Identifying Treatment Resistance Related Pathways by Analyzing Serum Extracellular Vesicles of Patients With Resistant Versus Regressed Retinoblastoma

Abstract

Purpose: To identify the genes and pathways responsible for treatment resistance (TR) in retinoblastoma (RB) by analyzing serum small extracellular vesicles (sEVs) of patients with TR active RB (TR-RB) and completely regressed RB (CR-RB).

Methods: Serum-derived sEVs were characterized by transmission electron microscopy and nanoparticle tracking analysis. sEV transcriptome profiles of two TR-RB and one CR-RB with good response (>20 years tumor free) were compared to their age-matched controls (n = 3). Gene expression data were analyzed by the R Bioconductor package. The CD9 protein and mRNA expression of CD9, CD63, and CD81 were studied in five RB tumors and two control retinae by immunohistochemistry and quantitative reverse transcription–polymerase chain reaction.

Results: The isolated serum sEVs were round shaped and within the expected size (30–150 nm), and they had zeta potentials ranging from −10.8 to 15.9 mV. The mean ± SD concentrations of sEVs for two adults and four children were 1.1 × 10¹² ± 0.1 and 5.8 × 10¹¹ ± 1.7 particles/mL. Based on log2 fold change of ±2 and P < 0.05 criteria, there were 492 dysregulated genes in TR-RB and 184 in CR-RB. KAT2B, VWA1, CX3CL1, MLYCD, NR2F2, USP46-AS1, miR6724-4, and LINC01257 genes were specifically dysregulated in TR-RB. Negative regulation of apoptotic signaling, cell growth, and proton transport genes were greater than fivefold expressed only in TR-RB. CD9, CD63, and CD81 mRNA levels were high in RB tumors versus control retina, with increased and variable CD9 immunoreactivity in the invasive areas of the tumor.

Conclusions: Serum sEVs could serve as a potential liquid biopsy source for understanding TR mechanisms in RB.

The globe salvage rates of patients with retinoblastoma (RB) have improved significantly in recent times due to advances in targeted chemotherapy and multimodal treatment protocols.¹ Nevertheless, challenges remain, especially for advanced RB, mainly because of persistent or recurrent tumors due to subretinal and vitreous seeding following chemotherapy.² Other prognostic factors such as younger age (<2 years), anterior segment disease, endophytic tumor, and increased tumor thickness (6 mm) at diagnosis have been identified as the critical factors for tumor recurrence.³ Apart from these, the expression of multidrug-resistant proteins (MRPs), enhanced drug efflux mechanisms, and the presence of cancer stem cells have been established as possible causative factors for tumor recurrence in RB.^4–6 In spite of having this knowledge, the predictive biomarkers for minimal residual disease monitoring in salvaged eyes of RB and assessing treatment response among patients with RB undergoing systemic chemotherapy are still unavailable. 

Extracellular vesicles (EVs) are a mixed population of small (30–150 nm), medium (150 nm–1 µm), and large (>1 µm) vesicles involved in intercellular communication by transferring the molecular cargo (nucleic acids, proteins, and metabolites) of cells of origin to recipient cells.⁷ EVs mediate cross-talk between tumor and stromal cells in the tumor microenvironment, induce epithelial-to-mesenchymal transition, promote angiogenesis, and mediate tumor cell migration and invasion.⁸ Moreover, EVs confer cancer chemoresistance via direct sequestering of anticancer drugs, thus reducing their concentrations at target sites, as well as by inducing resistance in drug-sensitive cells by transferring drug-resistant EV cargoes such as antiapoptotic proteins, drug efflux pumps, lipids, and nucleic acids (DNA, mRNA, microRNA, and long noncoding RNA) to drug-sensitive cells.^9,10 

Analysis of serum EVs from patients with RB is of great clinical interest, as the technique is minimally invasive and repeated serum sampling is feasible, in addition to the non-availability of RB tumor tissue prior to enucleation due to the relative contraindications of biopsy in cases of RB. Our group has established that serum sEV-derived RNA cargo from patients with advanced RB harbor RB tumor signatures involved in the epigenetic regulation of the cell cycle and oncogenic signaling pathways that drive RB tumorigenesis.¹¹ Several other groups have noted the involvement of RB small EV microRNAs (miRNAs) in tumor angiogenesis and RB vitreous seed-derived small EV proteins in invasion and metastasis.^12,13 

