Single cell proteogenomic sequencing identifies a relapse‐fated AML subclone carrying FLT3‐ITD with CN‐LOH at chr13q

Abstract Internal tandem duplication of the Feline McDonough Sarcoma (FMS)‐like tyrosine kinase 3 (FLT3‐ITD) is one of the most clinically relevant mutations in acute myeloid leukemia (AML), with a high FLT3‐ITD allelic ratio (AR) (≥0.5) being strongly associated with poor prognosis. FLT3‐ITDs are heterogeneous, varying in size and location, with some patients having multiple FLT3‐ITDs. Bulk cell‐based approaches are limited in their ability to reveal the clonal structure in such cases. Using single‐cell proteogenomic sequencing (ScPGseq), we attempted to identify a relapse‐fated subclone in an AML case with mutations in WT1, NPM1, and FLT3 tyrosine kinase domain and two FLT3‐ITDs (21 bp and 39 bp) (low AR) at presentation, then relapsed only with WT1 and NPM1 mutations and one FLT3‐ITD (high AR). This relapse‐fated subclone at presentation (∼2.1% of sequenced cells) was characterized by the presence of a homozygous 21 bp FLT3‐ITD resulting from copy neutral loss of heterozygosity (CN‐LOH) of chr13q and an aberrant, immature myeloid cell surface signature, contrast to the cell surface phenotype at presentation. In contrast to results from multicolor flow‐cytometry, ScPGseq not only enabled the early detection of rare relapse‐fated subclone showing immature myeloid signature but also highlighted the presence of homozygous 21 bp FLT3‐ITDs in the clone at presentation.


INTRODUCTION
Internal tandem duplication of the FMS-like tyrosine kinase 3 (FLT3-ITD) is one of the most common and clinically relevant mutations in acute myeloid leukemia (AML) [1,2]. FLT3-ITD is found in approximately 25%-30% of AML cases and often co-occurs with NPM1 (nucleophosmin 1) mutations [2][3][4]. Prior publications have commented on the importance of allelic ratio (AR), insertion size, location, and the number of Internal Tandem Duplications (ITDs) as being associated with diverse clinical outcomes [5][6][7][8][9][10][11][12]. Based on these observations, the 2017 European LeukemiaNet (ELN) recommendations commented that NPM1-mutated AML patients can be categorized into either intermediate or favorable risk groups depending on their FLT3-ITD status and the AR [13]. It is recommended that patients with an NPM1 mutation and FLT3-ITD ≥0.5 receive allogeneic hematopoietic cell transplantation, while high dose consolidation chemotherapy is considered sufficient for patients with FLT3-ITD AR <0.5 receiving curative-intent chemotherapy.
As information regarding the presence or absence of FLT3 mutations is required within days of diagnosis for the choice of proper treatment for AML, polymerase chain reaction (PCR) of bulk Deoxyribonucleic Acid (DNA) or Ribonucleic Acid (RNA) is employed in diagnostic laboratories [14]. This approach can determine the size and AR of FLT3-ITDs but does not provide information with regard to what is happening within individual cells. An AR of ≥0.5 means that in a significant proportion of cells, there has likely been a loss of the wild-type FLT3 allele, such that some cells contain only the ITD form of FLT3. This could occur due to loss of heterozygosity (LOH), or reduction to homozygosity, at the FLT3 locus. Such cells have a very high probability of causing relapse in the absence of allogeneic HCT [15][16][17][18]. However, when the AR is <0. 5, there is uncertainty about the nature of the FLT3-ITD carrying leukemic population. As most diagnostic laboratories use DNA from bulk peripheral blood or bone marrow nucleated cells, a spuriously low AR of <0.5 can come about if there is significant contamination of the sample by residual nonleukemic cells. It is also possible to miss cells with only the FLT3-ITD form of FLT3 if the FLT3-ITD occurred late in disease development, as is often the case, and is subclonal at the time of assessment. Finally, the bulk assessment does not inform whether the mutations in cases with several FLT3 isoforms are present in a single clone of cells or come about because of multiple clones.
While bulk methods cannot resolve questions of co-occurrence of mutations in a cell or identify subclones that have lost the wildtype allele FLT3, single-cell-based approaches can overcome these limitations and provide an opportunity to capture subclonal genetic events. Studies utilizing single-cell sequencing have generated clinically and biologically relevant information in AML including the pattern of acquisition of mutations and clonal evolution, as well as deconvolution of bulk AML samples based on surface markers and mutations [19][20][21][22][23][24][25][26][27]. With longitudinal samples, emerging mutation patterns post-FLT3 inhibitor treatment have also been observed [23].
In this report, we describe our investigation of the leukemic cells of a patient with AML using single-cell proteogenomic sequencing (ScPGseq) allowing for simultaneous determination of DNA mutations and cell surface proteins at the single-cell level. Through this approach, we demonstrate that it is possible to accurately characterize multiple FLT3-ITDs at the single-cell level. More importantly, by integrating DNA mutation and cell surface phenotypes, we show that a preexisting relapse-fated subclone could be identified at the time of initial diagnosis.

