Toward Integrated Genomic Diagnosis in Routine Diagnostic Pathology by the World Health Organization Classification of Acute Myeloid Leukemia

Significant milestones and seminal discoveries during 1674-1966, by individuals who have made crucial contributions toward progress in the diagnosis of hematologic neoplasms as we understand today are depicted chronologically in Figure 1, with selected references [1-11]. It is notable that the path to progress in the understanding of disease and neoplasms initially took centuries for significant discoveries (17th-18th centuries), and subsequently, many decades (19th-20th centuries) for a breakthrough or a change from the prevailing norm. Further, that progress always required perseverance, dedication, innovation, and collaboration among individuals that were not necessarily recognized by the majority at the time, as exemplified by the respectful collaboration between John Hughes Bennett and Rudolph Virchow, as described by Piller [1]. Abstract

"Leukemia," lymphomas, and cancer genetics: first descriptions As depicted in Figure 1, the first clear description for the thought of "leukemia" as the disease cause appears to have been in 1841 by David Craigie, a physician at Edinburgh Royal Infirmary, who observed a patient with thick blood and an enlarged spleen and who was puzzled by, and questioned, the cause of the clinical features in that patient [1,2]. Three years later, in 1844, Craigie attended to another patient with similar clinical features, and was convinced that the pathological cause was the same in both patients, and not pus and inflammation. John Hughes Bennett, a clinician, and pathologist at Edinburgh Royal Infirmary, who had received prior training with Alfred Donne, a French microscopist, performed an autopsy in Craigie's second patient and examined the deceased patient's blood under a microscope [1,2]. Bennett first reported in 1845 that the disease, which would now be recognized as chronic myeloid leukemia, was due to systemic involvement of blood (and not inflammation), along with Craigie's report of his first patient in the same journal. The second clinicopathologic report of leukemia was by Rudolph Virchow, a demonstrator of anatomy in Berlin, who, at age 24, described the unstained appearances of blood cells in the first report of a patient with chronic lymphatic leukemia. Subsequently, Virchow described the third case, also chronic leukemia with splenic enlargement, and recognized that there were two types of leukemiassplenic and lymphatic. While Bennett preferred the name "leucocythemia," Virchow coined the name "leukemia" for the disease, importantly, with both in agreement for the microscopic features in blood and the cause of the disease as systemic, instead of inflammation [1,2].
Five decades later, in 1900, the words "myeloblast" and "lymphoblast" were introduced by Naegeli, a Swiss hematologist, for precursor cells in the two types of acute leukemias. That nomenclature followed two crucial milestones: (1) first, in 1868, by Ernst Neumann, a Professor of Pathologic Anatomy at Konigsberg, who had stated that blood originated in the bone marrow, hematopoiesis is a continuous process, and that leukemias originate in the bone marrow, and (2) the seminal work of Paul Ehrlich, who, as a medical student in 1877, had described details of blood cells and developed stains to identify acidophil, basophil, and neutrophil granules in white blood cells [1].
Interestingly, malignant lymph node tumors (lymphomas), which represent the solid counterpart of hematologic lymphoid neoplasms, were first studied by post-mortem examination by Thomas Hodgkin at Guy's Hospital in London in 1832, a decade before Craigie observed his first patient with chronic leukemia in Edinburgh. Nevertheless, "Hodgkin disease" was first recognized by the medical world in 1865, two decades after Bennett and Virchow described leukemias due to Samuel Wilks, a curator at the same museum where Hodgkin had worked. Three decades later, at the turn of the 19 th -20 th century, Sternberg and Reed described the characteristic large nucleolated cells by histological examination in Hodgkin disease. Twenty-five years later, during 1925- J Clin Haematol. 2020 Volume 1, Issue 2 1927, Brill and Symmers described follicular tumors of germinal centers, and another 30 years later, in 1958, Dennis Burkitt described Burkitt lymphoma, which also arises from the germinal center [3,4].
In parallel with the milestones above in hematology in the second half of the 19 th century, the field of the underlying genetics in cancer started with the observation by a German pathologist, Hansemann, who in 1890 first observed mitoses in tissues of malignant tumors. Two decades later, in 1914, Theodor Boveri hypothesized that chromosomal abnormalities caused the transition from benign to malignant. However, it was not until 1955 when Tjio and Levan at the Institute of Genetics in Lund ascertained the correct number of human chromosomes as forty-six [5]. Subsequently, in the late 1950s, the abnormally minute "Philadelphia chromosome" was identified as a causative abnormality in 7 patients with chronic myeloid leukemia in the absence of any observed chromosomal abnormalities in 10 AML patients [2,6].
