Significance of Primary Immunodeficiency Disorder (PID)–associated Germline Mutations in Korean patients with ALL

Background In addition to somatic mutations, germline genetic predisposition hematologic malignancies is an emerging research issue. In this study, we investigated genetic alterations (both somatic and germline) in Korean acute lymphoblastic leukemia/lymphoma (ALL) patients using targeted gene panel sequencing. Method We selected ALL patients diagnosed at Samsung Medical Center. Eighty-two well known genes associated with 23 predisposition syndromes were included in the gene panel. In addition to sequence variants, gene-level copy number variants (CNVs) were investigated. Results Ninety-three ALL patients (65 children, 28 adults) were enrolled in this study. We identified 197 somatic sequence variants (excluding synonymous variants) and 223 somatic CNVs. Overall survival (OS) and relapse-free survival (RFS) differed significantly by cytogenetic group in childhood B-ALL. The IKZF1 alteration had an adverse effect on OS and RFS in childhood ALL. Among the 82 genes associated with 23 predisposition syndromes, known mutations (TP53, LIG4) and novel mutations (TINF2, CTC1) were identified. Nine patients (9.7%) had pathogenic or likely pathogenic variants associated with primary immunodeficiency disorder (PID) in a heterozygous state. Conclusions We found recurrent mutations and clinical effects similar to those reported in previous studies. We also found an unexpectedly high prevalence of PID-associated germline mutations. Our results suggest several important points: 1) paired leukemic and normal control samples are recommended to discriminate somatic and germline mutations; 2) gene panels must include genes associated with predisposition; 3) gene panels must include genes associated with PID. hybridization; HGMD: Human Gene Mutation Database; iAMP21: intrachromosomal amplification of chromosome 21; M: male; NGS: next generation sequencing; NOS: Not Otherwise Specified; OS: overall survival; RFS: relapse-free survival; T-ALL: T-cell precursor acute lymphoblastic leukemia/lymphoma; WHO: World Health Organization.

ALL) are two of the most common malignancies in children. ALL can be classified by genetic alterations, which are various and heterogeneous. Chromosome aneuploidy, structural alterations and rearrangements, copy number variations (CNVs), and sequence mutations all contribute to leukemogenesis. In 2016, the fourth edition of the World Health Organization (WHO) classification of lymphoid and myeloid neoplasms and acute leukemia included new provision entities of ALL: BCR-ABL1-like (or Ph-like) ALL, iAMP21 (intrachromosomal amplification of chromosome 21), and early T-cell precursor ALL (ETP-ALL) [1].
BCR-ABL1-like ALL is a high-risk form of ALL with peak incidence in young adults.  [2]. iAMP21 accounts for about 2% of ALL in older children and is associated with adverse outcomes [3]. In iAMP21, three or more additional copies of RUNX1 (AML1) are observed on chromosome 21 in metaphase fluorescence in situ hybridization (FISH). ETP-ALL is defined as CD1a − , CD8 − , CD5 − (dim), and positive for one or more stem-cell or myeloid antigens. Genetic alterations in ETP are somewhat different than in non-ETP, with FLT3, DNMT3A, and WT1 mutations found more frequently in ETP than in non-ETP [4,5].
In addition to somatic mutations, germline genetic predisposition to hematologic malignancies is an emerging issue in current research. Genes found to be associated with predisposition to myeloid malignancy have been included in the WHO classification, "Myeloid neoplasms with germ line predisposition" [1]. CEBPA, DDX41, RUNX1, ANKRD26, ETV6, GATA2, et al. are included in that category. Several syndromes, such as bone marrow failure syndrome and telomere biology disorders, also are included in that category. One early study estimated that childhood leukemia with hereditary genetic causes accounted for 2.6% of all cancers [6]. Down syndrome (DS) is the most common underlying genetic predisposition for ALL [6,7]. Various syndromes, such as Li Fraumeni (TP53), Bloom syndrome (BLM), Wiskott Aldrich syndrome (WAS), ataxia telangiectasia (ATM), and Nijmegen breakage syndrome (NBN), also increase susceptibility to ALL [8]. A germline mutation of PAX5 is highly associated with B-ALL [9]. Some genes associated with ALL are found to have both germline and somatic mutations.
For example, PAX5, ETV6, TP53, and IKZF1 are known to have important somatic alterations, and germline mutations of those genes also cause susceptibility to ALL.
Furthermore, somatic and germline mutations of those genes can be found at the same time in leukemic samples [9,10]. Therefore at initial ALL diagnosis, discrimination of somatic and germline mutations are crucial to accurately classify ALL patient genetic subtypes/risk groups and detect predisposition genes.
In this study, we used extensive gene panel sequencing to investigate genetic alterations (both somatic and germline) in Korean ALL patients. We also evaluated the clinical significance of recurrent somatic mutations and germline predisposition mutations in Korean ALL patients.

