Somatic Mutations in Oncogenes Are in Chronic Myeloid Leukemia Acquired De Novo via Deregulated Base-Excision Repair and Alternative Non-Homologous End Joining

Somatic mutations are a common molecular mechanism through which chronic myeloid leukemia (CML) cells acquire resistance to tyrosine kinase inhibitors (TKIs) therapy. While most of the mutations in the kinase domain of BCR-ABL1 can be successfully managed, the recurrent somatic mutations in other genes may be therapeutically challenging. Despite the major clinical relevance of mutation-associated resistance in CML, the mechanisms underlying mutation acquisition in TKI-treated leukemic cells are not well understood. This work demonstrated de novo acquisition of mutations on isolated single-cell sorted CML clones growing in the presence of imatinib. The acquisition of mutations was associated with the significantly increased expression of the LIG1 and PARP1 genes involved in the error-prone alternative nonhomologous end-joining pathway, leading to genomic instability, and increased expression of the UNG, FEN and POLD3 genes involved in the base-excision repair (long patch) pathway, allowing point mutagenesis. This work showed in vitro and in vivo that de novo acquisition of resistance-associated mutations in oncogenes is the prevalent method of somatic mutation development in CML under TKIs treatment.


INTRODUCTION
Tyrosine kinase inhibitors (TKIs) are effective for treating chronic myeloid leukemia (CML); however, 10-15% of patients develop resistance to first line imatinib treatment. 1 The 1 st generation TKI imatinib remains the most prescribed drug recommended for frontline therapy of chronic phase (CP) CML. While 2 nd -generation TKI frontline therapy has shown faster cytogenetic and molecular responses, no differences in overall survival were found between 2 nd generation TKIs and imatinib, which may be the preferred choice for older patients with comorbidities and for most adult patients with low-and intermediate risk CML-CP (1,2).
Mutations in the kinase domain (KD) of BCR-ABL1 represent a common molecular mechanism of imatinib therapy resistance and are responsible for 30% of acquired imatinib resistance cases (3,4). Furthermore, CML relapses and progression to advanced phases of the disease are often associated with recurrent mutations in oncogenes, which are frequently mutated in other hematological malignancies (5,6). Despite the major clinical relevance of mutation-associated TKI resistance in CML, the mechanisms underlying mutation acquisition in oncogenes of drug-resistant CML cells are not yet well understood.
It was postulated, that BCR-ABL1 KD mutations could be detected in newly diagnosed patients in accelerated or blast phases (7). This led to the general assumption that a preexisting mutated subpopulation of CML cells carrying a resistance phenotype is selected and expanded during TKI treatment. However, some studies observed the mechanism of de novo BCR-ABL1 mutation acquisition during imatinib treatment in CML cells. It was suggested that BCR-ABL1 expression, but not kinase activity, is required for the acquisition of KD mutations (8). The NAD +dependent histone deacetylase SIRT1 was shown to promote acquisition of BCR-ABL1 KD mutations in CML cells (9). Others proposed the activity of the alternative nonhomologous end-joining (alt-NHEJ) pathway in resistant CML cells driven by BCR-ABL1 and/or MYC transcriptional activity (10,11).
This study was based on the hypothesis that the appearance of somatic mutations during imatinib treatment is not just a passive process of selection and expansion of preexisting mutated clones; rather, it is based on de novo mutation acquisition involving deregulation of the DNA damage response and DNA repair pathways.

MATERIALS AND METHODS
The detailed methods are provided as Supplementary Methods.

Isolation of CML Clones
KCL-22 and CML-T1 clones were prepared by single-cell FACS sorting (BD FACS Aria III; BD Biosciences, San Jose, CA, USA) according to CD38 expression ([HB-7] anti-CD38 APC; Sony Biotechnology, San Jose, CA, USA) in 96-well plates containing optimal growing medium either imatinib-free (sensitive; S clones) or with a 0.004 µM imatinib that was subsequently increased up to 4 µM (resistant; R clones) (Supplementary Figure S1).

