Chemogenomics Mapping of Potential Drugs and Targets for Treatment of Multiple Myeloma

Multiple myeloma (MM) is the second common hematological malignancy affecting about 352,000 worldwide. Some subgroups of MM patients still cannot benet from the currently available anti-MM drugs and therefore are at high risk of death. The pathological mechanism of MM remains to be unraveled. The identication of a global gene signature for MM might lead toward development of novel diagnostics and therapeutic interventions. Here, we identied common differentially expressed genes (DEGs) shared by 30 MM microarray data sets and compared the common DEGs with those induced by genetic or chemical perturbations. We found some potential therapeutic targets for MM treatment, for example RARA, FGFR1, PML, ROR1, SLAMF7, MTDH and Daxx. as modulating them can reverse the MM-induced gene signature. Based on our analysis results, we also predicted and validated some drug reposition, such as Imatinib, Decitabine, Dexamethasone, Vincristine, Paclitaxel, as well as Bortezomib plus Balomycin A1 combination for MM treatment by a literature search, data mining, and in vitro bioassays. This study could provide guidance and indications for the development of MM specic diagnostic biomarkers, indication predictors and therapeutic treatment.


Introduction
Multiple myeloma (MM) [1] is a blood cancer characterized by the abnormal proliferation of plasma cells accumulated in bone marrow. In a healthy condition, the plasma cells make antibodies that ght infections. But in multiple myeloma patients, they release too much abnormal protein into the bones and blood, releasing chemicals and causing organ damage and lytic lesions in the bone. MM is the second most prevalent hematological carcinogenesis currently affecting 86,000 patients in the US. It is estimated that currently there are 352,000 MM patients worldwide. In the US, about 30,280 new adult cases increase annually, and about 12,590 subjects are reported to die from MM each year, accounting for 0.9% of all cancer deaths (https://www.cancer.net/cancer-types/multiple-myeloma/statistics) Unlike other hematological carcinogenesis with a single associated genetic aberration, MM is a heterogeneous disease with its major subtypes being classi ed, diagnosed and predicted according to several different genetic abnormalities of plasma cells [2]. With help from the development of techniques, large efforts have been made to advance our understanding of the complex multistep pathogenesis of MM, involving identi cation of abnormal gene expression pro les, bone marrow microenvironment change, cytogenetic rearrangement, cell survival signaling and other pathways. The enhanced insight of the molecular pathology of MM allows the development of increasing effective new-targeted therapies, such as thalidomide, bortezomib, and lenalidomide, which dramatically improve the clinical outcomes of MM by substantially increasing the overall survival of several subtypes of MM patients. Despite these dramatic improvements, unfortunately, this devastating disease is still incurable. Also, currently available MM drugs cannot bene t a large portion of newly diagnosed patients (15%) and these patients are still at high risk for relapse and death [3]. This indicates that some key aspects of the underlying pathobiology mechanism of MM remain elusive [2]. Furthermore, effective therapy is needed to treat this high-risk group.
The unsatis ed therapeutic effect on some subgroups of MM patients reveals the fact that some unknown underlying mechanism and complex genetic pathways are associated with the disease. In past decades, DNA microarray technology allowed us to interrogate a large number of genes and identify the disease-related genes. Analysis of the DNA microarray result often generates thousands of differentially expressed genes (DEGs), making it di cult to identify genes that are responsible for a particular disease condition from an individual experiment.
To address this problem, here we applied a meta-analysis approach on publicly accessible MM patient microarray datasets to generate a global gene pro le for MM. As shown in Figure 1, the MM datasets were analyzed separately at rst to discover the differentially expressed genes (DEGs) between the disease and health groups. Then DEGs obtained from each dataset were overlapped to nd the common DEGs shared by different datasets, which is credible relevant to MM. Pathway enrichment analysis was done to identify the most MM associated biological processes and other disease conditions based on common DEGs. In addition, we went a step further by comparing each individual MM gene expression dataset with a gene signature induced by a drug or gene perturbation, and attaining the most correlated medications and genes for each dataset. The common MM correlated gene regulations shared by most datasets will help us nd potential biomarkers for diagnosis and indication prediction as well as nd future therapeutic targets that might have a capacity for MM treatment. The common MM negatively associated drugs appeared in most datasets might have a potential to be repurposed to treat patients with MM, as they can reverse the gene expression change in MM disease conditions. Finally, we will con rm our prediction about drug reposition by literature study and experiment validation. The ndings can help to nd genes that have the potential to be novel biomarkers for diagnostic and therapeutic interventions against MM, as well as discover drugs that can be repurposed to treat MM patients. This study offers a drug reposition approach based on meta-analysis of gene signature data.