Study population and samples
We selected paired initial-diagnosis and complete remission (CR) bone marrow samples To detect germline mutations, bone marrow slides obtained when the patients were in CR were used. Morphology, chromosome, FISH, and immunophenotyping results were reviewed, and bone marrow slides with no apparent residual leukemic cells were selected as control samples.

Conventional study
A chromosome study was performed using a standard method, and the karyotypes are described according to the International System for Human Cytogenetic Nomenclature.
Multiplex reverse transcription polymerase chain reaction (RT-PCR) was performed to detect recurrent translocation (HemaVision kit, DNA Technology, Aarhus, Denmark). FLT-ITD (internal tandem duplication) mutation analyses (by fragment length polymorphism) and FISH for CDKN2A/B were performed.

Gene panel
From a literature review, we selected 500 genes found to be significantly mutated in ALL (Additional file 1: Table S1). Our gene panel included [11]: cell cycle and p53 signaling

Data analysis
Data analysis was conducted using previously described methods (Additional file 1: Fig.   S1) [12]. After sequencing, we aligned the reads to human genomic reference sequences (GRCh37) using the Burrows-Wheeler alignment tool. The Genome Analysis Tool Kit (Broad Institute) was used for variant calling. Pindel was used to crosscheck insertion and deletion mutations. All mutations were annotated using ANNOVAR and VEP software.
Variants were further examined by visual inspection using the Integrative Genomic Viewer.  [13]. An R package, ExomeDepth (1.1.10), was used to detect exon-and gene-level CNVs in target regions, followed by visualization using a base-level read depth normalization algorithm implemented in DxSeq Analyzer (Dxome, Seoul, Korea) [12].

Patients
Characteristics of the 93 enrolled Korean ALL patients are summarized in Table 1. We enrolled 65 pediatric (<20 years old) ALL patients and 28 adult ALL patients, including 12 T-ALL patients (7 children, 5 adults).
According to the immunophenotyping results, two adults with T-ALL were diagnosed with early T-cell precursor acute leukemia.
B-ALL was classified into three risk groups following a previous study [14]: (1) good risk

Pathogenic or likely pathogenic variants
Four variants were identified in four patients ( Table 2). Two variants were previously known, and two variants were novel. The TP53 rs28934575 variant was identified in a B-ALL, NOS patient. The rs28934575 variant was previously reported in Li-Fraumeni syndrome patients (multiple cancers, including breast cancer, liver cancer, and lung cancer) [15]. The LIG4 rs772226399 variant was identified in a patient with B-ALL with hyperdiploidy that had previously been reported in patients with clinical feature of dyskeratosis congenita (DC) [16]. The TINF2 c.1130-1G>C variant is novel and associated with DC and Revesz syndrome. It is found in about 10% of all DC cases in an autosomal dominant (AD) pattern [17]. The mutation hot spot of TINF2 is exon 6 [17], but the TINF2 c.1130-1G>C variant is located in an intron between exon 7 and exon 8. The CTC1 c.2249dup variant was identified in a patient with B-ALL with hyperdiploidy. CTCI is a causative gene for DC in an autosomal recessive (AR) pattern [18].

Pathogenic or likely pathogenic variants
We identified 30 variants in this study (Table 3 and Additional file 1: Table S2), of which four were novel and 26 had been previously reported. Most variants have AR pattern or AD/AR pattern inheritance. The PTCH2, THBD, PDGFRB, and CSF1R genes are associated with AD pattern disorders.

Germline copy number variants
Only one patient (male, 18 years old, B-ALL-NOS) had a known CNV, CASP10 (deletion of exon 6-exon 9, Additional file 1: Fig. S2). This same CNV was previously found in a patient with systemic juvenile idiopathic arthritis [19]. CASP10 is a causative gene for autoimmune lymphoproliferative syndrome (ALPS) type IIa, and its mutation hot spot is the protease domain (CASc) with an almost missense mutation.
The RAS pathway and chromatin structure modifiers and epigenetic regulators were the most common variants in ALL (Additional file 1: Table S4).
Sixty variants were previously known (registered in the COSMIC database), and 33 variants were in genes associated with the 23 predisposition syndromes.
One ETP patient (ALL0043) had a WT1 mutation (c.822_823insTCCCT), and FLT3-ITD was identified in it by fragment length polymorphism. The other ETP patient (ALL0047) had several sequence variants, including NOTCH1 and IL7R (Additional file 1: Table S3).