Significantly, also in the late 1950s and 1960s, although not frequently noted, Baikie in Edinburgh, and Sandberg at Roswell Park Memorial Hospital in Buffalo, had described cytogenetic abnormalities in AML cases and suggested that cytogenetics could classify AML [7][8][9]. However, cytogenetics was not formally incorporated to classify AML until four decades later, when the 2001 WHO classification of tumors was introduced [12].

The morphologic and subsequent immunologic era for hematologic cancers
Ehrlich's groundbreaking work in 1877 initiated the morphological era for hematology, which has progressed for almost 1½ centuries. Major progress in the diagnosis and classification of leukemias was achieved by careful examination of peripheral blood and bone marrow aspirate smears, including cytochemical-stained smears, by the FAB Co-operative Group. As shown in Table 1, the FAB group classified acute leukemias during 1976-1985, and subsequently, chronic lymphoid and myeloid leukemias [13][14][15][16]. Figure 2 shows the AML subtypes by FAB classification, based on the differentiation of the leukemic cells and the extent of myeloid and monocytic maturation, as determined by microscopic evaluation of Wright-Giemsa-stained and cytochemical-stained smears [13,14].

The incorporation of cytogenetics and molecular genetics for diagnosis in acute myeloid leukemia
As noted earlier, as early as the late 1950s, cytogenetics was suggested to classify AML possibly [7][8][9]. As previously described, the diagnostic paradigm shifted only in 2001 when, based on the principles of the REAL classification, the WHO classification for hematopoietic and lymphoid tumors was introduced, including cytogenetics incorporated for acute leukemia classification [12,24]. That WHO volume represented the 3 rd edition, which had Most importantly, the WHO 2001 classification represented a collaboration of pathologists, oncologists, and geneticists worldwide to standardize criteria for the definition and classification of cancer types and standardized nomenclature to ensure progress in clinical cancer care [12].
Since normal cytogenetics comprised 45-50% of all AML, molecular genetic abnormalities were investigated for risk stratification in a normal karyotype. That motivation led to studies that established the prognostic significance of abnormalities in the FLT3 and NPM1 genes. The inactive FLT3 receptor is monomeric and is normally present on bone marrow CD34+ stem cells and immature hematopoietic progenitors (myeloid, monocytic, and B-lymphoid) [29]. The FLT3 receptor dimerizes in its active form, with phosphorylation activating downstream pathways for cell proliferation [29]. Mutations in the FLT3 gene, located on chromosome 13q12, are most well-known to occur as internal tandem duplication in the juxtamembrane domain (FLT3 ITD ) and as point mutations in the tyrosine kinase domain (FLT3 TKD ), and these mutations lead to constitutive activation of the FLT3 receptor leading to uncontrolled cell proliferation [29]. In 2001, in a study of 854 AML, the presence or absence of FLT3 ITD mutations in the intermediate-risk cytogenetics group led to increased or decreased relapsed disease, respectively [30]. In 2002, another study of 224 AML showed the worst survival if FLT3 ITD -positive, intermediate if FLT3 TKD -positive, and best with absent FLT3 mutations [31]. In 2005, Falini et al. identified insertion mutations in NPM1 that led to abnormal cytoplasmic localization of the nucleolar protein in AML [32]. Subsequently, the prognostic role of FLT3 and NPM1 mutations in AML was ascertained, with the favorable effect of NPM1 mutations present only if FLT3 ITD mutations were absent [33].
In 2001, Pabst et al. described mutations in AML in the CEBPA gene, which encodes for the transcription factor CEBPA, crucial for granulocytic differentiation [34].  Interestingly, germline CEBPA mutations underlying familial AML were identified in 2004 [35], several years after anticipation was observed in one pedigree with the same type of familial AML [36].
After the Human Genome Project was completed in 2003, the role of mutations in AML pathogenesis was intensely investigated, as reviewed earlier [37]. The Cancer Genome Atlas study of 200 de novo AML revealed gene mutations with intricate co-operation patterns and mutual exclusion between and within eight categories of biologically functional genes [38]. Mutations in NPM1 co-occurred with mutations in FLT3 or DNMT3A [38], similar to prior genomic profiling [39], while gene fusions were mutually exclusive of mutations in NPM1, RUNX1, TP53 and CEBPA [38]. Importantly, NPM1 mutations were not detected in age-related clonal hematopoiesis, even with deep sequencing [37]. Subsequently, two types of short-read sequencers enabled analysis of the cancer genome by massively parallel sequencing, termed NGS, in clinical laboratories.