Whole transcriptome analysis provides a global view of all of the functional elements (RNA transcripts) of the genome for a specific physiological condition.¹⁴ Therefore, in the present study, we aimed to evaluate the transcriptome profiles of serum EVs from patients with TR active RB (TR-RB) who were scheduled for therapeutic enucleation, RB survivors with salvaged eyes with good treatment response, patients with completely regressed RB (CR-RB), and corresponding age-matched control subjects in order to identify genes and signaling pathways putatively linked to the TR phenotype. In addition, the expression of the exosome-specific markers CD9, CD63, and CD81 was evaluated in RB tumors and control retinae. 

Methods

Sample Collection

This study was conducted according to the tenets of the Declaration of Helsinki and carried out after obtaining approval from the institutional review board of L V Prasad Eye Institute, Hyderabad (LEC-BHR-P-01-21-575). Informed consent was obtained from the parents or legal guardians of children involved in the study. The diagnosis of RB was established based on clinical findings by examination under anesthesia, B-scan ultrasonography, and orbital imaging. Tumors were classified (groups A–E) based on the International Classification of Retinoblastoma classification system. Blood samples (2 to 3 mL) were collected from two patients with intraocular RB that were resistant to combined chemoreduction and focal therapy, and they underwent subsequent secondary enucleation (TR-RB); from one previously treated, completely regressed patient with RB with a history of good treatment response (CR-RB) (>20 years); and from two healthy age-matched children and one healthy adult (sibling of patient with CR-RB) with no known retinal pathology. 

The collected blood samples were centrifuged at 2000g for 15 minutes to remove any cellular debris. The supernatant samples containing the cell-free serum were stored in aliquots at −80°C. Fresh RB tumor tissues (n = 5) were collected following enucleation of the eye as part of the treatment protocol for advanced intraocular tumor. An area of maximum tumor volume based on orbital imaging was identified, a 5-mm scleral window was created, the tumor was identified, and 2 to 3 mm of the fresh tumor was obtained. The entire procedure was performed in the operation theater under aseptic precautions. The globe was then submitted for routine histopathological examination. The control retinae (n = 2) were acquired from human cadaveric eyeballs, which were collected within 6 hours of death and preserved in the sterile moist chamber by Ramayamma International Eye Bank at L V Prasad Eye Institute, Hyderabad. The globe was bisected with a blade adjacent to the optic nerve, the retina was identified and carefully dissected, and 2 mm of it was later transferred to an aliquot containing 400 µL of RNA and stored at −80°C until RNA isolation. The remaining retina was sectioned and stained to rule out any evidence of retinal disease. 

Isolation of Small Extracellular Vesicles From Serum

As described by our group earlier,¹¹ small extracellular vesicles (sEVs) were recovered from serum samples using the commercially available Invitrogen Total Exosome RNA & Protein Isolation Kit (Thermo Fisher Scientific, Waltham, MA, USA) based on exosome precipitation. Briefly, according to manufacturer's instructions, 400 to 500 µL of serum was mixed with appropriate volume of precipitation reagent. The solution was incubated overnight at 4°C. EVs were pelleted by centrifugation at 10,000g for 10 minutes, resuspended in 100 µL of phosphate-buffered saline (PBS), aliquoted, and stored at −80°C. Small EVs recovered from all the serum samples were characterized for physical properties. 

Transmission Electron Microscopy

The morphology of sEVs was examined by transmission electron microscopy (TEM) according to the technique described by Ahmed et al.¹⁵ Briefly, a 20-µL drop of EVs PBS solution containing EVs (1:100 dilutions) was loaded onto carbon-coated copper grids and permitted to stand overnight for air drying. The absorbed sEVs were negatively stained with 2% uranyl acetate for 10 minutes. Finally, the images of sEVs were captured under TEM (FEI Tecnai G2 Spirit; FEI Company, Hillsboro, OR, USA) at 80 kV after the grids were dried. 

Nanoparticle Tracking Analysis by Zeta View

The particle size, number, and zeta potential of the isolated sEVs were quantified using ZetaView (Particle Metrix GmbH, Meerbusch, Germany) and its inbuilt software (version 8.05.12 SP1) as described earlier.¹¹ The final concentrations were measured by multiplying the observed concentration with the dilution factor. The concentration of EVs present in each sample was expressed as particles/mL. Triplicates of each sample were measured, and the results were expressed as mean ± SD. 

