Designing high affinity target-binding peptides to HLA-E: a key membrane antigen of multiple myeloma

Identification of upregulated DEGs in MM

GSE6691 [9] and GSE39754 [10] datasets were found on Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo) database with MM patients and normal plasma cell samples. The number of upregulated DEGs on MM in the GSE6691 and GSE 39754 datasets were 580 and 682, respectively. The overlap between the two datasets consisted of 134 upregulated DEGs (Figure 1A).

Identification of HLA-E as an anti-myeloma target by PPI network construction and hub gene analysis

The protein-protein interaction (PPI) network was constructed from the 134 co-upregulated DEGs though STRING (http://www.string-db.org/, version 10.5) [11] and visualized by Cytoscape (http://www.cytoscape.org/, version 3.6.0) (Figure 1B). Five modules were recognized as key clusters by molecular complex detection (MCODE, version 1.4.2) [12] clustering algorithm that comes with Cytoscape. Three of these modules had MCODE scores greater than 4 (Figure 1C–1E) and, thus, were considered key modules involved in the occurrence of MM. These three modules consisted of 31 genes. Hub genes were involved in the occurrence of myeloma, and identified by cytoHubba (version 0.1) [13] (Figure 1F). The four genes (DDOST, HLA-E, NME1, and CANX) were found in the overlap between the MCODE and cytoHubba analyses, which indicating that these genes may play an important role in MM (Table 1 and Figure 1G).

To design a peptide drug candidate for the treatment of MM, the suitable targets had to be membrane proteins. Thus, DDOST, HLA-E, and CANX were selected as potential targets. NME1 was excluded because it is an intracellular protein. However, a protein lacking available structural information could not be used as a target for the computational design of a peptide drug. Following a search of the RCSB Protein Data Bank (PDB, https://www.rcsb.org), HLA-E was the only key membrane protein suitable for designing anti-myeloma peptide drug-candidate. And HLA-E was in the cluster 3 and hub genes as shown in Figure 1E, 1F.

Differential HLA-E expression in myeloma patients and controls

We demonstrated that HLA-E was overexpressed on MM samples based on GEO database analysis. To further investigate the HLA-E expression on MM cells, we collected bone marrow samples (including 6 MM and 6 normal controls) in Shengjing hospital. After sorted by CD138 beats, HLA-E mRNA was detected. There were significantly higher HLA-E expression levels in the MM samples compared with the controls (P < 0.05, Figure 2).

Design of HLA-E targeting peptides

Crystal structure of HLA-E complex interacting with CD94/NKG2A (PDB ID: 3CDG) [14] had been used as the foundation for designing affinity target-binding peptides (Figure 3A) in the present study. The pocket in which HLA-E interacts with CD94/NKG2A was considered as the dominant site[15, 16]. The crystal structure data 3CDG was uploaded into the Molecular Operating Environment (MOE) software 2018.01 (Chemical Computing Group ULC, Montreal, Quebec, Canada) (http://www.chemcomp.com/). After refining and energy minimization steps, the area of the HLA-E complex that interacts with CD94/NKG2A was identified. Subsequently, a detailed analysis of the binding sites was carried out to investigate the favorable ligand-receptor contact regions. Both CD94 and NKG2A could interact with the pocket of HLA-E. The model peptides were obtained from CD94 or NKG2A which could interact with the pocket of HLA-E. Before docking, peptides conformation searching was done to make the peptide more realistic. Finally, a peptide segment containing 12 amino acids obtained from CD94 which was more stable, could be used as a model peptide for binding more accurately to the pocket of HLA-E. Protein docking showed that the model peptide could interact with the pocket of the HLA-E complex (Figure 3B; HLA-E pocket, pink surface; peptide, green, red, and blue balls). The site view of the model peptide docking with the pocket of HLA-E could ensure the interaction but with a low affinity for forming few chemical bonds (Figure 3C, Table 2). With the aim of enhancing the affinity of the peptides to HLA-E, a peptide library was constructed by randomly replacing non-key amino acids in the model peptide using MOE software (Figure 3D). Docked the top three peptides from peptide library (i.e., highest affinity and stability) with the pocket of HLA-E (Figure 4, Table 2). The results showed that PEPTIDE 3 was predicted to have the highest affinity and S value (Figure 4E, 4F and Table 2). PEPTIDE 1 and 2 also had higher affinity to HLA-E compared to the model peptide (Figure 4A–4D and Table 2). Based on the results, PEPTIDE 3 could markedly increase the bonding affinity, indicating that it was a potential peptide drug candidate for targeting HLA-E in myeloma.

