High Co-Expression of PDCD1/TIGIT/CD47/KIR3DL2 in Bone Marrow Is Associated with Poor Prognosis for Patients with Myelodysplastic Syndrome

Cellular immune disorder is a common characteristic of myelodysplastic syndrome (MDS). Abnormal natural killer (NK) cell function has been reported in MDS patients, and this is closely related to disease progression and poor prognosis. However, little is known about the association between the abnormal immune checkpoint (IC) that results in abnormal immune NK cell function and the prognosis of MDS. In this study, RNA-sequencing data from 80 patients in the GSE114922 dataset and bone marrow (BM) samples from 46 patients with MDS in our clinical center were used for overall survival (OS) analysis and validation. We found that the NK cell-related IC genes PDCD1, TIGIT, CD47, and KIR3DL2 had higher expression and correlated with poor OS for MDS patients. High expression of PDCD1 or TIGIT was significantly associated with poor OS for MDS patients younger than 60 years of age. Moreover, co-expression of PDCD1 and TIGIT had the greatest contribution to OS prediction. Interestingly, PDCD1, TIGIT, CD47, and KIR3DL2 and risk stratification based on the Revised International Prognostic Scoring System were used to construct a nomogram model, which could visually predict the 1-, 2-, and 3-year survival rates of MDS patients. In summary, high expression of IC receptors in the BM of MDS patients was associated with poor OS. The co-expression patterns of PDCD1, TIGIT, CD47, and KIR3DL2 might provide novel insights into designing combined targeted therapies for MDS.


Introduction
Myelodysplastic syndrome (MDS) is a group of heterogeneous malignancies with distinct natural histories that are characterized by inefective clonal hematopoiesis, abnormal hematopoietic cell morphology, and varying degrees of cytopenia where one-third of patients progress to acute myeloid leukemia (AML) [1][2][3]. Tis disease is more prevalent in older patients aged 65-70 with less than 10% of patients younger than 50 [4,5]. In elderly patients, the incidence is 7-35 per 10 5 with men more likely to develop MDS than women [6]. According to the Revised International Prognostic Scoring System (IPSS-R), treatment strategies vary for patients with diferent risk stratifcations [7]. Although chemotherapy, hypomethylating agents, and allogeneic hematopoietic stem cell transplantation have partially benefted patients, MDS treatment still poses a great challenge [8][9][10]. Multiple factors have been involved in the pathogenesis of MDS, such as cellular immune dysfunction. Previous studies have reported abnormal natural killer (NK) cells in MDS patients with overexpression of immunosuppressive molecules [11], decreased expression of activating NK receptors, reduced antibody-dependent cytotoxicity (ADCC), and lowered direct NK cell lytic function [12]. Moreover, haploidentical NK cell therapy has been reported to achieve complete remission by reducing high-risk MDS clones [13]. Reconstructed NK cells achieve better functional education and help reduce relapse in patients when donors and hosts express all of the KIR ligands for donor KIRs [14]. Tese fndings suggest that targeting NK cell-associated receptors may be a novel immunotherapeutic strategy for MDS patients.
In contrast to T cells, NK cells can migrate to many tissues and initiate immune responses to infections or cancers, and they are able to dissolve certain target cells via cytotoxicity mechanisms that release substances, such as granzymes and perforin, without sensitizing the host [15][16][17]. Increasing evidence has demonstrated that NK cells are defective in patients with solid tumors [18] or MDS [19], indicating that NK-mediated immune surveillance of tumors may be disrupted, and immunosuppression and immune escape may contribute to disease progression [12,20]. Recent studies have suggested that immune checkpoint (IC) receptors, such as programmed cell death protein-1 (PD-1), programmed death-ligand 1 (PD-L1), T cell immunoglobulin and ITIM domain (TIGIT), CD47, and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4), are highly expressed in many types of cancer and are currently targeted to improve antitumor responses [21][22][23][24]. As the classical immune checkpoint, PD-1 binds to its ligands PD-L1 and programmed death-ligand 2 (PD-L2) to allow tumor cells to evade immune surveillance [25]. In numerous cancers, TIGIT signaling negatively regulates antitumor immunity [26], and CD47 overexpression has been reported to be associated with poor survival and a higher rate of progression to AML in MDS patients [27]. Moreover, cytokines secreted by KIR3DL2 expressing NK cells and T cells could promote the progression of malignances [28]. Our previous study found that dysregulated T cells, caused by low B-cell leukemia/lymphoma 11B (BCL11B) expression, are associated with the prognosis of MDS; however, the efect of abnormal IC receptor expression on NK cells on the clinical outcomes of MDS patients remains unclear [29]. Furthermore, the antitumor efects of IC inhibitors (ICIs) alone are limited, which may be due to heterogeneity in IC receptor expression levels and distinct dominant IC expression patterns in diferent MDS patients [30,31]. Terefore, it is worth studying the expression patterns of IC proteins in MDS.
In this study, we investigated the prognostic importance of IC proteins in relation to NK cells using RNA-sequencing data obtained from newly diagnosed MDS patients from the Gene Expression Omnibus (GEO) database, and the results were further validated by quantitative real-time PCR (qRT-PCR) in our clinical center.