Small EV RNA Isolation and Whole Transcriptome Analysis

Total RNA was isolated from sEV samples using the Invitrogen Total Exosome RNA & Protein Isolation Kit according to manufacturer's instructions. Briefly, the enriched exosome pellet was suspended in a resuspension buffer, and the sample was incubated for 10 minutes at room temperature. To this solution, one volume of 2× denaturation solution was added and incubated for 5 minutes at 4°C, followed by the addition of one volume of acid–phenol–chloroform solution, mixed properly and centrifuged at 10,000g for 5 minutes at room temperature. The aqueous upper phase was collected in a fresh tube. To this, 1.25 volumes of 100% ethanol was added, mixed thoroughly, and passed through a filter cartridge. After this, the RNA-bound cartridge was washed with wash buffers I, II, and III, and the total RNA was eluted using elution buffer and stored at −80°C until use. RNA purity and concentrations were determined using the Qubit RNA Assay Kit for use with a Qubit 4.0 Fluorometer (Thermo Fisher Scientific). 

Library Construction

The next-generation sequencing (NGS) libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England BioLabs, Ipswich, MA, USA) according to the manufacturer's protocol. The enriched total RNA was subjected to fragmentation and random priming, followed by first- and second-strand cDNA synthesis. Double-stranded cDNA was purified using 1.8X Agencourt AMPure XP Beads (Thermo Fisher Scientific). The End Repair/dA-tail of the cDNA library protocol and subsequent adaptor ligation were performed. Size selection and polymerase chain reaction (PCR) enrichment were done. Library quality details were checked with the Qubit 4.0 Fluorometer using the Agilent 2200 TapeStation System utilizing High Sensitivity D1000 ScreenTape analysis (Agilent Technologies, Santa Clara, CA, USA). 

RNA Sequencing and Bioinformatics Analyses

Paired-end sequencing reads of 2 × 150 bp were generated using the Illumina HiSeq 2500 System with a read length of 150 bp. The raw data files were downloaded as FASTQ files. Quality control parameters (total number of reads sequenced, GC content, and the overall base quality score), adapter trimming, quality filtering (Phred quality score > 30), and per-read quality pruning were conducted using the ultra-fast all-in-one FASTQ preprocessor fastp (version 0.20).¹⁶ The cleaned reads were then aligned to the reference genome (ftp://igenome:[e-mail]/Homo_sapiens/UCSC/hg38/Homo_sapiens) using the HISAT2 splice-aware read aligner¹⁷ and assembled into transcripts using Stringtie,¹⁸ which also estimates the expression levels of all transcripts as fragments per kilobase of transcript per million mapped (FPKM) reads. The FPKM values were used to calculate relative RNA abundances. FPKM values normalize RNA amounts produced from each gene to account for gene length differences; thus, the relative abundance of transcripts from healthy and RB serum EVs can be calculated. 

Principal Component Analysis

The gene expression data of sEV RNAs of TR-RB, CR-RB, and healthy controls (HCs) were analyzed by principal component analysis (PCA) using R (R Foundation for Statistical Computing, Vienna, Austria). A large number of variables (gene expression data) were reduced to the smallest number possible while still retaining as much as the original information. These variables were grouped into different principal components (PCs). If the data required higher PCs, that meant there was a high variance; otherwise, the data were closely correlated with each other. Scree plots revealed the percentage of variances explained by each PC. 

Analysis of Differentially Expressed Genes

Differentially expressed gene (DEG) analysis was performed using DESeq2 in the R Bioconductor package.¹⁹ The Benjamini–Hochberg procedure was used to control the false discovery rate.²⁰ A P < 0.05 and log2 fold change of ±2 were used as criteria for identifying DEGs with substantial fold change across the groups. Two independent analyses were performed: (1) TR-RB (n = 2) versus age-matched healthy controls (n = 2), and (2) a single CR-RB versus healthy adult. 

Functional Enrichment Analysis

The Database for Annotation, Visualization and Integrated Discovery (DAVID) 6.8 was used for functional annotation of the transcripts based on Gene Ontology (GO) terms (biological process [BP], molecular function [MF], and cellular component) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. P  <  0.5 was considered statistically significant. 