Patients
CML patients (n=32) were diagnosed and treated at the Institute of Hematology and Blood Transfusion in Prague (UHKT), Czech Republic (Tables 1, 2). All samples were collected after written informed consent was obtained according to the principles of the Declaration of Helsinki and approval by the UHKT Ethics Committee.

In Vivo Models
Experiments were performed on NOD.Cg-Prkdc scid Il2rg tm1Wjl / SzJ (NSG)-immunodeficient female mice (n=16) aged 8-10 weeks. Mice were subcutaneously inoculated with 1x 10 6 KCL-22R imatinib-resistant clones. Treated mice were orally administered imatinib (100 mg/kg/mouse) once per day, with the initiation of treatment 3 days after inoculation. The design of all experiments was approved by the institutional Animal Care and the Use Committee.

Droplet Digital PCR
Mutations were analyzed on DNA level using commercially available allele-specific assays for droplet digital PCR (ASO-ddPCR) (QX200 Droplet Digital PCR system and Auto Droplet generator, Bio-Rad, Hercules, CA, USA).

Gene Expression PCR Arrays
Human DNA Damage Signaling Pathway RT 2 Profiler ™ PCR Array (PAHS-029Z; Qiagen, Valencia, CA, USA) and Human DNA Repair RT 2 Profiler ™ PCR Array (PAHS-042Z; Qiagen) were performed on the StepOnePlus system (Thermo Fisher Scientific, Waltham, MA, USA). The complete list of 128 measured genes is provided (Supplementary Table S1). The differentially expressed genes were determined following stringent criteria: 1) the mean 2 DC value for R clones was ≥ 2.5 or ≤ 0.4-fold of the mean 2-(Cts-Ctc) value for S clones and 2) the difference of 2 DC values for R clones and S clones reached the level of significance p=0.05.
Sequence analysis and mutation identification were performed using NextGENe software (SoftGenetics, State College, PA, USA).

Mass Cytometry
The protein expression in leukemic cells and clones was determined by staining with CyTOF metal-conjugated antibodies against proteins of interest (Supplementary Table S3) and analyzed by a CyTOF2 mass cytometer (Fluidigm, South San Francisco, CA, USA) as described in detail previously (12).

Statistical Analysis
The statistical analyses were performed using Student´s t-test in MS Excel (Microsoft Corporation, Redmond, WA, USA). Fish plots were generated in the program Fishplot package for R (13).

TKI-Resistant Mutation Acquisition in CML Patients
From 253 CML patients diagnosed within the years 2010-2018 and treated with TKIs as frontline therapy 87 responded to therapy as warning or failure according to the ELN recommendation. 1 This group of non-optimally responded patients were regularly evaluated for the presence of BCR-ABL1 KD mutations. BCR-ABL1 mutations were detected in 24/87 patients (28%) at a median time of 12.2 months since TKI start (range 3.2-117.4; Table 1A). Presence of BCR-ABL1 mutations was retrospectively analyzed by NGS at the time of diagnosis in 21/24 patients with available RNA samples. Mutations were not detected in any of the 21 patients at the time of diagnosis with the sensitivity range of 1-3.7% (Supplementary Methods). Thus, it is highly probable, that these mutations were acquired de novo under TKI treatment. Table S2) was used to reveal co-occurrence of somatic mutations in oncogenes in the samples of non-optimal responders with BCR-ABL1 mutations. The oncogenic mutations were detected in 8/24 patients (33%) ( Table 1B). Presence of oncogenic mutations was retrospectively analyzed by NGS myeloid panel at the time of diagnosis in all 24 patients. The analysis revealed mutations in 5/24 patients at the time of diagnosis, while 3 mutations were likely acquired de novo under TKI treatment.

