Recurrent genetic defects on chromosome 5q in myeloid neoplasms

Background Deletion of chromosome 5q (del(5q)) is the most common karyotypic abnormality in myeloid neoplasms. Materials and Methods To define the pathogenic molecular features associated with del(5q), next–generation sequencing was applied to 133 patients with myeloid neoplasms (MDS; N = 69, MDS/MPN; N = 5, sAML; N = 29, pAML; N = 30) with del(5q) as a sole abnormally or a part of complex karyotype and results were compared to molecular features of patients diploid for chr5. Findings A number of 5q genes with haploinsufficient expression and/or recurrent somatic mutations were identified; for these genes, CSNK1A1 and G3BP1 within the commonly deleted 5q region and DDX41 within a commonly retained region were most commonly affected by somatic mutations. These genes showed consistent haploinsufficiency in deleted cases; low expression/mutations of G3BP1 or DDX41 were associated with poor survival, likely due to decreased cellular function. The most common mutations on other chromosomes in patients with del(5q) included TP53, and mutations of FLT3 (ITD or TKD), NPM1 or TET2 and were mutually exclusive. Serial sequencing allowed for definition of clonal architecture and dynamics, in patients with exome sequencing allelic imbalance for informative SNPs facilitated simultaneous approximation of clonal size of del(5q) and clonal burden for somatic mutations. Interpretation Our results illuminate the spectrum of molecular defects characteristic of del(5q), their clinical impact and succession of stepwise evolution.

The commonly deleted regions (CDR) in del(5q) have been extensively studied with a distal region (CDR1:5q32-33) often deleted in the 5q-syndrome [2] and a proximal region (CDR2:5q31) in higher-risk MDS and AML [13]. SNP-array-based karyotyping helped to further refine the boundaries of the CDR (CDR1;145,299,747-153,828,955 and CDR2;137,500,665-139,471,723), and commonly retained regions (CRR1;from the centromere to 5q14.2 and CRR2;from 5q34 to the telomere) [5]. Patients with small interstitial deletions were shown to have a better outcome as compared to those with larger deletions [5]. Haploinsufficiency of several genes in the CDRs likely contributes to specific phenotypic features in del(5q). For instance, heterozygous deletions resulting in haploinsufficient expression of RPS14 is a key determinant of ineffective erythropoiesis [14], while thrombocytosis and megakaryocytic dysplasia may be related to haploinsufficient miR-145/miR-146a [15]. However, experimental knockdown of these genes did not result in a growth advantage and haploinsufficiency is not uniformly present in all cases to explain clonal dominance. Recurrent hemizygous mutations of genes within the deleted locus have not been identified with the notable exception CSNK1A1 missense mutations found in only 3/40 del(5q) cases, but absent in heterozygous configuration [16]. Moreover, most of del(5q) cases involve large deletions encompassing both CDRs, and thus identification of genes contributing to individual clinical phenotypes has been challenging. It is likely that pathogenetic mechanisms in del(5q) may involve hemizygous mutations or haploinsufficiency and be modified by additional somatic lesions affecting genes on other chromosomes. Furthermore, the position of del(5q) within the clonal hierarchy might also affect the phenotype and clinical behavior.
To characterize the genetic and genomic complexity and clonal hierarchy in myeloid neoplasms with 5q abnormalities, we used next generation sequencing (NGS), including whole exome sequencing (WES) and targeted multiamplicon deep sequencing in a cohort of patients with del(5q) in a comparison to patients with diploid chr5. In addition, to explore ancestral events in del(5q), we compared clonal size of individual somatic mutations with that of del(5q) identified by WES.
Interpretation: Our results illuminate the spectrum of molecular defects characteristic of del(5q), their clinical impact and succession of stepwise evolution.
When we focused on mutations in the CDR, we found 70 alterations in 57 genes (27% of all alterations on 5q) including CSNK1A1 (5q32) and novel recurrently   Supplementary Table S3 for individual genes). Copy number status of chr5 was demonstrated as follows: diploid 5q, green and blue lines; and del(5q), a red dashed line. Two commonly retained regions (CRR1 and CRR2) and a commonly deleted region (CDR), defined by SNP-A karyotyping analyses, are represented by vertical rectangles.