B-ALL
NRAS, FLT3, SETD2, and KRAS were the most common variants overall in B-ALL (Additional

Somatic copy number variants
We identified 223 somatic CNVs in this study (Additional file 1: Table S5). Lymphoid development and differentiation and the cell cycle and p53 signaling pathway were the most common CNVs overall.

CDKN2A/B by NGS and FISH
FISH for CDKN2A/B was performed at the initial diagnosis of ALL. We compared the CDKN2A/B results between the FISH and NGS CNV analysis. The overall agreement rate for CDKN2A/B was 83.7% (Table 4). Nine cases were positive for CDKN2A/B deletion by NGS but negative by FISH. Six cases were normal by NGS analysis, but deletion/duplication was confirmed by FISH.

T-ALL
CDKN2A/B was the most common CNV in T-ALL (Additional file 1: Figure S5). Five of seven childhood T-ALL patients had a CDKN2A/B CNV. One ETP patient (ALL0047) had a CDKN2A/B deletion.

Overall survival and relapse-free survival
Overall survival (OS) and relapse-free survival (RFS) are shown by cytogenetic groups in Discussion NGS technology has been applied to various hematologic diseases. Many gene panels and several methods have been used to detect not only sequence variants but also large gene deletions and duplications or fusion genes [12,20,21]. In this study, we found a significant agreement rate between NGS and FISH for CDKN2A/B CNV detection. We also found significant IKZF1 deletions using an NGS CNV analysis. A presumed diagnosis of BCR-ABL1-like ALL can be enabled by using an NGS CNV analysis to test for a high prevalence of genetic alterations in IKZF1 and JAK2 because IKZF1 alterations (~ 80%) and JAK2 alterations (~25%) are more prevalent in BCR-ABL1-like ALL than in other B-ALL subtypes [2]. Although we did not perform gene express profiling and FISH or RT-PCR for alterations commonly found in BCR-ABL1-like ALL, we found seven cases with IKZF1 alterations in non-ALL with BCR-ABL, and they had adverse clinical effects. We assume that the seven patients with non-ALL with BCR-ABL1 and an IKZF1 alteration probably had BCR-ABL1-like ALL.  [11,22], and we found a similar distribution of genetic changes in this study. In the adult high-risk cytogenetic group, IKZF1 deletion was the most common alteration (6/12, 50%).
This high prevalence might have occurred because most cases (11/12) were ALL with BCR-