Resistance-Conferring Mutations Are Acquired De Novo After Exposure to Imatinib
The hypothesis of de novo mutation acquisition in CML cells resistant to TKI was tested on FACS sorted single cells of KCL-22 exposed to imatinib. To address whether the putative BCR-ABL1 mutation acquisition is related to CD38 (non)expression as suggested previously (14), cells were sorted according to the expression of CD38. Out of the 120 sorted single cells, 33 growing clones emerged. No BCR-ABL1 KD mutations were detected by NGS in KCL-22 clones in day 21 post-sorting. Finally, 6 KCL-22R clones resistant to 4 µM imatinib were established from both CD38-(n=4) and CD38+ (n=2) cells of origin (Supplementary figure S1A).
To support data suggesting de novo acquisition of mutations conferring imatinib resistance, cells of the original isolated KCL-22 Clones 1-4 stored at day 20 after the sort to 0.004µM IM were used to repeat experiments of resistance development in biological triplicates. Interestingly, the newly isolated resistant clones (KCL-22R´) developed a different spectrum of BCR-ABL1 and KRAS mutations (Figure 2).
The de novo acquisition of mutations was also confirmed in the experiment with CML-T1 cell line by similarly designed experiment  To decipher the putative pathway involved in mutation acquisition, the expression of genes associated with mismatch repair (RPM1, RFC1, POLD3, PCNA, LIG1, EXO1, MLH1, MSH6, MLH3, MSH3 and PMS2) was determined in KCL-22R clones and compared to that in KCL-22S clones. The increased expression of MLH1 and genes with more general functions in DNA repair (e.g., PCNA, POLD3, LIG1) was found, while the expression of genes involved in the direct recognition of mismatched DNA (MSH6, MLH3, MSH3, PMS2) was not significantly altered, making mismatch repair a questionable candidate to participate in mutation acquisition in KCL-22R clones (Supplementary Figure S5A). Next, the expression of genes participating in BER was determined. The expression of genes involved in short-patch BER was not altered, while significant upregulation of genes participating in long-patch BER was found ( Figure 3A). In addition to its roles in BER and mismatch repair, LIG1 is also associated with the error-prone alt-NHEJ pathway of DNA double-strand break (DSB) repair. To investigate the possible role of DNA DSB repair pathways, the expression of DNA ligases and their functional partners was determined. The expression of LIG1 and PARP1 was found to be significantly increased, while the expression of LIG3, LIG4 and their respective partners was not altered ( Figure 3B). Increased expression of BER and alt-NHEJ genes was also confirmed at the time of mutation acquisition in the imatinib-resistant KCL-22R´and CML-T1R clones (Supplementary Figure S5B, C). Moreover, regulatory regions of the LIG1 and PARP1 genes were found to be epigenetically active during and after mutation acquisition and, in the case of LIG1, occupied by the MYC transcription factor, while the DNA methylation was found in KCL-   Table S4).
The expression of DNA ligases and their functional partners was examined in CD34+ primary cells of CML-CP patients ( Table 2). The expression of LIG1 and PARP1, representing alt-NHEJ pathway, was found to be relatively augmented to that of LIG4 and XRCC6, representing classical NHEJ (c-NHEJ) pathway, at the time of diagnosis and resistance to imatinib (therapy failure) in comparison with prevalently non-leukemic CD34+ cells at the time of response to TKIs ( Figure 4A). Gene expression was also examined in the total leukocytes from peripheral blood (PB) of CML patients at diagnosis (n=3), who on TKI therapy subsequently acquired mutations, compared with the pooled samples of PB leukocytes of healthy donors (n=5) ( Figure 4B). No difference in the expression of DNA ligases was found, while the increased expression of FEN1, PCNA and, to a lesser degree, UNG was confirmed in CML-CP. Suggested molecular mechanisms were studied during mutation acquisition in CML patients resistant to TKIs. The expression of LIG1, FEN1, UNG and, to a lesser degree, POLD3 was augmented during mutation acquisition ( Figure 4B).