As previous report, only with leader peptide could HLA-E protein express on the surface of the cells [16]. There were several ways HLA-E could obtain the leader peptide (nine amino acids with different sequences) [17]. We compared the affinity between the designed affinity peptide and leader peptide with HLA-E. The final best pose with the highest score was analyzed and compared to the binding mode of the PEPTIDE 3 with the HLA-E protein. According to the binding free energy of HLA-E-peptide complex, we found the designed PEPTIDE 3 was binding more stronger than the original leader peptide (PEPTIDE 3 with the highest score of 17.06 kcal/mol and leader peptide with the highest score of 15.26 kcal/mol). In addition, the interactions between peptides and HLA-E were analyzed. As seen in Figure 5, the designed PEPTIDE 3 (Figure 5B) had more interactions with HLA-E protein than the original peptide (Figure 5A). Figure 5A showed that leader peptide can form hydrogen bonds between Val1 and Tyr159, Met2 and Glu63, His9, Arg5 and Trp133, Ser147, Leu9 and Arg65 of HLA-E. Hydrogen bonds between Leu9 and Gln113 of CD94, Lys164 of NKG2A protein could also be formed to increase the binding energy. However, the designed PEPTIDE 3 could form more interactions which include hydrogen bonds between Ala2, Asp4, Glu5, Glu8, Asp9, Lys10, Asn11, Arg12 and Thr163, Gln156, Ser143, Tyr84, Lys146 of HLA-E protein. It could also form several hydrogen bonds between PEPTIDE 3 and CD94 protein, such as Glu5, Asn11, Arg12 and Gln112, Ser110, Asn158, Ser109 (As seen in Figure 5B). The computational study indicated that the designed PEPTIDE 3 bound more stronger to the HLA-E protein and had more binding free energy.

Verify the targeting peptides’ affinity against HLA-E by FCM and confocal laser microscopy

The expression of HLA-E on the surface of 293T cells and HLA-E plasmid transfected 293T cells were detected by flow cytometer (FCM, ACEA NovoCyte^TM, Hangzhou, China). The result showed that 293T cell line didn’t express HLA-E protein on cell surface (Figure 6A). The expression rate of HLA-E on 293T cell line was 72.53% after HLA-E plasmid transfection (Figure 6A). The HLA-E high expressed 293T cells and 293T cells were incubated with different concentration fluorescein isothiocyanate (FITC) labeled model peptide and PEPTIDE 1-3 for 1h. And then detected the affinity of targeting peptides to HLA-E by FCM. The result showed that with the concentration growing, more peptide could be detected in HLA-E high expressed 293T cells (Figure 6B–6E). PEPTIDE 3 had the highest affinity to HLA-E which was consistent with computational result (Figure 6E, 6F). After that HLA-E blocking antibody was used to incubate with HLA-E plasmid transfected 293T cells for 30mins, and then incubated with 50μg/ml PEPTIDE 3. The result showed that fluorescence intensity nearly disappeared (Figure 6E, 6F).

U266 cell line is a common myeloma cell line. HLA-E protein could not be detected on the surface of U266 cell line by FCM (Figure 7A). As previous experiment introduced that HLA-E protein need leader peptide to express on myeloma cells surface [18]. With this protocol, HLA-E protein was detected on the surface of U266 cells by pretreating with leader peptide (Figure 7A). This HLA-E expressed U266 cell line was used to detect the FITC labeled peptides’ affinity by confocal laser microscope. PEPTIDE 3 which had highest affinity to HLA-E was used in this part. With the concentration growing, more FITC labeled PEPTIDE 3 could be detected on the HLA-E expressed U266 cells (Figure 7B–7E). When HLA-E expressed U266 cells pre-treated with HLA-E blocking antibody, the affinity of P3 to HLA-E disappeared (Figure 7F).

Microarray data

GSE6691 and GSE39754 datasets were downloaded from the GEO database and combined for analysis. The datasets included bone marrow samples from 182 MM patients and 11 healthy volunteers (control samples).

Identification of upregulated differentially expressed genes (DEGs)

The upregulated DEGs in MM samples compared to the controls were screened using GEO2R (http://www.ncbi.nlm.nih.gov/geo/geo2r). Log2FC (fold change) > 1.5 and P-value <0.05 were set as standard criteria. The co-DEGs between GSE6691 and GSE39754 were identified by Venn diagram online (http://bioinformatics.psb.ugent.be/webtools/Venn/).

PPI network construction and cluster analysis

PPI network of co-upregulated DEGs was constructed using the STRING online database. A combined score greater than 0.4 was considered statistically significant. Cytoscape was used to visualize the PPI networks. The whole PPI network was clustered into several key modules using MCODE. The criteria for selection were as follows: MCODE score > 4, degree cut-off = 2, node score cut-off = 0.2, k-core = 2, and maximum depth = 100.

Identification of hub genes

CytoHubba was employed to determine the hub genes of the PPI network. In this study, BottleNeck calculations were used to identify the top ten hub genes.

Patients and bone marrow samples

Bone marrow samples from six MM patients and six non-hematological malignant patients were collected from the Hematological Department, Shengjing Hospital of China Medical University from Sep 2016 to Jun 2017 with informed consent. All experimental protocols were approved by the Ethics Committee in Shengjing Hospital. MM cells were purified with CD138 magnetic beads using positive selection according to the manufacturer’s instructions (Miltenyi Biotec, Germany) [44].

