Microarrays have become a cornerstone in genomic research, providing a high-throughput platform to explore genetic variations across the genome. This technology enables researchers to study the impact of rare variants on common diseases, identify mutations in causal genes, and explore the link between environmental factors and phenotypes [1]. At present, hundreds of different microarray chip types are available, some containing over a million genetic markers [2]. Microarray platforms have revolutionized tumor cell analysis, offering a powerful tool to detect genome-wide genetic alterations [3]. Particularly, single-nucleotide polymorphism (SNP) arrays allow for the large-scale evaluation of single-nucleotide variants (SNVs), which could reveal new genetic markers with potential prognostic significance in hematological malignancies [3,4].

Multiple myeloma (MM) is a mature B-cell malignancy characterized by the proliferation of abnormal monoclonal plasma cells in the bone marrow and the overproduction of non-functional immunoglobulins or free immunoglobulin light chains [5]. Patients show a variable clinical course, with some cases progressing rapidly, while others have a more favorable outcome with an overall survival greater than ten years. The disease is associated to multiple genetic events that influence different signaling pathways, modifying the biological features of myeloma cells, and determining proliferative and selective advantages [6].

Despite significant advances in understanding MM genomics, the potential of microarray technology to understand disease biology and identify novel biomarkers remains unexplored. Early studies identified candidate genes involved in the transition from monoclonal gammopathy of uncertain significance (MGUS) to MM [7]. For instance, López-Corral et al. showed an increased incidence of copy number aberrations during disease progression [8]. Similarly, Kim et al. found that increased genomic complexity was associated to poor outcomes in patients receiving bortezomib plus melphalan and prednisone therapy [9]. A recent review highlighted the importance of microarray-based approaches for a better understanding of MM complexity and heterogeneity [10].

The present study conducted a large-scale analysis of SNVs in newly diagnosed MM (NDMM) patients to identify variants with potential prognostic implications. Additionally, as a secondary objective, this study assessed the utility of SNP arrays to differentiate between MGUS and MM.

An initial analysis was conducted to identify SNVs associated with the four IMWG treatment response categories. This approach identified 692 statistically significant SNVs distributed across 527 distinct genes, each meeting the criterion of t-score >3. These 692 SNVs effectively classified the dataset according to PIR, although some overlap was observed between the very good partial remission (VGPR) and complete remission (CR) groups (Supplementary Figure S1). The application of a more stringent criterion (t-score >4) identified 42 of the most significant SNVs, located in 39 genes and exhibiting a differential association with cancer (Table 2). Among them, three SNVs were associated with CR, one with VGPR, five SNVs with partial remission (PR), and 33 with

Subsequently, a search was conducted for SNVs associated with the ER and non-ER categories in patients undergoing front-line autologous hematopoietic stem cell transplantation (auto-HSCT). In this analysis, 151 SNVs distributed across 93 genes were identified as statistically significant based on the criterion of t-score >3 for each group (Supplementary Figure S2). Only 12 overlapping SNVs were found between ER and ≤PR groups, seven of which significantly affect clinical outcomes (Table 3). The examination of allele frequencies for these variants revealed no deviation from HWE. Two patient groups were differentiated based on the presence or absence of alternative alleles at the following loci: Protein tyrosine phosphatase receptor type D (PTPRD rs12343415, rs77411943, rs10978084), Notch Receptor 4 (NOTCH4 rs8192588), SH3 Domain Containing Ring Finger 3 (SH3RF3 rs76256617), DCC Netrin 1 Receptor (DCC rs72920200), and CUB and Sushi Multiple Domains 1 (CSMD1 rs11781684).

Group A (n = 9) included patients who were either heterozygous or homozygous for the alternative allele, whereas Group B (n = 47) included those homozygous for the reference alleles. Patients in Group A exhibited a significantly higher ≤PR rate compared to those in Group B (88.89% versus 31.91%; p-value = 0.005) (Figure 1 A) and had a higher ER rate than Group B (88.89% versus 40.43%; p-value = 0.021) (Figure 1 B).

Fig. 1.