De Novo TKI-Resistant Mutation Acquisition and the Evolution of Resistant CML Clones Depends on the Culture Conditions, Time, and Dose of Imatinib
The processes of natural mutation appearance and mutated cell expansion was studied in detail in KCL-22 cells. Invariably, no BCR-ABL1 KD mutations were detected in imatinib-naïve cells by NGS. However, BCR-ABL1 mutations were repeatedly detected by NGS after the exposure of cells to 0.4 µM imatinib. The competition of mutated clones was driven not solely by the type of mutations but also by their frequency and by the environmental conditions  Figure S11).
To address a clonal evolution upon the constant imatinib dose, equal numbers of cells from the 4 KCL-22R clones growing in 4 µM imatinib were mixed, split, and cultured in 4 µM, 10 µM imatinib and imatinib-free medium. The T315I-bearing clone outgrew other 3 clones in both 4 µM and 10 µM imatinib ( Figures 5A, B). The superior proliferation of the T315I-bearing clone in 4 µM imatinib was also confirmed using a mixture of cell tracking dyes (Supplementary Figure S12A) and by cell cycle analysis (Supplementary Figures S12B, C). In the mixture of clones growing without imatinib, the Y253H-bearing clone showed a proliferative advantage over the other 3 clones ( Figure 5C). Clonal evolution was followed for the artificially mixed 3 CML-T1R clones, where Y253H-mutated cells outgrown in all conditions (Supplementary Figure S13).
The signaling properties of leukemic cells and resistant clones were studied at the protein level using a CyTOF custom-designed panel. BCR-ABL1-dependent pathways via STAT5, ERK1/2 and AKT were inhibited after the imatinib exposure and remained diminished in mutated resistant clones with slightly distinct pathway preferences dependent on the type of mutation and/or imatinib dose. Imatinib-resistant clones KCL-22R displayed persistently high expression of MYC, while overexpression of BCL-2 was additionally found in CML-T1R clones (Supplementary Results; Supplementary Figures S14-S17).
All 4 KCL-22R clones were able to engraft in the mice without any apparent difference between the imatinib-treated and untreated groups. Individual clones exhibited different engraftment abilities ( Figure 6A). T315I-mutated Clone 1 exhibited the fastest growth in the imatinib-treated cohort, similar to in vitro conditions ( Figure 6B). Similar data were observed in the imatinib-untreated cohort.