Enrichment Analysis for Common DEGs shared by MM datasets
We then applied 1439 common DEGs shared by at least 15 MM datasets as an input to do enrichment analysis for the biological process network (Figure 2A), the pathway map ( Figure 2B), GO process ( Figure   2C), and disease phenotypes ( Figure 2D) using MetaCore. Figure 2 displays the top 10 highly related pathways and the top 20 diseases. Not surprisingly, we found that more than half of the top 10 pathway maps and several process networks highly associated with MM datasets are tightly involved in cell proliferation, translation, transcription, and cell cycle, which are important biological pathways that are highly expressed in cancer cells due to the rapid and unlimited cell proliferation potential of cancer cells. In addition to these common pathways induced by multiple cancers, we identi ed several pathway maps, biological process networks and GO processes that are closely related to immune response and in ammatory signaling, especially IL-6 and B cell antigen receptor (BCR) signaling, which is speci c for the immune response caused by B cell. As we mentioned before, multiple myeloma is a type of cancer caused by malignant plasma cell, the DNA impairment in B cell make abnormal B cells, which will turn to malignant plasma cells in response to a stimulus, followed by the release of abnormal antibodies, leading to tissue damage. Thus, the BCR signaling is speci c for MM, same as the IL-6 signaling in MM ( Figure 2B). Moreover, viral process and parasitism symbiosis are identi ed as top listed GO process, which indicates a potential correlation between viral infection and multiple myeloma. Furthermore, ubiquitin-proteasome system is also emphasized here in the top ranked pathway maps, which further indicates a role of proteolysis via USP and autophagy in MM, in accordance with our previous analysis. Figure 2D shows the top 20 diseases that have a higher correlation with common DEGs genes shared by MM datasets. First, many of the other types of neoplasms and carcinoma are intensely connected to MM, as different cancer types often share a lot of similar pathological pathways. They occur in different tissues and organs, including uveal tracts, breast, ducts, lobules, endocrine gland, eye, and ovary. Additionally, we also nd some diseases speci cally related to multiple myeloma, such as the bone marrow diseases and hematological diseases. Multiple myeloma is a cancer of plasma cells t located in bone marrow, and it is the second most common hematologic malignancy. Interestingly, virus diseases, RNA virus infection, and immunological de ciency syndromes are ranked at the top of the diseases, indicating a tight association and potential common pathological pathways shared by viral infections like HIV and MM, which is consistent with our previous results.