Real-time quantitative RT-PCR (RT-qPCR)

Total RNA was extracted using Trizol (TaKaRa, Japan). Reverse transcription (2.5 μg total RNA) was performed using a cDNA synthesis kit (TaKaRa, Japan). qRT-PCR was performed with SYBR Premix EX Taq (TaKaRa, Japan) using the QuantStudio 3 real-time PCR system (ThermoFisher, USA). RT-qPCR primers were synthesized as described [45]: HLA-E forward primer, 5'-ccgtcaccctgagatgga-3'; HLA-E reverse primer, 5'-agcaatgatgcccacgat-3'. β-actin was used as a reference and amplified using primers 5'-ccaaccgcgagaagatga-3' and 5'-ccagaggcgtacagggatag-3'. qRT-PCR was performed by denaturation at 95°C for 3 min followed by 40 cycles of denaturation at 95°C for 12 sec and annealing at 62°C for 40 sec. The relative levels of HLA-E expression were calculated as ΔCt = Ct(gene) - Ct(reference). The 2^-ΔΔCt method was used to calculate the fold-change of gene expression.

Design of high-affinity targeting peptides against HLA-E

Crystal structure of homo sapiens’ HLA-E complex interacting with CD94/NKG2A was obtained from the PDB website. The result showed that there were two crystal structures available listed as PDB code 3CDG and 3CII. 3CDG with the higher resolution which would be used as the foundation for the design of affinity targeting peptides. Interactions between proteins and peptide-protein docking were performed by MOE software. To design affinity peptides targeting HLA-E, we firstly fulfilled the structure preparation of the HLA-E-CD94/NKG2A complex by addition of hydrogen, removal of H₂O molecules, and energy minimization. Next, we identified the key interacting area between HLA-E and CD94/NKG2A by analyzing the bonding and mining literature data. A peptide obtained from CD94/NKG2A was used as a model peptide for interaction with the HLA-E pocket. The peptide library was constructed based on the model peptide using the residue scan module in MOE software for random replacement of non-key amino acids. All peptides with predicted high affinity for HLA-E were optimized as follow. Firstly, conformational searching of these peptides was conducted with LowModeMD method; Secondly, docking was conducted based on two steps (placement and refinement) of Dock module in MOE. Triangle matcher and rigid receptor method were used in these two steps. 100 poses would produce in the placement step and at last exported 10 poses. At last, London dG and GBVI/WSA dG scoring function were used to estimate the affinity of peptide to HLA-E. The binding affinity of these peptides could be predicted by forming interaction bonds and binding free energy (S value obtained from docking simulations).

Verify the targeting peptides’ affinity against HLA-E by FCM

293T cells as a tool cell line which collected by our laboratory were used for construction HLA-E high express cell lines by transfecting HLA-E plasmid with Lipofectamine 3000 (Invitrogen) as previous report [46]. Affinity peptides which labeled by FITC were purchased from Chinese Peptide (Hangzhou, China). The expression of HLA-E was detected using antihuman fluorescent monoclonal antibody allophycocyanin (APC) and performed by flow cytometer (FCM, ACEA NovoCyte^TM, Hangzhou, China). The HLA-E antibodies were purchased from Biolegend (clone: 3D12, San Diego, CA, USA). 5μg/ml, 10μg/ml, 25μg/ml, 50μg/ml model and PEPTIDE 1-3 were cultured with HLA-E high expressed 293T cells and 293T cells. Then affinity of peptides against HLA-E were detected by FCM. Firstly, a ‘live gate’ was circled. Secondly, Cells were first gated by their FSC/SSC properties, and then the percentage of each concentration FITC-labeled peptides were computed (Supplementary Figure 1).

Verify the targeting peptides’ affinity against HLA-E by confocal laser microscopy

U266 cells was a common kind of myeloma cell lines which could be inducted to express HLA-E by adding leader peptide as previous report [28]. U266 cells were incubated with 500 μM of HLA-B7 (VMAPRTVLL) or a control peptide (VGRGRAFVLI) for 12 h at 37 °C. Cultured U266 cells without peptide or with the solvent of the peptides (DMSO) considered as negative control. HLA-E expression was determined by FCM.

HLA-E high expressed U266 cells and U266 cells were used for verification the affinity of the targeting peptides against HLA-E by confocal laser microscopy [47]. Preparations for microscopic observations as follows: FITC-labeled targeting peptides solution of 1μL (50mg/ml) and 1mL cells (4×10⁵) were mixed into a six-hole plate for 1h and then fixed by polyformaldehyde for 30mins. Membrane and nuclei were dyed by DLI and DAPI for 30mins after washing and resuspending the cells with phosphate buffer solution (PBS). Nikon C2 confocal microscope was used for microscopic observations [48]. The sample was stored in dark place at 4°C, and then observation was carried at 20mins after dyeing.

Statistical analysis

The data were expressed as the mean ± SD and two group statistical comparisons of means were calculated using the Student's t-test (SPSS software 23.0). A P-value < 0.05 was considered statistically significant.