GSE114922 Dataset.
Te RNA-sequencing data from 80 de novo MDS patients and corresponding clinical information in the GSE114922 dataset were downloaded from the Gene Expression Omnibus (GEO) database (https:// www.ncbi.nlm.nih.gov/geo/). Overall survival (OS) was defned as the time from the date of diagnosis to the date of death or last follow-up. Te clinical information of the patients, including age, gender, OS status, cancer type, and risk stratifcation, is listed in Table S1. Te analysis process is shown in Figure 1. Because the GSE114922 dataset is publicly available, approval by a local ethics committee was not required.   Table S2. Te expression levels of PDCD1, TIGIT, CD47, and KIR3DL2 are presented as 2 −ΔCT .

Statistical Analysis.
All statistical analyses were performed using R (version 4.0.2, https://www.r-project.org/) and Statistical Product and Service Solutions (SPSS) (version 22.0, IBM, Armonk, NY, United States) software. Te function "surv_cutpoint" in the R package "survminer" determined the optimal cutof value for continuous variables ( Figures S1 and S2). Te log-rank test was used to compare diferences in Kaplan-Meier curves. Cox proportional hazards models were constructed with the R package "survival." A two-tailed P value <0.05 was considered to be statistically signifcant.

High Expression of PDCD1, TIGIT, CD47, and KIR3DL2 in the BM of MDS Patients Is Associated with Poor OS.
To investigate the prognostic contribution of NK cell-related IC receptors in MDS patients, we performed survival analysis based on expressions of these genes in the GSE114922 dataset. After the optimal cutof values for PDCD1, TIGIT, CD47, and KIR3DL2 were obtained, the patients were divided into low-and high-risk groups ( Figures S1 and S2).

PDCD1, TIGIT, and CD47 Have Potential for Stratifcation Prediction in MDS Subgroups.
To study the correlation between PDCD1, TIGIT, CD47, and KIR3DL2 and clinical characteristics, we performed a subgroup analysis using the GSE114922 dataset. As shown in Figure 3  PDCD1, TIGIT, CD47, and KIR3DL2 on the clinical outcomes of MDS patients, we analyzed diferent combinations of these genes in Kaplan-Meier curves. Signifcantly, MDS patients who were PDCD1 high TIGIT high , TIGIT high CD47 high , PDCD1 high CD47 high , or PDCD1 high KIR3DL2 high had a worse OS in the GSE114922 dataset (HR > 3, P < 0.001) ( Figure S3). Tese results were confrmed in the JNU-SMU dataset (HR > 2, P < 0.05) (Figures 4(a)-4(f )). Importantly, to further identify which co-expression pattern has the greatest contribution to OS, we used HR as an evaluation criterion. Interestingly, the top three combinations that contributed to OS prediction were PDCD1/TIGIT, TIGIT/CD47, and PDCD1/KIR3DL2, and PDCD1/TIGIT had the greatest contribution in the GSE114922 dataset (HR = 5.45, Figure 4(g)). Tis fnding was again confrmed in the JNU-SMU dataset (HR = 3.07, Figure 4(h)).

New Risk Stratifcation Based on the Nomogram Model
Shows Better Performance on OS Prediction. Because standard risk stratifcation based on IPSS-R and the expression levels of PDCD1, TIGIT, CD47, and KIR3DL2 were significantly associated with the prognosis of MDS patients, these were all used to construct a nomogram model to visually predict the 1-, 2-, and 3-year survival rates for MDS patients in the GSE114922 dataset ( Figure 5(a)). Te detailed points and OS rates are shown in Table S3. To provide more precise prognostic prediction for MDS patients, we generated a new risk stratifcation for patients based on the total points derived from the nomogram model. After obtaining optimal cutof values 131 and 203 using X-tile software, the patients were divided into low-, intermediate-, and high-risk groups ( Figures S4A and S4B). Interestingly, patients with higher risk (total point >203) based on the nomogram model had worse OS than those with low or intermediate risk in the GSE114922 dataset (P < 0.001) ( Figure 5(b)). Tis result was also shown in standard risk stratifcation based on IPSS-R in the GSE114922 dataset where patients with high risk had worse OS compared with those with low or intermediate risk