Positive and negative correlated genetic perturbations
Top positively ( Figure 3A) and negatively ( Figure 3B) correlated genetic knockdown with MM datasets is identi ed. We identi ed the signi cant related genetic perturbations with each MM dataset using Knockdown Altas function in BaseSpace. We then took the highly connected gene perturbations with a gene perturbation score higher than 50 into account for summing up data of 30 MM datasets, nally to determine and rank the highly positive or negative correlated gene perturbations, including gene knockout, knockdown, mutation, and overexpression. Positive correlations between DEGs induced by speci c single gene knockdown/knockout and MM induced DEGs may imply that the functions of these genes are impaired during the MM process, because knockdown/knockout of these genes will induce similar gene signature change as MM. Our ndings are consistent with other literature reports. Negative correlations between a single gene knockdown/knockout DEGs and MM DEGs might suggest that this gene overexpresses in the MM disease condition, and blocking this single gene expression or inhibiting the protein encoded by this gene will reverse the differentially expressed genes in the pathological process of MM and help treat the disease.
The gene signature induced by ROR1 knockdown is highly negatively connected to the MM induced data. A functional study suggested that ROR1 is a part of Wnt signaling pathway and can promote cancer cell survival [1,4]. Overexpression of the receptor tyrosine kinase-like orphan receptor 1 (ROR1) has been identi ed in B-cell chronic lymphocytic leukemia (B-CLL) and chronic myelogenous leukemia (CML), but not in normal white blood cells and reactive lymph nodes, suggesting ROR1 might serve as a tumorspeci c target for therapy [4,5,1].
SLAMF7 is negatively associated with MM by 16 negative associations and no positive association with 30 MM gene sets. This gene encodes Signaling Lymphocyte Activation Molecule Family member 7 (SLAMF7), a therapeutic target for MM treatment. In 2015, the FDA approved Elotuzumab (a humanized monoclonal antibody targeting SLAMF7) as a treatment for patients with multiple myeloma who have received one to three prior medications [6]. Now Elotuzumab is a global approved medication used in combination with other anti-MM drugs to treat relapsed/refractory and advanced MM. It demonstrates its e cacy to improve symptoms when added into other MM therapy without causing any signi cant additive toxicity [7].
Another top negatively associated knockdown gene is MTDH (Metadherin), which has shown a crucial role in several aspects of tumor progression, including transformation, evasion of apoptosis, invasion, metastasis, and chemoresistance. MTDH overexpression is frequently observed in different cancers, including melanoma, glioma, neuroblastoma, as well as liver, breast, prostate, and esophagus cancers. Overexpression of MTDH activates downstream PI3K/AKT, NF-κB, and Wnt/β-catenin signaling pathways to stimulate cancer cell proliferation, cell survival, invasion and chemoresistance. MTDH also mediates the transformation of oncogene Ha-Ras and c-Myc and the adhesion of tumor cells to remote organs to promote transformation and metastasis. MTDH can be a therapeutic target to block several processes of tumor progression [8].
Sixteen positive correlations were identi ed between MM datasets and DEGs induced by knock down of SATB1, "a cell-type speci c global gene regulator" by folding chromatin into loop domains [9]. Scientists reported that SATB1 can regulate gene expression to promote growth and metastasis of breast tumors [10]. Here, we observe the reverse correlation between SATB1 and MM: the knockdown of SATB1 has a positive association with MM, indicating blocking of SATB1 might regulate genes to promote MM.  Figure 3), which are also demonstrated to be involved in carcinogenesis pathologies. Our analysis suggested that FGFR1 mutations might cause FGFR1 overexpression in tumor cells. Fibroblast growth factor receptor 1 (FGFR1) plays an essential role in the regulation of cell proliferation, differentiation and migration via mediating the downstream signaling, such as MAPK, AKT/PI3K, and Ras. Ampli cation of FGFR1 was reported in breast cancer [12], non-small [13] and squamous cell lung cancer [14], urinary bladder cancer [15], and prostate cancer [16]. Several tyrosine kinase inhibitors targeting FGFR1 are currently applied for cancer treatment. Regorafenib, Sorafenib, Lenvatinib, Nintedanib and Pazopanib are approved by the FDA to treat carcinoma that occurs in different tissues and organs, such as hepatocellular carcinoma, advanced renal cell carcinoma, non-small cell lung carcinoma and differentiated thyroid carcinoma (DTC) that is refractory to radioactive iodine [17]. Negative correlated overexpression also shares some common gene with negative correlated mutations, such as ZBTB16 and MAPT. ZBTB16 plays a role in myeloid maturation and serves as a component of E3 ligase that mediates protein ubiquitination and subsequent proteasomal degradation for abnormal proteins, which might be associated with acute leukemia and multiple myeloma [18]. MAPT encodes the microtubule-associated protein tau (MAPT), and its mutations have been associated with some neurodegenerative disorders such as Alzheimer's disease, Pick's disease, and frontotemporal dementia.