RNA Isolation and Real-Time PCR

Total RNA from the RB tumors (n = 5) and control retina (n = 2) was isolated by the TRIzol Reagant method. DNA contamination was removed by DNase treatment. The extracted RNA was quantified using a Thermo Scientific NanoDrop spectrophotometer, and cDNA was prepared using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). Quantitative real time PCR (RT-PCR) was then performed in a CFX Real Time PCR Thermo Cycler (Bio-Rad) using the SsoAdvanced Universal SYBR Green Master Mix (Bio-Rad) according to the manufacturer's protocols. The primers used for amplification of CD81, CD9, CD63, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) are listed in Supplementary Table S1. The relative quantity of target mRNA against the standard reference gene (GAPDH) was calculated according to the 2^−∆∆Ct method.²¹ Student's t-test was used to compare the normalized mRNA expression levels. P < 0.05 was considered to be statistically significant. 

Immunohistochemistry

The enucleated RB specimens (n = 4) were processed for routine histopathological examination and were subjected to automated tissue processing (Leica TP1020; Leica Biosystems, Wetzlar, Germany). Formalin-fixed and paraffin-embedded sections of RB enucleated specimens and one control retina were used for immunohistochemical staining of CD9 and RB1. The unstained sections were subjected to deparaffinization by three cycles of xylene washes with each cycle being 10 minutes in duration, and they were then rehydrated by subsequent incubation in ethanol (100% EtOH for 5 minutes, 70% EtOH for 5 minutes, 50% EtOH for 5 minutes, 30% EtOH for 5 minutes) followed by rinsing the slides in running water for 5 minutes. Then, microwave-based antigen retrieval was performed in 10-mM citrate buffer (pH 6.0) followed by blocking the endogenous peroxidase by incubation with peroxidase-1 (Biocare Medical, Pacheco, CA, USA). Later, tissue sections were blocked with Background Sniper (Biocare Medical) for 10 minutes and incubated with the primary antibodies against CD9 (1:50, mouse monoclonal; Abcam, Cambridge, UK) and RB1 (1:25, rabbit monoclonal; Abcam) overnight at 4°C. Post-incubation, sections were incubated with recommended polymer and probe (Biocare Medical) according to the manufacturer's protocol. Further, diaminobenzidine chromogen was added over sections along with hematoxylin counterstaining. Immunoreactivity was visualized under the microscope, and staining was assessed by two ocular pathologists. 

Results

Supplementary Table S2 shows the demographic and clinical features of the study subjects. Among two female patients with TR-RB, one had unilateral RB and another had bilateral RB. One of the TR-enucleated specimens showed high-risk features of massive choroidal invasion. The adult with bilateral CR-RB was male, 25 years old, at the time of sample collection. The mean age of patients with RB was 3.8 years, and control subjects were 8, 5, and 22 years old, respectively. 

Characterization of RB Serum sEVs

The TEM images showed that the isolated sEVs were round shaped with heterogeneous sizes ranging from 30 to 150 nm (Supplementary Figs. S1A, S1B). Nanoparticle tracking analysis revealed that the majority of the vesicles had a diameter below 200 nm (Supplementary Fig. S1C) and ZP (Zeta potential) ranging from −10 to −20 mV (Supplementary Fig. S1D). The mean ± SD size and ZP of the sEVs ranged from 110.6 ± 3.8 to 142.7 ± 23 and from 10.8 ± 0.7 to 15.9 ± 2.3, respectively (Supplementary Table S3). However, the concentration of sEVs in adults was higher compared to children's sEVs; 1.1 ± 0.1 E × 10¹² and 1.2 ± 0 E × 10¹² particles/mL were measured for patients with CR-RB and the healthy adult. But, the concentration of sEVs present in patients with TR-RB and healthy children ranged from 4.2 ± 0.3 E × 10¹¹ to 8.0 ± 1.4 E × 10¹¹ (Supplementary Table S3). 