Discussion
It has been demonstrated that aberrant expression of IC receptors is signifcantly associated with NK cell dysfunction in MDS, but little is known about their association with the prognosis of MDS patients [11,12]. In this study, we found that high expressions of PDCD1, TIGIT, CD47, and KIR3DL2, which  Harnessing the immune response to target malignant tumors through the use of immune checkpoint blockade (ICB) has now become a vital breakthrough in treating solid tumors [32][33][34] and hematological malignancies [35]. In this study, we found that overexpression of PDCD1, TIGIT, CD47, and KIR3DL2 predicted poor OS for MDS patients, and this may indicate that the proteins of these genes might serve as targets for immunotherapy. Nowadays, various clinical trials of ICBs in MDS have been carried out; however, remission could be observed in only a small percentage of MDS patients who were treated with a single dose of monoclonal antibody to ICBs [30,[36][37][38][39]. Due to the limited response activities of ICB as a single agent, combination therapy with diferent ICBs may be used to overcome primary and acquired resistance to single-agent administration and provide more clinical beneft for patients [40]. In a variety of solid tumors, the results of multiple clinical trials of ICB combination therapy for malignant tumors have shown that combining ICBs can lead to higher response rates with manageable safety profles and good tolerability [41][42][43][44]. Primary data from an ongoing Phase II clinical trial suggested that combination of PD-1 and CTLA4 monoclonal antibodies can achieve 58% projected 1-year survival in R/R AML patients [45,46]. However, clinical trials of ICB combination therapy in MDS patients are recruiting and ongoing, and there are currently no relevant preliminary data [47][48][49][50]. Our previous research has found that the combination of two IC proteins was a better predictor for the prognosis for AML patients than individual ones [51]. In this study, we demonstrated that co-expression of PDCD1 and TIGIT may work the best for OS prediction in MDS, which may provide a feasible and efective scheme for ICB combination therapy for MDS patients.
It is common practice to detect cytogenetic changes in patients during the diagnosis of hematological malignancies to stratify risk and predict clinical outcomes, but due to the heterogeneity and complexity of the diseases, individuals difer in responses to chemotherapy, duration of remission, and recurrence although they have the same cytogenetic changes [52]. In recent years, genetic aberrations have signifcant practical value, and some have been used in National Comprehensive Cancer Network (NCCN) guidelines [53,54]. Although the standard classifcation method based on IPSS-R includes chromosomal abnormalities, it lacks precise gene expression changes and mutations and cannot further individualize risk stratifcation for MDS patients. Tus, in this study, we constructed a new risk stratifcation consisting of PDCD1, TIGIT, CD47, KIR3DL2 and the risk stratifcation based on the IPSS-R, which may provide more precise prognosis predictions for MDS. However, the small sample size of this study may have statistical bias, which may account for the lack of signifcant correlation between standard risk stratifcation based on IPSS-R and prognosis of patients in JUN-SMU dataset ( Figure 5(e)). Moreover, there were a large number of low-risk patients in the GSE114922 dataset, while the proportion of high-risk patients in the JUN-SMU dataset was relatively high. Due to clinical sample limitations, we could not collect enough low-risk patients with sufcient follow-up time; thus, we will need to expand the sample size and include more samples from lowrisk patients to verify our results in the future.
In conclusion, we demonstrated that high expression of PDCD1, TIGIT, CD47, and KIR3DL2, which are related to NK cells, is associated with poor OS for MDS patients, and co-expression of PDCD1 and TIGIT might be the best OS predictor for MDS. Moreover, a new risk stratifcation paradigm consisting of PDCD1, TIGIT, CD47, and KIR3DL2 expression and the standard risk stratifcation based on IPSS-R could provide more precise prognosis predictions for MDS. Tese fndings might provide novel insight into prognosis stratifcation and designing combined targeted therapy for MDS.

Data Availability
Te raw data used and/or analyzed during the current study may be available from the corresponding authors upon reasonable request.

Disclosure
MZ and CC are co-frst authors.

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Authors' Contributions
MD, YL, and XH contributed to the concept development and study design and edited the manuscript. MZ and CC performed the experiments, analyzed and interpreted the data, and wrote the manuscript. WZ, JT, and JC collected the clinical information. All authors have read and approved the fnal manuscript. MZ and CC contributed equally to this work.
(bottom) of CTLA4. (B) Te optimal cutof value (upper) and OS analysis (bottom) of KLRC1. Figure S2: the optimal cutof values of PDCD1, TIGIT, CD47, and KIR3DL2 in the GSE114922 (A-D) and the JNU-SMU (E-H) datasets. Figure S3: co-expression patterns of PDCD1, TIGIT, CD47, and KIR3DL2 were related to poor OS in MDS patients in the GSE114922 dataset. Figure S4: the optimal cutof value of the total points in the monogram model was provided by Xtile software (version 3.6.1). Table S1: clinical information of the MDS patients. Table S2: primers for qRT-PCR. Table S3: points and OS rates in the nomogram model. (Supplementary Materials)