Positive and negative associations between MM-and drug-induced gene pro les
Signi cant differentially expressed genes with a score over 50 compared to each MM datasets were collected and summarized for a total of 30 datasets. Top ranked drugs, which induced DEGs with more than 15 negative associations compared to MM induced DEGs among 30 datasets, are listed in Table 1, with their therapeutic targets and number of negative and positive associations. Drugs that induced DEGs with negative correlations with MM, might be able to reverse the gene expression change caused by myeloma, and have a bene cial effect to treat MM. Most of the highly negatively related medications are anticancer agents, anti-infective drugs, dietary supplements and immune regulators. Among the top listed drugs with negative associations (Table 1), several drugs have already been approved and applied to treat MM or investigated in clinical trials. Table 2 lists nine currently used anti-MM drugs that are also identi ed from the top list of MM gene negatively associated drugs with their development stage, regimen, number of positive and negative associations, as well as reference. For example, both Imatinib and Ergocalciferols have 18 negative correlations with MM gene datasets and only a few positive correlations (3 for Imatinib, 1 for Ergocalciferols). Imatinib, a tyrosine kinase inhibitor, is now in clinical trial Phase II in a single drug form and in combination use for MM treatment. Ergocalciferol, also known as Vitamin D2, is used in combination with other anti-tumor drugs as a nutrient supplement to treat MM [20]. Another dietary supplement, Folic Acid, also known as Vitamin B9, was also identi ed to have potential for treating MM, which is consistent with the presence of this drug in clinical trial Phase II for combination use in MM. Dexamethasone, a type of corticosteroid medication with 18 negative associations and 8 positive ones, was approved by the FDA to treat patients with MM in combination with other drugs [21]. Additionally, four other antitumor agents, including Sirolimus, an mTOR inhibitor, Decitabine, a nucleic acid synthesis inhibitor, as well as Vincristine and Paclitaxel, tubulin inhibitors, are in different clinical trial phases (Phase I for Sirolimus and Decitabine, Phase III for Vincristine, Phase II for Paclitaxel) to treat MM along or with other medications. For some of these drugs listed in Table 1, they have the potential to be repurposed for the treatment of multiple myeloma because of their possible capacity to reverse the mRNA level change in pathological conditions. We checked the literature rst and did some simple in vitro assays using some easily accessible medications to verify the gene signature analysis prediction for drug reposition.
As shown in Figure 5A, although Folic Acid and Ba lomycin A1 themselves did not have much antiproliferation effect, the combination use of Folic Acid or Ba lomycin A1 with Bortezomib will improve the anti-proliferation activity of Bortezomib (a well-known FDA approved anti-MM drug) in MM.1.S cells (a multiple myeloma cell line) with 16h medication treatment. The other two drugs predicted to be repurposed for MM treatment in our study, Sorafenib and Doxorubicin, also showed some antiproliferation effects on the MM cell line. These results indicate that our prediction via gene signature analysis might be reliable on the cellular level.
To further investigate the signaling pathway these medications go through to kill the tumor cells, we treated MM1.S cells with these drugs and their combinations for 16h at their effective concentrations.
Then we collected the cell lysate samples and ran Western blot to measure the caspase-3 level changes, a well-known apoptosis marker, in uenced by different drugs. As shown in Figure 5B we found that cotreatment of Bortezomib with folic acid or Ba lomycin A1 induced more caspases 3 cleavage (19kDa and 17kDa) than treatment with Bortezomib alone. Ba lomycin A1 and Folic acid also induce caspase 3 cleavage compared with DMSO alone, but the effect is much weaker than bortezomib. In addition, in accordance with our MTT result, treatment of Sorafenib also induces the cleavage of caspase 3, induces apoptosis signaling and causes tumor cell death.
To summarize, both the common DEGs from the MM datasets and the drugs that induce negative correlated transcriptional pro les are involved in the pathogenesis of infection, immune, and in ammatory pathways or are indicated to treat viral, bacterial and fungal infection, in ammation and immune diseases. This result suggests a tight relationship between multiple myeloma and these diseases, from molecular mechanism to therapeutic treatment. It was reported that infections with the HIV and hepatitis C virus appeared associated to an enhanced multiple myeloma risk. Infections are a major threat to multiple myeloma patients, with majority of morbidity and mortality due to infections [31,32]. In a nationwide population-based study in Sweden, multiple myeloma patients showed a 7-fold risk to develop a bacterial infection and a 10-fold risk for viral infection compared to matched controls (P<0.001) [33,32]. The complications of MM are the major cause for the death of patients. It is necessary to introduce the anti-infective agents in combination with anti-tumor therapy for the treatment of infectious complications in MM patients, which is consistent with our drug reposition prediction [32].
As for the immune response, multiple myeloma is a cancer of plasma cells, which are cells responsible for producing antibodies in response to foreign antigen and debris stimulations. Multiple myeloma disease will lead to the dysfunction of plasma cells, which is followed by immunode ciency. It has been shown that MM patients have a low immune response to infection, which is one of the causes for a higher risk of infection [34,35]. "Managing the complications of the disease and its treatment, such as infections, thrombosis and neuropathy, has become more important as MM patients survive longer [32]." Drugs with negative associated gene signature compared to MM in our study also include cardiovascular agents and CNS medications, suggesting possible medication reposition to treat the complications of the disease.

Declarations
Compliance with Ethical Standards Con ict of interest: The authors declare that they have no competing interest.
Ethical approval: This article does not contain any studies with human participants or animals performed by any of the authors.
Informed consent Informed consent was obtained from all individual participants included in the study
Author Contribution SM with the idea of writing this manuscript. XX supports the work. SM wrote the rst draft. All authors read and approved the nal version.

Available of data and material
We collected the AD-induced DEGs from commercial software Illumina BaseSpace software (Santa Clara, CA, USA, http://www.nextbio.com). The data can also be reached using GEO ID in Table 1 in http://www.ncbi.nlm.nih.gov/geo/. The associated gene and chemical perturbations from each dataset were reached and downloaded through Pharmaco Atlas search and Knockdown Atlas search in BaseSpace. It's a commercial software, we will upload the data in this study we downloaded using the software.