Genotype-Based Analysis of partial remission (PR) and early relapse (ER). A) Comparison of ≤PR rates between Groups A and B (Grey ≤PR, White ≥ GPR); B) Comparison of early relapse (ER) rate between Groups A and B (Grey = ER, White = non-ER). Group A included patients carrying homozygous or heterozygous genotypes for alternative alleles, whereas Group B comprised patients homozygous for the reference alleles for variants of the PTPRD (rs12343415, rs77411943, rs10978084), NOTCH4 (rs8192588), SH3RF3 (rs76256617), DCC (rs72920200), and CSMD1 (rs11781684) genes.

An interpopulation comparison of allele frequencies showed that alternative alleles for SH3RF3, PTPRD, and CSMD1 gene variants were significantly more frequent in the present cohort than in the LA2 population, with no differences observed compared to Europeans. The minor allele frequencies (MAF) for SH3RF3 rs76256617 and PTPRD rs10978084 observed in the present cohort were 0.018, compared to 0.002 in the LA2 (p-value = 0.047). Additionally, for CSMD1 rs11781684, a MAF of 0.045 was observed in this study, while in Latin Americans it was 0.040 (p-value <0.001) (Table 4).

Survival analysis revealed that Group A had a 5-year OS rate of 40%, that is, significantly lower than Group B (76.1%). Similarly, the median PFS was markedly shorter in Group A (eight months) compared to Group B (47 months). A multivariate analysis, adjusted for first-line auto-HSCT and International Scoring System (ISS), confirmed that Group A had significantly poorer outcomes for both OS (p-value = 0.008) and PFS (p-value = 0.017) (Figure 2).

Fig. 2.

Kaplan Meier curves comparing overall survival (A) and progression free survival (B) between Group A (dotted line) and Group B (solid black line). Group A included patients presenting the alternative allele for the PTPRD (rs12343415, rs77411943, rs10978084), NOTCH4 (rs8192588), SH3RF3 (rs76256617), DCC (rs72920200), and CSMD1 (rs11781684) genes, and Group B patients homozygous for the reference allele.

Because the majority of Group A cases (8/9) were not transplanted, a sub-analysis was conducted of non-transplanted patients (n = 18: Group A = 8; Group B = 10). ER occurred in 7/8 (87.5%) of Group A versus 8/10 (80.0%) of Group B. Additionally, Group A had a higher proportion of cases with relapse at 12 months; 75% (6/8), compared with Group B 60% (6/10).

The comparison between MM and MGUS patients, following rigorous filtering, resulted in the identification of 103,942 SNVs. Subsequent Binary Discriminant Analysis revealed 251 significant SNVs mapped to 193 genes and successfully clustered MM and MGUS using PCA (Dim1 16.9%, Dim2 4.5%) (Supplementary Figure S3). Differentially represented SNVs between MM (13) and MGUS (8) are shown in Table 5.

This large-scale analysis of SNVs in NDMM patients identified variants associated with IMWG treatment response categories, highlighting the genetic complexity that influences therapy outcomes in this disease. The genetic profile enabled stratification between ER and non-ER in auto-HSCT patients and revealed only 12 overlapping SNVs between the ER and ≤PR groups, seven of which have a significant impact on clinical outcomes, suggesting shared genetic drivers.

The analysis of overlapping SNVs revealed that alternative alleles in the PTPRD, NOTCH4, SH3RF3, DCC, and CSMD genes were strongly associated with poor outcomes in the present cohort. These genes participate in different cancer-associated signaling pathways. PTPRD, located on chromosome 9p23, interacts with pathways such as STAT3, JAK, mTOR, and β-catenin; its inactivation has been implicated in various malignancies including colon, gastric, and breast cancers, glioblastoma, melanoma, and with MM pathogenesis [14,15]. Similarly, NOTCH4 (6p21.32), contributes to oncogenic signaling through interactions with molecules like Sox2, c-MET, and CD44 [16]. SH3RF3 (POSH2)(2q13), defined by its SH3 and ring finger domains, enhances cancer stem cell properties in a JNK-dependent mechanism [17]. The DCC gene (18q21), encoding the tumor-suppressor Netrin-1 receptor, is frequently silenced in colorectal cancer [18]. The CSMD1 gene (8p23) is commonly deleted in cancers such as head and neck squamous cell carcinoma (50%), breast cancer (55%), and lung cancer (46%). In addition, it is associated to tumor mutational burden, mismatch repair deficiency, and PD-L1 expression [19].