The World Health Organization classification of AML WHO 2001 -WHO 2016/2017
The evolution of the diagnostic criteria for AML was previously described, with a schematic presented in Figure  3 3) Therapy-related AML. The 4 th category, AML, not otherwise specified (AML-NOS), was created for those AML cases that could not be classified into any of the 3 specific categories. By WHO 2008 , (1) two major categories, AML-RGA and AML with myelodysplasia (MDS)related changes (AML-MRC), were refined with additional inclusive cytogenetic criteria, (2) additional categories were recognized that needed to be excluded prior to rendering a diagnosis of AML, including acute leukemias with ambiguous lineage, myeloid/lymphoid neoplasms with eosinophilia and gene abnormalities, and blastic plasmacytoid dendritic cell neoplasm. The category of AML-NOS was retained similar to WHO 2001 . By WHO 2016/2017 , AML with mutations in NPM1 and biallelic CEBPA were included as 2 new subtypes. A major change in the WHO 2016 was that multi-lineage dysplasia alone by morphology (microscopic examination) was no longer a criterion for AML-MRC, the diagnosis of which required either MDS-specific cytogenetic abnormalities and/or a prior history of MDS or MDS/ myeloproliferative neoplasm. A new category was introduced for germline predisposition in myeloid neoplasms, to encourage recognition of familial myelodysplastic syndromes and AML.

Integrated Genomics for Acute Myeloid Leukemia Diagnosis and Classification
Most significantly, in 2020, we understand AML to be an aggressive clonal hematologic malignancy characterized by marked heterogeneity in genetic and clinical features, with this "single" disease comprised of numerous subtypes, with different prognosis and treatment options. Precise classification of AML, necessary for risk stratification and therapy, is best achieved by the WHO 2016/2017 classification [41,42], which requires not only microscopic skills, but also availability and collaboration for multiple modalities of tests, including clinical history, laboratory hematology, FCI, cytogenetics, and molecular genetics. Table 2 shows the WHO 2016/2017 AML classification, including seven subtypes based on chromosomal translocations/inversion, AML-NPM1 mut , and AML-biCEBPA mut based on gene mutations, and two provisional subtypes [41,42].
As reported in 2019, careful application of the WHO diagnostic criteria leads to the precise classification of the genetically-defined AML subtypes [24]. Here, features previously reported in text/tables [24] are highlighted in the following Figures: Figure 4A-4C, Study design, methods, and excluded cases; Figure 4D, Distribution of all cases by WHO 2008 and WHO 2016/2017 ; Figure 4E, WHO 2008classified cases reclassified by WHO 2016/2017 ; Figure 5A, Median age, numbers of males and females in all AML categories. Additionally, the median age for females with AML-NPM1 mut was lower (49 versus 62 years) than in males with AML-NPM1 mut [24]; Figure 5B, Hematologic findings in all categories; Figure 5C, Cytogenetics findings, and survival outcome in all categories; Figure 6, Smoking status in all molecularly-confirmed AML-NPM1 mut ; all three non-smokers with familial AML-NPM1 mut had a history of female relatives with leukemia [24]; also, one young AML-NOS patient with features suggestive of AML-NPM1 mut and with history of familial leukemia, was a non-smoker [24]; 5B: Overall abnormal hematologic values for all patients, showing (a) Leucocytosis was present in a greater percentage of AML NPM1 mut cases than in AML with recurrent genetic abnormalities (AML-RGA), AML with myelodysplasiarelated changes (AML-MRC), and AML-NOS. (b) AML-NPM1 mut was characterized by anemia, leucocytosis, and thrombocytopenia. As previously noted [24], by WHO criteria, all AML-NPM1 mut were de novo AML, explaining possibly why thrombocytosis was observed in prior FAB-classified cohorts of AML with NPM1 mutations, which had included secondary AML.      J Clin Haematol. 2020 Volume 1, Issue 2 [44]. Table 4 highlights that in the cohort by Green et al., the percentage of AML-biCEBPA mut patients was highest in the youngest (15-29 years) age group [45].   [45].