ABL1.
In T-ALL, we found recurrent somatic sequence variants and CNVs similar to those reported in previous studies [4,5]. Sequence variants of NOTCH1, IL7R , and FBXW7 and CDKN2A/B deletion are common mutations in T-ALL [5,23]. Although only two cases of ETP were enrolled in this study, one of those patients had an FLT3-ITD and WT1 mutation [4].
More ETP cases are needed to reveal genetic alterations with T-ALL among Koreans.
Detection of gene mutations that predispose individuals to cancer is important for several reasons, including modification of cancer management, early detection of secondary malignancies or non-malignant complications, and genetic counseling. The characteristic clinical and physical manifestations of multiple cancers and familial cancer histories can be suggestive of predisposing gene mutations. Some researchers have published a guideline for referral indications for cancer predisposition assessment [8]. However, cancer predisposition assessments can be delayed or ignored for several reasons.
Heterogeneous clinical and physical manifestations can be missed by physicians.
Sometimes, patients have no familial history because they have a de novo mutation [24]. Therefore, the germline mutation detection rate from guideline-based referrals for cancer predisposition assessment (using clinical, physical, and familial history criteria) is lower than from universal, targeted panel sequencing of cancer genes using tumor-normal matched samples [25]. In this study, we analyzed unselected sporadic ALL cases to investigate the overall germline mutation rate in Korea.
In this study, we used CR-state bone marrow slides for germline mutation detection. Skin fibroblasts are the only recommended control sample for germline mutations. Peripheral blood (PB) and bone marrow can be contaminated with leukemic cells. Other samples, such as saliva or buccal swab, can be also contaminated with PB. Furthermore, Agerelated clonal hematopoiesis can be observed in ~10% of the healthy population, which increases with age [26]. However, a skin biopsy is an invasive procedure, so such samples are not easily available. Therefore, we used CR-state bone marrow slides (acquired to test for residual leukemic cells) as the control for germline mutations. No apparent leukemic samples were obtained in our review of bone marrow morphology or the FISH, chromosome, flow-cytometry, and RT-PCR results. The variant allele fraction (VAF) and public population databases were used as filtering tools [27]. Germline variants can have a VAF of >33% even in tumor samples [27]. Variants within that range have a high possibility of germline origin. Variants registered in public databases, such as the Single Nucleotide Polymorphism database (dbSNP), the 1000 Genomes Project, the Exome Aggregation Consortium (ExAC) database, and the Human Gene Mutation Database (HGMD), are probably also of germline origin. We double checked the germline variants both in CR and leukemic samples. True germline variants (identified in CR samples) were also identified in paired leukemic samples with similar VAF. On the contrary, true somatic variants (identified in leukemic samples) were not found or were found with very low VAF in paired CR samples.
At initial diagnosis, only leukemic samples without a normal control might be available to analyze genetic alterations. In that case, discrimination of somatic and germline genetic alterations is difficult. Some genes could have both somatic and germline alterations. For example, PAX5 is a predisposition gene for ALL, and a PAX5 alteration can also be found in B-ALL as a somatic alteration. Moreover, if the VAF of a PAX5 alteration is around 50%, it is difficult to determine whether its origin is somatic or germline. Many clinically important genes in ALL, such as IKZF1, ETV6, and RUNX1, can have both somatic and germline alterations.
To overcome the difficulty of tumor-only analysis, some researchers have sought computational or automatic algorithms to discriminate somatic and germline alterations using VAF and population databases. The accuracy of those methods depends on tumor purity, CNVs, and population allele frequency cut-off values. Sensitivity and specificity of more than 90% were reported [28,29], but no consensus guidelines or cut-off values for tumor-only filtering strategies have been established. Therefore, further studies are still needed to accurately discriminate somatic and germline mutations in tumor-only samples.
Various syndromes increase the risk of ALL, with variable penetrance and preference. DS is the most common genetic cause of childhood leukemia. In an analysis of the National Registry of childhood tumors in the United Kingdom, 131 of 142 leukemia patients with underlying genetic causes were DS patients [6], and an analysis of approximately 18,000 European childhood ALL cases found that 2.4% of ALL patients also had DS [7]. Other genetic diseases with connections to ALL are ataxia telangiectasia, Nijmegen breakage syndrome, neurofibromatosis type 1, familial ALL, and Noonan syndrome [7]. Germline PAX5 and ETV6 mutations carry a high risk (high penetrance) of cancer, mainly ALL [7,9,30]. Bloom syndrome and constitutional mismatch repair deficiency syndrome also carry a moderate risk of ALL. Constitutional mismatch repair deficiency syndrome is more associated with T-cell lineage leukemia and lymphoma than B-cell lineage [31]. However, in this study, we did not find pathogenic or likely pathogenic variants among those gene mutations (PAX5, ETV6, NF1, BLM, ATM , et al.), which have high penetrance for ALL [7], possibly because we enrolled a relatively small number of unselected sporadic cases.
In this study, we did identify four pathogenic or likely pathogenic variants (TP53, CTC1, TINF2, and LIG4 ) among 82 genes associated with 23 syndromes well known for their connection to hematologic malignancy. All four of the variants were heterozygous. Li-Fraumeni syndrome (TP53) is a well-known rare cancer syndrome. The most common cancers in patients with Li-Fraumeni syndrome are solid cancers (such as breast cancer, lung cancer, and bladder cancer) [15]. However, a somatic TP53 alteration is strongly associated with low hypodiploidy ALL (~90%), disease relapse, and germline origin (~40%) [32]. Although hypodiploid ALL accounts for only 5% of childhood ALL cases, hypodiploid ALL patients should be tested for Li-Fraumeni syndrome because of its poor prognosis and the possibility of a germline TP53 mutation [33]. CTC1, TINF2, and, LIG4 are genes associated with inherited bone marrow failure syndrome, whose clinical features overlap with DC [16]. Fanconi anemia, DC, Diamond-Blackfan anemia, and Shwachman-Diamond syndrome are also associated predominantly with solid cancers, whereas myelodysplastic syndrome and acute myeloid leukemia are predominant among hematologic malignancies rather than ALL [34]. Therefore, it is unclear whether these variants are causative genetic factors of ALL in our patients. A functional study or familial study, along with a physical and clinical investigation, is needed to confirm the correlations between ALL and these variants.
We also found 31 germline pathogenic or likely pathogenic variants in genes other than those associated with the 23 syndromes, including a CNV (CASP10). Among these 31 germline variants, ADA, CASP10, IL12RB1, JAK3, LPIN2, MEFV, and TYK2 are genes associated with PID [35]. Given that CTC1, TINF2, and, LIG4 are also associated with PID, nine of our patients (9.7%) had pathogenic or likely pathogenic heterozygous variants associated with PID. All these genes are inherited in the AR pattern except MEFV (AR or AD, familial Mediterranean fever) and TINF2 (AD, DC).
More than 300 distinct disorders and genes of PID have been classified by the International Union of Immunological Societies PID expert committee [35]. Familial Tyk2 deficiency (TYK2) cause mendelian susceptibility to a mycobacterial disease associated with defects in intrinsic and innate immunity.
An increase in leukemia/lymphoma with PID is well known. A large analysis of PID patients in the United States Immune Deficiency Network database found significant increases in lymphoma (>8 fold-changes in PID cancer incidence over what was expected) [36].
Leukemia also increased, with 1.43 fold-changes in men and a 1.0 fold-change in women.
The mechanism of leukemogenesis in PID is unclear, but intrinsic (cancer predisposition parallel to the immunological defect) and extrinsic (following chronic infections, inflammation, or loss of immunosurveillance) mechanisms were proposed by Hauck et al [37].
We were unable to determine whether our patients with variants associated with PID were primary immunodeficiency patients, carriers, or carrying non-pathogenic variants because all of them were heterozygous, and appropriate clinical/laboratory findings or functional studies for the identified variants were unavailable. Nevertheless, the high prevalence of PID-associated gene variants in our ALL patients suggests that PID-associated germline mutations are important in hematologic malignancies in Koreans, especially ALL. However no comprehensive study of the prevalence of cancer or lymphoma/leukemia among Korean PID patients has been done. There is not even a national registry for PID in Korea. One large study of PID in Korea observed 152 PID patients (<19 years) from 2001 to 2005 [38].
The most common diseases were antibody deficiencies (such as congenital agammaglobulinemia or IgA deficiency) and chronic granulomatous disease, though one patient died from lymphoma [38]. Further study of PID and the risk of malignancies is needed because of the relatively high prevalence of genetic alterations association with PID that we found in Korean ALL patients.
In addition to gene variants associated with hematologic/immunologic disorders, we also identified various pathogenic or likely pathogenic variants in this study. Genes associated with bilirubin metabolism (UGT1A1), thrombophilia (THBD), thalassemia (HBD), platelet glycoprotein IV deficiency (CD36), and medulloblastoma (PTCH2), among others, were identified in this study. As reported in previous studies, variants not directly associated with hematologic malignancies are frequently identified incidentally [7,24]. The genes/diseases identified are representative of common genetic diseases in this ethnic group. The clinical significance of these variants is the diagnosis of unexpected co-morbid genetic diseases and supporting the management of specific complications [7].
Our study has several limitations. First, we enrolled a relatively small number of cases.
We enrolled few or no patients with ETP, hypodiploid ALL, and adult ALL. Second, skin fibroblasts were not used in our search for germline mutations. Third, we did not perform a familial study or a clinical or physical investigation of the identified germline variants.