DISCUSSION
Both the selection of preexisting BCR-ABL1 mutated clones (15) and de novo acquisition during TKIs therapy (8,9) were proposed to explain the emergence of TKI-resistant mutations.  This study worked with the hypothesis of de novo acquisition of mutations conferring resistance to imatinib as the prevalent mechanism in CML cells.
The established in vitro model of CML KCL-22 cells has the ability to relapse after exposure to imatinib upon acquiring the T315I mutation (8). Despite differences in the procedure of establishing imatinib-resistant clones using single-cell FACS sorting, de novo acquisition of BCR-ABL1 KD and other somatic mutations was confirmed in this work by sensitive NGS and ASO-ddPCR. The same clonal population identically exposed to imatinib in parallel experiments was able to either develop different mutations or fail to acquire any mutation conferring resistance, thus proving the de novo acquisition process and supporting the hypothesis that the existence of the specific pool of pre-mutated or mutationally destined cells is unlikely for KCL-22 cells. De novo acquisition of BCR-ABL1 mutations was also confirmed in the CML-T1 model.
Recently, it was established that the genetic landscape of CML is significantly more diverse than assumed in the past, particularly in recurrent disease and advanced phases (6). Indeed, several mutated genes in addition to BCR-ABL1 were identified by the DNA NGS myeloid panel in resistant KCL-22R clones, namely, BCOR, ATRX, RUNX1 and KRAS. All these genes were previously found to be mutated in CML, although their clinical importance for CML patients remains unclear (5,16,17). RUNX1 was found to be commonly mutated during both CML-CP diagnosis and blast phase, often in cooccurrence with other recurrent mutations (17,18). RUNX1 mutations are associated with the block of cell differentiation and with aggressive AML (19,20). Notably, KCL-22 carries isochromosome 21 [i(21)(q10)], possessing 3 alleles of RUNX1, and mutations in RUNX1 were found to be associated with Chr21 trisomy (21). In the case of KRAS, de novo mutation acquisition was proven in the clone bearing mutations in ATRX and RUNX1. While KRAS mutations in CML seem to be particularly rare events (22), it should be noted that the pattern of acquisition of a KRAS mutation following RUNX1 genetic lesions was described in AML (23). The association of the BCOR and ATRX mutations with prognosis remains contradictory in CML (5,24). In contrast to KCL-22R clones, no somatic mutations beside BCR-ABL1 KD were identified in CML-T1R clones. However, it must be mentioned that some of the genes frequently found to be mutated in the lymphoid blast phase (e.g., IKZF1) (17,25) were omitted in the NGS myeloid gene panel.
The exact molecular mechanism of mutagenesis of BCR-ABL1 KD and other oncogenes remains unclear. It was proposed that BCR-ABL1-induced reactive oxygen species (ROS) are responsible for genomic instability and BCR-ABL1 mutation acquisition in CML cells (26)(27)(28). While ROS may definitely increase the genomic instability of CML cells and contribute to deregulation of DNA repair (29), the proposed pathways of ROS induction via STAT5 and PI3K/Akt were found to be inhibited in KCL-22 and CML-T1 cells treated with imatinib and remained diminished even after mutation acquisition, making their contribution to de novo mutagenesis questionable.
Therefore, the DNA damage response and DNA repair genes with notably changed expression at the time of mutation acquisition were identified and subsequently annotated to the functions in the mismatch repair and BER pathways. Only a few studies are examining the role of mismatch repair in CML. Rather, an inhibitory effect of BCR-ABL1 on the expression of mismatch repair genes was found in CML cells (30). The data from expression arrays in this work revealed that the expression of most of the mismatch repair-specific genes was unchanged during mutagenesis. By contrast, the expression of genes involved in the long-patch variant of the BER pathway was significantly upregulated during mutation acquisition. FEN1 has been observed overexpressed in a broad spectrum of cancers and was previously associated with PCNA-dependent genetic instability and DNA damage (31,32). The increased expression of POLD3 was also observed during mutation acquisition. The POLD3 subunit of DNA polymerase d was found to induce mutagenesis by promoting translesion synthesis. POLD3 is required for inducing C-to-T mutagenesis in an abasic site (33). Notably, C-to-T base exchange is behind the T315I BCR-ABL1 mutation, which is frequently acquired in KCL-22R clones and KCL-22 cells and is one of the most common mutations in advanced CML.
In addition to their role in BER (34,35), UNG and LIG1 are also associated with the error-prone alt-NHEJ repair pathway of DSB (36). Alt-NHEJ was originally depicted as a backup pathway in cells with defective c-NHEJ (37), while some recent studies suggest that inactivation of c-NHEJ is not required for alt-NHEJ activity (38). The group of Chen et al. (8,9) proposed that KCL-22 cells utilize NAD+ -dependent c-NHEJ to induce point mutations (14). Overexpression of CD38 NAD+ hydrolase leads to abrogation of BCR-ABL1 mutation acquisition (14). The data of this work do not provide strong support for c-NHEJ involvement in de novo mutation acquisition. First, in contrast to Wang et al., CD38+ resistant KCL-22R clones with BCR-ABL1 KD mutations were also isolated. This discrepancy may be caused by the different origins of CD38+ KCL-22 cells tested in both