The literature offers limited evidence regarding the application of SNP-arrays in MM. Kamada et al. [20]. found a deletion of PTPRD (9p23) in MM patients with a non-hyperdiploid karyotype as well as in human MM cell lines. More recently, a study by Campo et al. [21]. identified four genetic risk variants associated with the development of bortezomib-induced peripheral neuropathy, including DCC rs17748074. These findings, together with the results of the present study, support the use of SNP arrays to identify clinically relevant variants in MM and highlight the need for further studies in diverse MM cohorts.

The population-based comparison of allele frequencies showed that MAFs of SH3RF3 rs76256617, PTPRD rs10978084, and CSMD1 rs11781684 were higher than expected in the current cohort compared to the LA2 population but not to the European population. These results are likely influenced by the highly heterogeneous genetic structure of Latin American populations. Both Uruguay and Argentina share a strong European ancestry, particularly from Spain and Italy, and have undergone similar demographic processes, including the admixture with Native American and African populations [22–24]. The differences observed between the current cohort and other Latin American populations may reflect distinct genetic backgrounds, likely shaped by regional variations in ancestry components. This is in line with recent genomic studies showing that Latin American populations exhibit marked substructure due to complex admixture histories and demographic processes [25]. Given the underrepresentation of Latin Americans in global genetic studies [26], the findings of this study highlight the importance of further research in genetic characterization of these populations.

The present analysis demonstrates that specific SNVs, identified through comprehensive genotyping, indicate that microarray technology can adequately classify distinct groups of PIR and ER patients. Moreover, specific SNVs were associated with poorer induction response, shorter PFS, and reduced OS. Furthermore, the differential representation of several SNVs between MM and MGUS highlights genetic distinctions that may be clinically relevant for both diagnosis and prognosis. These findings underscore the role of SNVs in predicting treatment response and clinical outcomes in MM. Interestingly, a recent meta-analysis using four GWAS datasets found a significant genome-wide association between the G allele of SKT10 rs28199 and increased MM risk, with no observed heterogeneity among studies [27]. More recently, Zhu et al. [28]. combining microarray analysis and Mendelian randomization, identified the myeloperoxidase (MPO) gene as a novel biomarker potentially implicated in MM pathogenesis and progression. Thus, the identification of predictive markers could be used to guide treatment options, as well as to monitor residual disease, providing significant benefits in MM management [10].

In this study, high-resolution SNV genotyping using microarray technology identified several SNVs that can effectively classify patients in different PIR groups. Furthermore, despite the limited number of patients, the results suggest that the presence of alternative alleles may be associated with inferior induction response, PFS, and OS in NDMM. To our knowledge, this is the first study showing the involvement of specific genes and variants in the context of MM. The findings reported here support the potential predictive role of the identified SNVs in disease progression and prognosis. However, given the small sample size and number of events, further investigation is necessary to elucidate the precise mechanisms underlying these associations and their clinical implications.

All authors contributed to the study´s conception and design. DG, FS, and CG were responsible for methodology, data acquisition, analysis, and interpretation of results. DG, SL, and ER were responsible for clinical data collection. ML contributed with bioinformatics analysis. DG, ER, AF, and IS were responsible for the study conceptualization. DG, AF, and IS writing the first draft of the manuscript, and reviewing and editing the final version. All authors read and approved the final version of the manuscript.

This work was supported by grants from IMS and the Laura Rodger Riney Foundation and CONICET (National Research Council), Argentine (PUE: 2018-0042 and PIP 1179).

The study was approved by the Local Ethics Committees, and it is in accordance with the current version of the Helsinki Declaration. All patients provided their written informed consent.