Expression of frequently mutated genes on chr5
Haploinsufficiency caused by deletion of 5q involving multiple genes is likely the key pathogenetic mechanism in 5q-syndrome. We hypothesized that heterozygous hypomorphic mutations ultimately result in haploinsufficient expression, thereby phenocopying haploinsufficiency due to deletions. We investigated the expression levels of genes located on 5q in comparison between cases with/without del(5q) focusing on genes which have been previously reported or we have found affected by heterozygous mutations. For definition of haploinsufficiency we set the cut-off value < 60% of normal. In total, 12/27 genes showed haploinsufficient expression in del(5q), with the majority of haploinsufficient genes located in the CDR (7 genes) or CRR2 (3 genes) ( Figure 2). Within the CDR, G3BP1, CD74 and CSNK1A1 exhibited both haploinsufficiency and somatic mutations, whereas PPP2CA, CTNNA1 and CDC25C showed haploinsufficient expression but no mutations. SH3RH2 and SH3TC2 genes did not display haploinsufficiency, while somatic events of these genes were noted. Other recurrently mutated genes located outside of the CDR: such as APC, DDX41 and MAML1 also showed haploinsufficient expression; however, mRNA levels of GPR98, FAM170A, a cluster of protocadherin family genes and NPM1 were not decreased in deletion cases.

Del(5q) and genetic events on other chromosomes
The associated mutational landscape outside of the del(5q) region may also affect the clinical and biological features of del(5q) cases. We thus analyzed the potential relationship of somatic mutations observed in other chromosomes in del(5q) cases. Globally, 5q-syndrome cases were associated with lower numbers of mutations (average 2.5 mutations/case) compared to IDR deletions (9.5 mutations) and patients with extreme deletions (involving CRR1 and CRR2; on average 18 mutations) by WES. TP53 mutations were associated with del(5q) as previously described by our group [19] and others [20][21][22]. In contrast, 10/15 of top mutated genes showed a significantly mutual exclusivity with del(5q) (e.g., TET2, NPM1, FLT3-ITD/TKD) ( Figure 3A). Mutations of DNMT3A, SF3B1, ZRSR2, NRAS and BCOR were evenly distributed. The correlation of TP53 mutations with del(5q) was most prominent. TP53 mutations with del(5q) was mostly occurred with other chromosome abnormality, though, only 1 case was seen with isolated del(5q) (Supplementary Table S2). In low risk MDS, TP53 mutations were detected in 13% of del(5q) cases and in only 0.5% of diploid 5q cases ( Figure 3B; P = .0001). Among high-risk MDS, TP53 was mutated in 42% of del(5q) vs. 4% of diploid 5q patients ( Figure 3B; P < .0001). When we focused on the extent of the deletion, somatic TP53 mutations were particularly frequent (39%, 17/38) in cases whose deletion involved both CRR1 and CRR2, and in 32% (12/38) of interstitial deletion cases ( Figure 3C, P = .03). Moreover, large deletions tended to be a part of complex karyotypes and 17p abnormality ( Figure 3D).

Prognostic impact
To investigate clinical implications, we initially assessed the impact of the deleted lesion on clinical outcomes, in which follow-up data were available. As expected, patients with isolated del(5q) showed better prognosis compared with del(5q) with other chromosomal abnormality (P < .001, Figure 4A). When we investigated the size of deletion, patients with both CRR1/CRR2 lesions (involving the 5q extremes) showed a worse prognosis compared with cases including CRR1 or CRR2, or with IDR lesions (P = .01, Figure 4A). We also investigated the impact of the presence of TP53 mutations (TP53MT) as the most common mutational event associated with del(5q). Predictably, survival among patients with del(5q) was inferior for TP53MT compared to wild type TP53 (TP53WT) cases (P < .001, Figure 4A). Furthermore, there were significant survival differences reflective of the previously described differences in the extent of deleted regions on 5q in TP53WT cases (P < .001, Figure 4B). TP53MT cases showed inferior outcome regardless of deleted region ( Figure 4B).
When we focused on the prognostic value of lowexpressed genes in primary AML cohort, low expression of G3BP1 and DDX41 correlated with a shorter survival (P < .001 and P = .04, Figure 4C), an effect that was not seen in CSNK1A1 cases ( Figure 4C). We also could not find association between low-expression and outcome in MAP1B, GPR98 and FAM170A which did not reach the haploinsufficiency cut off in our cohort (data not shown).