Single-cell proteogenomic sequencing
Cryopreserved peripheral blood mononuclear cells at initial diagnosis and relapse were obtained from a 46-year-old female who was diagnosed with de novo AML; these were used for quantitative Polymerase Chain Reaction (qPCR) for identification of FLT3-ITD mutations, targeted bulk DNA sequencing and ScPGseq. ScPGseq was performed using the Mission Bio's AML panel and 16 barcoded oligonucleotideconjugated antibodies following the manufacturer's protocols (Table   S1).
Using ScPGseq, we identified the mutation profile and abundance of 16 cell surface markers (Table S2). Detailed procedures on data filtering (

Statistical analysis
All statistical analyses were performed using the R programming language (R Foundation for Statistical Computing) [28]. To compare discrete and continuous variables, Fisher's exact test and Student's t-test or Mann-Whitney U test were used accordingly. Bonferroni correction was used to adjust for multiple comparisons [29].

Description of clinical testings and study subject
Cells were obtained from a 46-year-old female with de novo AML at the time of initial diagnosis and relapse (Table S4). Using bulk RNA and a qPCR-restriction fragment length polymorphism (RFLP) assay, two

Clonal analyses of single cells detect three AML subclones with three distinct FLT3 mutations at diagnosis and one AML clone at relapse
In the analysis of 2367 cells from the diagnostic sample, mutations were detected in WT1 S386*, NPM1 W288Cfs*12, FLT3 D835Y, 21 bp  Figure 1C and Figure S4). The largest fraction accounting for 86.5% (1925/2226 cells) carried WT1 S386*, NPM1 W288Cfs*12, and 21 bp FLT3-ITD ( Figure 1B,C, Figure S2 and S3B). In contrast to C3 (WT1 + /NPM1 + /21 bp FLT3-ITD + cells at initial diagnosis), there was no wild-type FLT3 allele present in these cells (median single-cell VAF [scVAF] = 100%). We refer to this clone as C3R as it most likely arose from C3 by loss of the wild-type FLT3 allele ( Figure 1C and Figure S5). The remaining 301 cells (13.5%, 301/2226 cells) did not carry any mutations. An interesting feature of the C3R cells was that only the 21 bp FLT3-ITD and not the wild-type FLT3 allele was detected ( Figure   S4A), suggesting reduction to homozygosity at the FLT3 locus. As can be seen in Figure 1D, the loss and duplication of chromosome 13q was confirmed using SNP array analysis.
By incorporating clonal architectures at diagnosis and relapse, we inferred clonal evolution from diagnosis to relapse ( Figure 1E). Based on the mutation pattern (Table S5), we inferred that the WT1 mutation occurred first, followed by the acquisition of the NPM1 mutation.
Subsequently, three FLT3 mutations developed as individual events in unique C1 cells, establishing three distinct subclones. Following treatment and relapse, only the subclone with the 21 bp FLT3-ITD of C3 was found, but as noted above C3R had lost the normal FLT3 allele ( Figure   S5A).  It is of note that C3-ITD Hom cells had a similar cell surface phenotype as observed in C3R cells, with high expression levels of CD7, CD33, CD117, and CD123 and decreased levels of CD11b and CD4 ( Figure 2E and Figure S9).

Cell surface phenotype identifies a subclone at diagnosis, which is dominant in subsequent AML relapse
To further confirm the presence of ITD Hom cells, we compared results from ScPGseq with flow cytometry (Figure 3). Based on flow cytometry, we were able to observe two populations of AML cells at diagnosis according to expressions of CD117 and CD7 ( Figure 3A). On

DISCUSSION
The current study utilized ScPGseq to characterize multiple FLT3-

CONFLICT OF INTEREST
The authors declare no competing financial interests.