 lncRNAs was progressively deregulated across the spectrum. Examples include RLIM-6 at Xq13, with a correlation coefficient of 0.37; ANGPTL-1 at 1q25, with a correlation coefficient of 0.3; and SERPINC1-1, also at 1q23, with a correlation coefficient of 0.29, suggesting roles in disease progression. A regulatory role may be present as these three lncRNAs are antisense to SCL16A2, RALGPS2, and ZBTB37, respectively. Butova et al. used a similar approach, with Illumina NGS and RT-qPCR to report that 52 different lncRNAs were significantly deregulated between MM and PCL samples [319]. Of these, the expression of both LY86-AS1 and VIM-AS1 was significantly reduced in PCL compared to MM, suggesting possible links with disease progression.

NGS and qRT-PCR were also used by Todoerti and colleagues to focus on lncRNAs in MM and pPCL patients with t(11;14). These authors found 38 lncRNAs to be differentially expressed, with three increased and 35 decreased in pPCL [320]. Furthermore, lower expression of Linc00886 was linked to adverse PFS and overall survival, NINJ2-AS1 and Linc02728 to poor PFS, while higher expression of SNHG6 was linked to poor overall survival, although in multivariate analysis, only the latter remained a significant predictor of overall survival. Data from Li and colleagues reported increased expression of the lncRNA UCA1 in MM, which correlated inversely with that of miR-331-3p. Other data pointed to a role of UCA1 in suppressing apoptosis, thus promoting proliferation [321].

Yang et al. reviewed lncRNAs in MM, showing their relationships with miRNAs, their effect on genes such as MYC and on second messengers, and their roles in PFS and overall survival, several of which, such as ANRIL and HOTAIR, have roles in numerous malignancies [322]. A review by Lei and colleagues summarised the potential of lncRNAs such as PCAT1 and LINC0017 as therapeutic targets and described the mechanisms by which lncRNAs and miRNA may act on genes to effect malignant transformation [323].

Although only recently (in terms of miRNAs and lncRNAs) described in depth, numerous roles for circular RNAs (circRNAs) have been reported. One of the earliest circRNAs to be discovered, circ_0000190, was found to be downregulated in MM tissues and in plasma compared to controls, and high/low expression levels were linked to PFS and overall survival. A possible mechanism for this may be via repression of miR-767-5p and the MAPK4 transduction pathway [324]. Similarly, Liu et al. reported increased expression of circRNA-101237 in MM, especially in those with cytogenetic abnormalities del(13q14), amp(1q21), del-p53, t(4;14) and t(11;14), and that high levels were linked to poor PFS and overall survival [325]. Other studies have shown the possible influence of circRNAs on drug resistance, on bone metabolism, and as tumour suppressors [326–328]. Mirazimi et al. reviewed circRNAs in MM, Peres and colleagues reviewed circRNAs in leukaemia, lymphoma and MM, and Lei et al. summarised data on small interfering RNAs in MM, such as IONIS-AR-2.5Rx, which may target GALNT2 at 1q42.13, a gene that codes for a metabolic enzyme [235, 323, 329].

Described in detail in 1997 [351], this fascinating condition provided the opportunity to directly link a precise genetic lesion with clinical and laboratory features. VHL at 3p25.3 codes for the 24 kDa Von Hippel Lindau protein that binds to the hypoxia-inducible factor 1α (HIF-1α), coded for by HIFA at 14q23.3, which itself has many roles, including as an important regulator of the body’s response to hypoxia, and therefore, the levels of EPO [352]. Loss-of-function mutations in VHL lead to the over-expression of HIF-1α, which is the basis of polycythaemia, and acts as a tumour suppressor [353, 354]. The most common form of polycythaemia (C598T) is estimated to have arisen from a single founding event 14,000 to 62,000 years ago, while VHL ^R200W may provide heterozygous protection against anaemia, thus explaining its persistence [355–357].

Having discussed excessive megakaryocytopoiesis as the root lesion in ET, a discussion of megakaryoblastic leukaemia naturally follows. First described in the mid-20th century, a genetic abnormality was reported 50 years ago, with strong links to myelofibrosis subsequently published, and an early report of three cases pointing to 11%–75% megakaryoblasts and platelet counts of 35–53 × 10⁹/L [379–383]. Breton-Gorius used monoclonal antibodies and electron microscopy to note the close association between megakaryoblastic leukaemia and acute erythroblastic leukaemia, while Cuneo and colleagues reported numerous chromosomal abnormalities, such as −5/5q, −7/7q, +8 and +21, rearrangements in 3q21 and 3q26, inv(16)(pl3;q22), t(13;20)(ql3 or l4;ql1), and der(7)t(7;17)(p14;q22) [384, 385].