Characteristics
Epidemiologic studies are required to determine if there is a real increase in AML with gene mutations in NPM1 and CEBPA in young patients, especially women and if yes, why? Why are women more affected by AML with genetic mutations in NPM1 and CEBPA? Further, studies are necessary for familial AML, including AML-NPM1 mut , which comprises the most frequent genetic AML subtype to date in European cohorts [24,46], including for molecular epidemiology, which could potentially identify AML subtypes (familial or sporadic) that may be preventable in the future by lifestyle changes [47], including smoking, which increases the risk of AML [48], or by continued advances in precision gene editing [49].
For routine integrated genomics diagnosis, the WHO classification is meant to be applied worldwide, including in countries with limited resources or advanced molecular genetic techniques. Even in the developed world, laboratory hematologic tests with microscopic evaluation of peripheral blood and bone marrow smears are crucial for a prompt diagnosis of AML. Indeed, a specific diagnosis of APL may be preliminarily rendered with blood smear examination even before a bone marrow biopsy and rapid testing for PML-RARA are performed, particularly if combined with rapid myeloperoxidase cytochemical staining that typically shows intense diffuse positivity in APL even in the hypogranular variant, and if present with disseminated intravascular coagulation. The distinction of an APL from a non-APL--AML is the most urgent priority of the pathologist in communication with the clinician, for the correct life-saving therapy. Morphologic smear review is also invaluable to diagnose a possible AML with monocytic or myelomonocytic differentiation, particularly when a neoplasm might not be clinically suspected. Even with minimal monocytosis, a careful smear examination must be performed for abnormal, possibly neoplastic monocytic cells, as illustrated in the WHO books [12,40,42], with bone marrow biopsy performed if needed to rule-out or rule-in an AML, and as described for one AML-NPM1 mut patient [24]. The identification of even rare but unequivocal Auer rods by morphologic review, although that may require time, confirms the diagnosis of a myeloid neoplasm.
FCI using a panel of markers that are required to assign lineage in acute leukemias correctly, is crucial to diagnose AML, including to definitively exclude MPALs, which were first classified as ALALs by WHO 2008 , as depicted in Figure  3. ALALs also include acute undifferentiated leukemia, which is extremely rare and must only be considered after all other leukemia subtypes (including blastic plasmacytoid dendritic cell neoplasm and leukemias of other unusual lineages) and non-hematopoietic neoplasms are definitively excluded [42]. Immunohistochemical positivity of the leukemic blasts for myeloperoxidase, lysozyme, or CD117 may be useful to identify myeloid lineage. The reader is referred to a recent expert review by Porwit and Bene for recommendations (and pitfalls) for lineage determination in acute leukemias and diagnosis of MPALs by the EGIL and WHO 2001 -WHO 2016/2017 classifications [50]. The myeloid antigens, CD13, CD33, and CD117, included by the EGIL and WHO 2001 but not by WHO 2008 , were myeloid lineage-specific for MPALs by WHO 2016/2017 . Also, while the EGIL classification included CD15 and not CD11c, the WHO 2001 -WHO 2016/2017 included CD11c and not CD15 as myeloid lineage-specific antigens [12,40,42,50]. Importantly, FCI also provides the patientspecific leukemia-associated immunophenotype that serves as a signature for subsequent disease detection, including the expression of CD33 on leukemic cells for anti-CD33 therapy.
Cytogenetics analysis is essential, with molecular analysis for genetic mutations. The presence of the t(8;21), t(15;17), t(16;16) or inv (16) abnormalities is diagnostic of AML even if blasts are less than 20%, and even if a MPAL is suggested J Clin Haematol. 2020 Volume 1, Issue 2 by FCI [12,[40][41][42]50]. MPALs may also show a complex karyotype, which alone does not indicate a diagnosis of AML-MRC [50]. Additionally, MPALs may harbor gene mutations similar to those present in AML [50], further necessitating diagnostic distinction between MPAL and AML. Finally, in communication with the clinician for clinical history and laboratory scientists, the pathologist can best integrate findings from all necessary testing modalities to correctly classify AML, which guides the clinician for risk stratification and therapy. Figure 8 shows the current NCCN guidelines for risk stratification [51,52]. Table 5 shows the CAP-ASH 2017 guidelines and NCCN guidelines for mutation testing in AML [51][52][53].