Del(5q) and hierarchical clonal architecture
It has been presumed that the del(5q) is the ancestral event in the myeloid neoplasms harboring this lesion [23]. Using deep sequencing and 'allelic imbalance' we can determine the position of del(5q) in the hierarchical clonal architecture [24]. In our cohort, del(5q) was present in 17-98% of tumor cells and there was good correlation to the size of del(5q) clone by FISH (r = .94; Figure 5A). We identified three patterns of recurrent clonal architecture in del(5q) cases ( Figure 5B) i) apparent pathogenic somatic mutations precede the deletion event (31%), ii) del(5q) appears to precede any other somatic mutation (19%) and iii) the succession cannot be determined because of expanded clones with similar size ("clonal saturation") i.e., these cases were not informative. In our cohort, among the majority of cases in which del(5q) was a secondary lesion, in 64% of instances a TP53 mutation was the ancestral event and in 27% of cases the primordial lesion was DNMT3A mutation ( Figure 5C). When we compared different time points in same case (MDS-phase and leukemic phase), the proportion of cells affected by TP53 mutation was more prominent than that of del(5q) (92% vs. 17%, Figure 5C right). In contrast, CSNK1A1 mutation occurred in the remaining allele after del(5q) ( Figure 5D). We also detected 1 case in which del(5q) was asserted to be ancestral, and the TP53 mutation was detected as a secondary event.

DISCUSSION
Our study identified a cohort of 178 patients with various forms of del(5q) to answer several fundamental questions related to the pathogenesis of myeloid malignancies associated with these deletions: i) is del(5q) Oncotarget 6489 www.impactjournals.com/oncotarget Oncotarget 6490 www.impactjournals.com/oncotarget associated with recurrent hemizygous mutations; ii) may heterozygous mutations corresponding to haploinsufficient genes mimic the phenotype of the deletion; iii) are gene mutations on other chromosomes recurrent in del(5q); iv) what is the architecture and clonal evolution pattern in del(5q) myeloid neoplasms? Does 5q still stand as the primordial lesion in the light of data generated from the use of the new genomic platforms?
We found several somatic hemizygous mutations in del(5q) cases, including G3BP1 and CSNK1A1. Mutations in CSNK1A1 were found in a canonical E98 position as recently reported [16,25]; only hemizygous mutations were found with del(5q) in the context of various clinical subtypes, including aggressive diseases RAEB-1 or therapy-related MDS. CSNK1A1 E98 mutations increase β-catenin activity thereby providing selective growth advantage. G3BP1 is another gene encoded within the CDR, and unlike CSNK1A1 mutations those in G3BP1 occurred both in heterozygous and hemizygous configuration. G3BP1 is known to control p53 activity through a dual pathway involving direct protein interaction of G3BP1-p53 and deubiquitination by regulating the ubiquitine specific peptidase USP10 [26].
Several other hematopoiesis-related genes and tumor suppressor genes are located in the CDR (e.g., CTNNA1, PPP2CA, EGR1, SPARC, RPS14 and CDC25C) with previously reported [14,[27][28][29][30][31], however, we were unable to detect hemi-or heterozygous mutations in Oncotarget 6491 www.impactjournals.com/oncotarget these genes. It is possible that they contribute to clinical heterogeneity, shape the clinical phenotype or modulate the growth advantage of the del(5q). For the purpose of our investigations we hypothesized that haploinsufficient genes in del(5q) may also be affected by loss of function/ hypomorphic mutations in diploid cases, and we have identified several genes fitting this profile. They were affected only in a minority of patients, and most did not recapitulate the clinical features of del(5q). Of note is that even the haploinsufficient expression showed variability  . (B) Initial genetic events (mutations or del(5q)) were determined by the size of affected clonal cell populations based on the following assessments (bar graph). Variant allelic frequencies (VAF) of somatic mutations were adjusted by copy numbers and LOH was based on SNP-A results. Pathogenic genes mutated as initial events included TP53, DNMT3A, NCOR2 and PRPF8 as shown in Figure 5C. VAF of SNPs with a deleted-allele due to del(5q) were calculated for allelic imbalance as mentioned in the methods section. Inconclusive cases without discrimination were categorized as clonal saturation. Distribution of the disease phenotypes in each clonal pattern was demonstrated by colors as indicated. (C) Clonal architecture of the cases (N = 9) with initial driver mutations prior to del(5q) was demonstrated by overlaid double-oval figures (left) and a serial-assessment figure (right). Percentages indicate the fraction of the cells affected by mutations or del(5q). (D) In 3 other cases, del(5q) was shown to be a primary event.