In an elegant series of experiments, Terui and colleagues provided details of the precise cell biology and aetiology of the disease. They found that a soluble product of megakaryoblasts stimulates collagen production by bone marrow fibroblasts and that this soluble factor is transforming growth factor-β. The authors also found that the megakaryoblasts express substantially increased levels of their mRNA [386]. Other authors have reported potential roles for additional growth factors, PDGF and FGF, although these may have a role in the myelofibrosis aspect of the disease [387]. Genes mutated in megakaryoblastic leukaemia include GATA1 (especially in G21 trisomy associated with Down’s syndrome, where RUNX1 is also implicated), a fusion of RBM6 to CSF1R, KIT, and FLT3, JAK2, JAK3, and MPL [388–392].

These conditions were originally thought of as being a form of lymphoma, often presenting with dermal manifestations and defined by immunophenotyping [448–452]. Cytogenetics has since pointed to abnormalities in 5q, 6q, 9 (such as del 9p21.3, which is the location of tumour suppressors CDKN2A and CDKN2B), 12p, 13q, and 15q, with +7q, +22, and del(3p). Molecular genetics may indicate lesions in TET2, TP53, NPM1, FLT3, IKZF1, SRSF2, and the proto-oncogene MYC, the latter often forming a t(6;8)(p21;q24) [451–455].

A presumptive diagnosis of Langerhans cell histiocytosis (LCH) can be made based on clinical, radiological, histological (e.g., abundant eosinophilic cytoplasm) and immunocytochemical grounds (e.g., CD1a, CD68, CD163, CD207), although molecular genetics are likely to aid this process [456–458]. An early cytogenetic analysis of LCH reported del(7q22), del(7q22), −15, −16, −17, −18, t(2;4)(p21;q33), t(9;19)(q12; q13), and inv(13)(q21;q33) [459]. A genetic analysis of 61 cases revealed that 57% had BRAF ^V600E [457], while others noted altered MAP2K1 in 27.5% of cases [458] and abnormalities in KRAS, ARAF, NRAS, and CSF1R (located at 5q32 and coding for the receptor for the cytokine macrophage colony-stimulating factor 1, also known as colony-stimulating factor 1), which may be linked to transformation [460]. This sub-section also includes Langerhans cell sarcoma and indeterminate dendritic cell tumours and sarcomas [91], the precise molecular pathology for which is yet to be determined.

In 1988, Benz-Lemoine [461] and colleagues reported a malignant histiocytosis characterised by t(2;5)(p23;q35), which, with the benefit of 20 years of additional research [462], may have been a variant linked to ALK (coding for anaplastic lymphoma kinase at 2p23.2-p23.1), now recognised as an important malignancy [463]. In many cases, ALK has gene fusion partners that include CLTC, COL1A2, DCTN1, EML4, TFG, TPM3, and TRIM33 [464]. An important second member of this group is Erdheim-Chester disease, for which the leading genetic lesion is BRAF ^V600E , present in approximately 50% of cases [464]. This mutation, therefore, points to the possibility of treatment with an inhibitor of the mitogen-activated protein kinase (RAS-RAF-MEK-ERK) pathway (as is the case with Langerhans cell histiocytosis), although there may also be mutations in ARAF, MAP2K1, NRAS, and PI3KCA [465]. Pai and colleagues have estimated that Erdheim-Chester disease comprises 37% of all histiocytic disorders, marginally exceeding Langerhans cell histiocytosis at 34% [466]. Other conditions in this sub-section include Rosai-Dorfman disease (most commonly linked to RAS isoforms [KRAS, NRAS], MAP2K1, and ARAF), juvenile xanthogranuloma (linked to RAS isoforms and MAP2K1), and histiocytic sarcoma (linked to alterations in KRAS, BRAF, and MAPK1) [466–468].