Moreover, AML treatment now includes drugs targeted against FLT3 and IDH mutations, anti-CD33 antibody therapy, and therapies specific for AML-MRC, therapyrelated AML, and AML in the elderly, as shown in Table 6, with several additional drugs, including in combinations, in clinical trials [54,55].
All of the above advances now increasingly necessitate comprehensive, integrated diagnostics, including genomics for AML diagnosis, prognosis, and therapy. Considerations to implement such testing for AML include (1) first, correct diagnostic classification of AML with combined testing modalities and multidisciplinary teams as described above, with the pathologist best able to integrate all findings, (2) cost of testing, which is likely to be much less than treatment costs [56], and (3) turnaround time, which can take much longer by NGS than single-gene PCR assays, especially for FLT3 and IDH1/IDH2 mutations. Apart from institution-specific assays [57], commercial assays to allow clinical NGS results in possibly two days from sample collection will soon be available for AML [58]. (4) Several commercially available NGS assays include bulk RNA sequencing for detecting gene fusions [58,59]. (5) Clonal hematopoiesis may co-exist with myeloid neoplasms, including AML, with mutant clone detection dependent upon genomic methods, leading to challenges for interpretation of the significance of genetic mutations identified by NGS [37,[60][61][62]. (6) Technically, in addition to sequencing depth and coverage, variability arising from the entire multi-step NGS process, including nucleic acid extraction, target enrichment, and library preparation, bioinformatics analysis, data analysis, and reporting for genomic aberrations must be considered, including for multi-institutional research, given the variability in NGS  This information should include the sample source, adequacy, and collection information, as applicable. Strong recommendation: To avoid duplicate procedures, associated patient discomfort, and additional costs, if after examination of a peripheral blood smear, it is determined that the patient will require immediate referral to another institution with expertise in the management of acute leukemia for treatment, the initial institution should, whenever possible, defer non-emergent invasive procedures, including bone marrow aspiration and biopsies, to the treatment center. Mutational testing for risk stratification: FLT3, NPM1, CEBPA, CKIT, TP53, ASXL1, and RUNX1, with FLT3 ITD , allelic ratio (mutant/normal <0.5 low; >0.5 high), as per 2017 European Leukemia Network guidelines [52] Recommend testing for both FLT3 ITD and FLT3 TKD Abbreviations: AML, acute myeloid leukemia; CBF-AML, core-binding factor AML (AML with t(8;21)(q22;q22.1); RUNX1-RUNX1T1 or inv (16) [53]. 3 Serious limitations in quality of evidence (inadequate [low] or insufficient), balance of benefits and harms, values or costs, but panel consensus was that a statement was necessary [53].
Quality of evidence: Convincing, high confidence that available evidence reflects true effect, and further research is very unlikely to change the confidence in the estimate of effect; Adequate, moderate confidence that available evidence reflects true effect, and further research is likely to have an important effect on the confidence in estimate of effect and may change the estimate; Inadequate, Little confidence that available evidence reflects true effect, and further research is very likely to have an important effect on the confidence in the estimate of effect and is likely to change the estimate [53]   methods (including manual versus automated) used in different laboratories [63][64][65][66][67][68]. A recent study showed genomic sequencing results to be institution-specific, with inter-institution heterogeneity in sequencing methods [69]. (7) Particularly for the AML classification, a technical challenge pertains to the detection of mutations in CEBPA [70], a GC-rich single-exon gene, and FLT3 ITD mutations by short-read sequencers, as per the experience of many laboratories. Since NGS results for both targets are methoddependent, both genes are often preferentially analyzed by separate PCR-based assays. (8) Differences in NGS assays alone, including customized for academic institutions and commercially available (off-the-shelf or customized), add to the variability in sensitivity and specificity that might vary for different gene targets. Necessarily, if mutations in the genes necessary for precise diagnostic classification might be variably detected inter-institutionally, then ultimately, that would affect the collective diagnoses rendered universally, leading to inadequate or inaccurate assessment of the overall "true" AML subtypes. Similar to the standardized nomenclature required by the WHO classification, standardization of genomic testing for clinical diagnosis, albeit challenging, would be a major step forward towards the very challenging task of integrating genomics in AML for routine clinical care worldwide.
In conclusion, due to progressively rapid advances in technology and genomics, the last two decades have shown exponential progress for depth in our understanding of AML pathogenesis. The WHO 2016/2017 classification of AML represents a significant milestone for the care of patients with AML and has paved the way for future continued collaboration and innovation toward further progress for patient care.