Conclusions
In conclusion, using targeted NGS, we found recurrent somatic alterations (sequence variants and CNVs) in Korean ALL patients. The somatic alteration results enabled us to perform a comprehensive genetic analysis. The presumed diagnosis of clinically important sub-types of ALL, such as BCR-ABL1-like ALL and ETP, was also enabled by targeted NGS.
We identified significant germline mutations in Korean ALL patients. In addition to wellknown predisposition genes, we identified variants associated with PID in 9.7% of our Korean ALL patients. Taken together, the results of our study suggest several important points: 1) paired tumor-normal matched sample analyses are recommended to discriminate between somatic and germline mutations; 2) gene panels must include genes associated with predisposition, in addition to recurrent somatic genes; 3) genes associated with PID are recommended for the gene panel, in addition to genes known to predispose patients for hematologic malignancies.

Consent for publication
All patients provided written informed consent for publication.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
All authors have declared no conflicts of interest.     Figure S1. Workflow for the NGS data analysis. Figure S2. CASP10 gene deletion detected by the NGS CNV analysis in a patient with B-ALL. Figure S3. Somatic sequence variants in ALL patients.    Tables   Table S1. Genes included in the hematologic cancer panel.  Figure 1 Clinical outcomes by cytogenetic group. Clinical outcomes by IKZF1 alteration.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.