In several instances, clinical studies have not specified or have merged the three major MPN groups into a single cohort with varying results [474–476], although others have presented parallel analyses. Hussein and colleagues reported the differential expression of 365 miRNAs in megakaryocytes from 18 cases of pMF, 18 cases of ET, and 8 cases of megakaryocyte hyperplasia, all controlled by 5 cases of normal haematopoiesis. The leading result was increased expression of miR-146b in pMF alone [477], although such small sample sizes raised concerns about false positives and negatives. Tombak et al. investigated six peripheral blood miRNAs in 22 cases of pMF, 22 cases of PV, 49 cases of ET and 40 controls. The principal findings were increased expression of miR-155 and reduced expression of miR-451 in all three disease groups, increased expression of miR-221 in pMF and ET, lower expression of miR-222 in PV, but higher expression of miR-222 in ET, and increased expression of miR-223 in pMF and ET. Although ET and pMF patients had higher levels of miR-223, there was no difference in the expression of miR-181a [478]. Afar and colleagues studied bone marrow from 40 cases of PV, 27 cases of ET, 40 cases of MF secondary to PV or ET, and used peripheral blood from 90 subjects as a control group. Their primary result was an increased frequency of the TT genotype in miR-146a rs2431697 in MF, with an increased frequency of the expression of IL-1β in the TT genotype in all three groups compared to the controls. However, the TT genotype was not linked to the transformation of PV or ET to MF [479]. Several research groups have reported miRNAs in each of the three major disease groups.

Bruchova and colleagues studied changes in the expression of miR-150, miR-155, miR-221, miR-222, miR-451, miR-16, miR-339, and miR-378 in peripheral blood mononuclear cells expanded in vitro by growth factors [480]. They found a positive correlation between the frequency of JAK2 V617F and the expression of miR-143, together with an inverse correlation with let-7a, miR-30c, miR-342 and miR-150 [481]. Data from Guglielmelli et al. suggest that deregulation of miR-16–2 contributes to the abnormal expansion of the erythroid lineage in PV [482]. Zhan and colleagues [483] probed circulating CD34-positive cells in 8 PV patients and 6 healthy controls, finding that 71 miRNAs were either up- or downregulated, hypothesising that four core species may act on target genes and thus have an effect on malignant transformation:

miR-575, targeting SFRS2, SFRS1, EPOR, HMGA2, and TFPI,

Navarro and colleagues extracted total RNA from the platelets of 19 patients with ET and 10 controls, quantifying the expression of 384 mature miRNAs. Of these, the ET samples showed a distinct signature of 70 species, 68 of which were downregulated compared to the control samples (with miR-9 and miR-431 being upregulated). Forty miRNAs differed between ET patients depending on their JAK2 status (mutated or wild type), 8 of which were proposed as being likely to activate the JAK/STAT pathway via SOCS1 and SOCS3 [484]. Trans et al. also probed platelets, finding that miR-10a, miR-28, miR-126, miR-155, miR-221, miR-222, miR-223, and miR-431 were downregulated, and that miR-9 (confirming data from Navarro et al) and miR-490 were upregulated compared with healthy controls. The expression of certain miRNAs correlated with metrics of platelet biology, such as mean platelet volume and P-selectin expression [485].

Calura et al. used NGS to profile 584 miRNAs and 18,654 genes in 73 CD34⁺ cells from 42 cases of pMF, and from 31 controls, ordering them into a large series of interacting networks with potential targets. Among the many observations were links between key genes that code for transcription factors (MYCN, ATF, CEBPA, REL, IRF, and FOXJ2) and miR-106a-5p, miR-20b-5p, miR-20a-5p, miR-17-5p, miR-19b-3p, and let-7d-5p [486]. Not only did Norfo et al. find 58 differentially expressed genes in pMF (with DEFA1 showing a 60-fold increase), they also observed changes in numerous miRNAs, describing gene/miRNA networks. The strongest interaction in these networks was found to be between miR-155-5p and tumour suppressor JARID2 at 6p22.3, with implications for megakaryocyte hyperplasia [487]. Rontauroli and colleagues suggested a role for upregulated miR-494-3p in megakaryocyte hyperplasia via an interaction with SOCS6 at 18q22.2, coding for suppressor of cytokine signalling 6 [488]. Similarly, Fuentes-Mattei et al. showed that miR-543 is increased in pMF and that it targets TET1 and TET2, both coding for enzymes with roles in the epigenetic regulation of DNA [489].

As discussed above, mutations in ASXL1 are a common feature of MDS (and other blood cancers). One such mutation causes an upregulation of miR-125a, which in turn leads to the repression of Clec5a at 7q34, coding for a lectin-like molecule with links to SYK, a tyrosine kinase [490]. Micheva and Atanasova summarised the potential diagnostic and prognostic roles of nine miRNAs in MDS, with the upregulation of miR-22 possibly involved in transformation by interacting with TET2 and PTEN. The authors also pointed to 9 upregulated and 13 downregulated species linked to del(5q), along with 15 with implications for therapy [491]. Many of these miRNAs were also described by Georgoulis and colleagues, who highlighted those of prognostic value together with other ncRNAs (circular, long non-coding RNAs, small nucleolar RNAs, and Piwi-interacting RNAs) [492].