studies, which include naturally occurring CD38+ cell subpopulation vs. cells with artificial CD38 overexpression, which may lose or diminish the mutagenesis ability. Second, the expression of LIG4, XRCC4 and XRCC6 genes was not changed in the KCL-22 (and CML-T1) model, and DNA methylation associated with repressive chromatin marks was found in the regulatory regions of LIG4. However, it should be noted, that upregulation of LIG4 expression, along with upregulation of alt-NHEJ genes, was observed during mutation acquisition in PB leukocytes of CML patients. Notwithstanding, data in general are in agreement with the studies proposing that the alt-NHEJ pathway is activated in TKIresistant CML cells (10,39). Increased expression of LIG1, UNG and PARP1 was found in cells undergoing mutagenesis. The expression of alt-NHEJ factors in CML was previously found to be augmented by the oncogenic transcription factor MYC (11). Accordingly, MYC is strongly expressed in both KCL-22R and CML-T1R clones and CML-resistant primary cells, as described previously (40), and its binding was enriched in the LIG1 transcription start site (TSS) at the time of mutation acquisition.
The growth abilities of individual KCL-22R and CML-T1R clones were assessed in different environmental conditions in vitro and in vivo with the conclusion that clonal evolution is driven by both mutation-type specific oncogenic properties and additional genetic/cytogenetic lesions in concert with environmental selection pressures. At high imatinib concentrations, T315I-bearing Clone 1 showed superior growth over other KCL-22R clones, in agreement with the superior resistance associated with the T315I mutation (41). The growing abilities of individual KCL-22R clones only partially reflect kinase activity revealed by CyTOF, a fact already described by others (42,43). Conversely, the finding that the activity of the main BCR-ABL1-associated signaling pathways in KCL-22R clones is generally lower than that in imatinib-naïve KCL-22 cells is in accordance with the fact that BCR-ABL1 mutations do not confer a growth advantage in the absence of imatinib (44).
Taken together, this study provides evidence for the responsibility of de novo mutation acquisition for acquired imatinib resistance in CML-CP. Imatinib has been the most used frontline TKI to combat CML-CP, and it likely will remain in this position in the future owing to the benefits of low toxicity and cost-effectiveness even after emerging therapy goal of treatment-free remission has been considered (45,46). For CML-CP patients, no robust evidence has conclusively shown that BCR-ABL1 KD mutations conferring resistance to imatinib may already be detectable at diagnosis, which was also confirmed in this work (47,48). Somatic mutations acquired de novo in the follow-up constitute a bulwark of lesions associated with refractory and progressed CML with poor outcomes (5).
The data support the considered role of the alt-NHEJ repair pathway (11) and provide the first hint that BER, particularly the long-patch variant, may be involved in the acquisition of somatic point mutations in oncogenes in CML. Contrary to some other natural in vitro models of BCR-ABL1 mutagenesis (49), BCR-ABL1 KD mutations acquired in resistant KCL-22R and CML-T1R clones belong among the mutations most commonly detected in both CML-CP and advanced CML (50). Moreover, the same mechanism of acquisition may be undergoing in the case of somatic mutations in other cancer genes, thus underlying the clinical relevance of the presented findings.
In conclusion, somatic mutations associated with resistance to imatinib are acquired in CML-CP de novo, thus early mutation detection is highly required to prevent progression because several therapeutic options exist for cases with mutated KD of BCR-ABL1. However, even when mutations in other cancer genes can be uncovered early by currently feasible NGS approaches, the resistance associated with other somatic mutations or in combination with mutated BCR-ABL1 remains therapeutically challenging, which will require highly individualized treatment management in the future.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: BioProject PRJNA750307.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by UHKT Ethics Committee. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Animal Care and the Use Committee, First Faculty of Medicine, Charles University, Prague.

AUTHOR CONTRIBUTIONS
NC wrote the manuscript, performed experiments, evaluated data, and contributed to the interpretation of the results. VP performed the majority of the experiments, evaluated data, and contributed to the interpretation of the results. PB performed experiments, analyzed data, and revised the paper. JK performed NGS analysis and evaluated data. AL performed in vivo experiments. TK and DKuz designed and interpreted mass cytometry data. JB and SR performed cytogenetic analysis and interpreted data. DKar performed expression arrays. KR prepared samples for mass cytometry analysis and interpreted data. VK and MB performed cell tracking and cell cycle analysis. DKun evaluated clonal development. JL performed resistant clone establishment. HK and CS provided patient samples and clinical data. PK supervised the in vivo experiments. OH supervised mass cytometry and flow cytometry analysis, interpreted data and critically revised the manuscript. KMP designed experiments, coordinated the research, interpreted the results, and revised the paper. All authors contributed to the article and approved the submitted version.