Oncotarget 6492 www.impactjournals.com/oncotarget among del(5q) cases: while average expression values may be decreased in del(5q) cohorts, specific expression is indeed haploinsufficient only in a portion of cases, and may in part reflect the relative percentage of the malignant clone in each patient. These differences in the degree of haploinsufficiency may explain, in addition to the size of deleted region, the intrinsic diversity of del(5q). Epigenetic regulation also affects the expression of each genes on del(5q) whereby deletion of unsilenced allele could even lead to gain of silencing. However, there would be no impact if silenced allele is deleted and thus may not be a key determining factor for the degree of haploinsufficiency. Because del(5q) occurred in one allele but epigenetic regulation (hyper-or hypo methylation) occurred in both allele, most likely at random. Nevertheless, several genes were found to be haploinsufficient and affected by somatic mutations, including HDAC3, CSNK1A1, G3BP1 and DDX41. Moreover, the functional role of 5q genes in hematopoiesis has been shown using a murine model.
The number of somatic mutations on other chromosomes increased with the increasing length of the 5q deletion. Co-occurrence of a TP53 mutation was particularly prominent in this del(5q) cohort, as reported in other studies [19,21,22]. It is still unclear why TP53 mutations selectively coincide with del(5q), one could speculate that loss of p53 function might overcome p53 tumor suppressor effects and foster leukemia evolution. Of note is that there are several gene clusters of negative regulators of TP53 on chr5q, such as, PPP2CA, RPS14, CSNK1A1 and G3BP1.
Among del(5q) patients we found that inferior survival was associated with patients with both deletion of CRR1 and CRR2. This relationship became evident in the TP53-wild type cohort, whereas there was no survival difference in existence of TP53 mutation. The larger deletions were frequently associated with other chromosomal abnormalities, which associated with inferior survival [32,33]. However, the reasons are still unclear. One possibility is that there are several tumor suppressor genes in this location and long deletion causes multifunctional loss of these genes. Alternatively, other gene mutations or loss of function of tumor suppressor gene may result in different clinical phenotype of extended version of del(5q). We also analyzed impact of a cohort of individual genes with low expression, which indicated recurrent somatic mutations. Low expression of G3BP1 or DDX41 correlated with inferior survival but there was no prognostic impact in CSNK1A1, CDC25C and EGR1 (data not shown). These results indicate that the presence of low expression did not always correlate with survival, and the loss of tumorsupressive function may affect their outcome. Thus, a loss of tumor suppressive function of G3BP1 or DDX41 may lead to leukemic evolution in del(5q).
To define the position of del(5q) within the clonal hierarchy, we have compared the clonal size of somatic mutations with that of del(5q) using a novel approach focusing on allelic imbalance. While it has been reported that del(5q) occurs in stem cells as an ancestral event in patients with the 5q-syndrome [23], our results indicate that the mutation of TP53 or other driver gene mutations such as DNMT3A occurred as initial events followed by deletion of 5q in a majority of case. Previously, cooccurrence of a TP53 mutation was described in various del(5q) cohorts [19,21,22], but the position of del(5q) and TP53 mutation within subclonal hierarchy could not be precisely established using Sanger sequencing. In this report we were able to overcome this shortcoming using NGS. We also found somatic mutation of APC (5q22) as initial event in del(5q) case, a role for low expression of APC in the pathogenesis of myeloid neoplasms [34].
In summary, comprehensive molecular analyses using SNP-A karyotyping, WES and targeted sequencing revealed recurrent somatic mutations involving CSNK1A1 and G3BP1 in the CDR and DDX41 in the CRR in myeloid neoplasms with the del(5q). These genes showed haploinsufficiency in deleted cases and low expression of G3BP1 or DDX41 is associated with poor survival, which may be due to loss of function. In addition, in assessing allelic imbalance in del(5q), our results suggested that del(5q) is not an universal ancestral event. Mutation of TP53 is the most common mutation in del(5q) cases and may serve as an ancestral event. These data illuminate the impact of the del(5q) in myeloid malignancy, providing deep insights into the identity and role of key genes.

Samples
Paired bone marrow and germ line (GL, CD3 + lymphocytes) DNA was obtained from 389 patients with various myeloid neoplasms and additional 631 DNA samples were included for further targeted resequencing, including a total of 178 cases of -5/del(5q) ( Table 1). All samples were obtained following written informed consent approved by the institutional review boards at Cleveland Clinic and the University of Tokyo. The Cancer Genome Atlas (TCGA) AML data set was obtained from http:// cancergenome.nih.gov/.

Determination of clonal burden
The detection of clonal size of del(5q) was accomplished by calculation of allelic imbalance for informative SNPs present within deleted region in heterozygous configuration in GL. For all heterozygous SNPs in the region, label the lost allele A and the retained allele B. For reads covering a heterozygous site in the sample, the probability that the read will carry the B allele is: P

Statistical analysis
Comparisons of proportions and ranks of variables between groups were performed by the χ 2 test, Fisher exact test, Student t test or Mann-Whitney U test, as appropriate. We used the Kaplan-Meier and the Cox method to analyze overall survival (OS) with a 2-sided P less than or equal to .05 determining significance.