RNA Sequencing Details

RNA sequencing was performed on total RNA extracted from serum sEVs of patients with RB and HCs. More than 50 million raw reads were generated for each sample. The number of raw reads and filtered reads for each sample are shown in Supplementary Figure S2A. After quality filtering, high-quality reads were mapped to the human (hg38) genome. The percentage of mapped reads, average GC content, and sequence length (bp) are listed in Supplementary Table S4. More than 20,000 transcripts were identified in each sample. Among these, 12,744 transcripts were common to all of the samples. Supplementary Figure S2B and Supplementary Table S5 show the unique and common transcripts shared by sEVs. 

sEV RNA Profiling

Supplementary Figure S3A shows the presence of various RNA biotypes and their proportions in sEVs. Protein-coding RNAs (messenger RNAs [mRNAs]) constituted the main fraction (>1500 transcripts) of sEVs, followed by minor fractions of miRNAs, long noncoding RNAS (lncRNAs), and a few transfer RNAs (tRNAs), small nuclear RNAs (snRNAs), and small nucleolar RNAs (snoRNAs). However, most repeated reads (>100 FPKM) corresponded to miRNAs, followed by protein-coding RNAs (40–60 FPKM) and lncRNAs (20–40 FPKM). Supplementary Figure S3B represents the number of reads related to specific RNAs and their distribution in each sample. 

Children and Adult Serum sEVs Showed Distinct Transcriptome Profiles

Figure 1A is the PCA plot of RNA-seq data for serum sEV RNAs of patients with RB and HCs. Adults showed a distinct sEV RNA signature compared to the children. The two groups were separated particularly in the first component, with a 78.2% variance, and clustered exclusively. However, 99% of the data were covered in the first three dimensions, as shown in the scree plot (Fig. 1B), which means that variance in the data (gene expression) across samples was present but not very high. The heat map also shows that children (TR-RB and HCs) had a distinct sEV RNA pattern in comparison to adults (CR-RB and HC) (Fig. 1C). 

Figure 1.

 

(A) PCA plot showing the variation in transcriptome-wide expression profiles of serum sEVs in children and adults: healthy children (C1 and C2), a healthy adult (C3), therapeutic-resistant (TR) retinoblastoma children (RB1 and RB2), and a treatment-responsive adult with completely regressed RB (RB3). (B) PCA scree plot. Cluster 1 (PC1) contributed about 86% of the variance; cluster 2 (PC2), 20%; and cluster 3 (PC3), about 8%. (C) Heat map of 20 downregulated and 20 upregulated genes with top log2 fold change. (D, E) Volcano plots of differentially expressed genes. Green dots indicate downregulated genes, and red dots indicate upregulated genes based on P < 0.05 and log2 fold change of ±2. Gray dots are genes that did not meet these criteria. TR-RB, treatment-resistant active retinoblastoma; CR-RB, completely regressed retinoblastoma after treatment.

Identification of Clinically Relevant Treatment Resistant Genes

To identify genes associated with the TR phenotype, serum sEV RNA profiles were compared between two groups (TR-RB vs. HCs and CR-RB vs. adult HCs). Based on P < 0.05 and log2 fold changes of ≥2 or ≤−2 criteria, 218 genes were upregulated and 273 downregulated in TR versus HC sEVs. However, only 101 genes were upregulated and 83 genes were downregulated in CR-RB versus HC (Figs. 1D, 1E). This indicates that high transcriptomic reprogramming occurred in TR-RB as compared to CR-RB. 

Protein-coding genes such as KAT2B, VWA1, CX3CL1, MLYCD, and NR2F2 were among the top 20 upregulated genes, whereas CYP19A1, KLK3, CYLD, STAG3, and MICAL3 were among the top 20 downregulated genes (Fig. 1C, Table 1) for TR-RB versus HCs. Noncoding RNAs (ncRNAs) such as USP46-AS1, MIR6724-4, and LINC01257 were upregulated, and ZNRD1-AS1, BRE-AS1, MGC12916, and LINC01164 were downregulated for TR-RB versus HCs. However, none of these above-mentioned protein-coding and ncRNAs was dysregulated in CR-RB versus adult HC (Table 2). 

Table 1.

 

Differentially Expressed Protein Coding Genes in the Treatment-Resistant Retinoblastoma Group Versus Healthy Children and Treatment-Responsive Adults With Completely Regressed Retinoblastoma Versus Healthy Adult Siblings

 

Table 2.

 

Differentially Expressed Non-Coding Genes in the Treatment-Resistant Active Retinoblastoma Group Versus Healthy Children and Treatment-Responsive Adults With Completely Regressed Retinoblastoma Versus Healthy Adult Siblings

 

Biological Processes and Signaling Pathways Associated With TR-RB Phenotype

To identify the biological processes (BPs) and pathways associated with TR phenotype, GO and KEGG analyses were performed and compared between DEGs of TR-RB and CR-RB with respect to their HCs using the DAVID database. DEGs corresponding to GO–BP terms GO:0045944: positive regulation of transcription from RNA polymerase II promoter (22 vs. 0) and GO:0000122: negative regulation of transcription and (13 vs. 0) were enriched only for TR-RB. Whereas GO:0007601: visual perception related to DEGs was high for CR-RB (1 vs. 8) (Fig. 2A). GO:0005515: protein binding (172 vs. 78), GO:0046872: metal ion binding (49 vs. 19), GO:0003677: DNA binding (44 vs. 18), and GO:0003700: transcription factor activity (27 vs. 10) were the top enriched GO–MF terms for both groups (Fig. 2B). Most of the DEG genes were found to localize to GO:0005634: nucleus (107 vs. 52), GO:0016021: integral component of the membrane (101 vs. 46), GO:0005654: nucleoplasm (57 vs. 23), and GO:0070062: extracellular exosome (53 vs. 23) (Fig. 2C). The enriched KEGG terms were hsa01100: metabolic pathways (22 vs. 6), hsa04024: cAMP (10 vs. 4), hsa04151: PI3K-Akt (10 vs. 3), cGMP-PKG signaling pathways (8 vs. 3), and hsa04810: regulation of actin cytoskeleton (5 vs. 1). However, hsa04080: neuroactive ligand–receptor interaction (8 vs. 0), hsa05206: micro RNAs in cancer (7 vs. 0), and hsa05016: Huntington's disease (6 vs. 0) were enriched only for DEG genes of TR-RB (Fig. 2D). 

Figure 2.

 

The functional enriched Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) terms for dysregulated genes in RB serum-derived sEVs. (A) Biological processes. (B) Molecular function. (C) Cellular component. (D) KEGG.

Apart from the above-enriched terms, the significant GO and KEGG terms with higher (>5) fold enrichment for TR-RB were related to GO:1902236: negative regulation of endoplasmic reticulum stress-induced intrinsic apoptotic signaling pathway, GO:2001234: negative regulation of apoptotic signaling pathway, GO:0016049: cell growth, and GO:0015992: proton transport. GO:0006342: chromatin silencing, GO:0000186: activation of MAPKK activity, GO:0030512: negative regulation of TGFβ signaling, GO:0002039: P53 binding, and GO:0034599: cellular response to oxidative stress were additional enriched terms associated with TR-RB (Fig. 3). 

Figure 3.

 

Significant Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways identified in circulating sEVs of children with treatment-resistant retinoblastoma compared to age-matched controls.

RB1 Gene Mutation Status and Clinical Phenotype

Out of the three, two patients with RB underwent genetic testing. NGS analysis on the patient with bilateral TR-RB2 revealed the presence of a heterozygous nonsense variation in exon 17 of the RB1 gene (chr13:g.48955432G>A; depth, 25×) that resulted in a stop codon and premature truncation of the protein at codon 516 (p.Trp516Ter; ENST00000267163.4). The variant was classified as pathogenic. For the patient with CR-RB (RB3), the strand multiplex ligation-dependent probe amplification (MLPA) test detected c.(607+1_608-1)_(1695+1_1696-1)del (exon 7-17 deletion), in the RB1 gene (RefSeq id: NM_000321.2). The variant was detected in a heterozygous state and labeled “likely pathogenic.” 

Relative mRNA Expression of CD81, CD9, and CD63 in RB Tumor and Control Retina

To examine the expression pattern of CD81, CD9, and CD63 in RB tumor tissues quantitative reverse transcription–polymerase chain reaction (qRT-PCR) was performed. All three mRNAs were present at low levels in the control retina. Even in the RB tumors, variable expression levels were observed. Out of five tumors, only two showed CD9 expression. CD81 and CD9 levels were significantly higher in two RB tissues with respect to control retina (Supplementary Figs. S4A, S4B), and CD63 levels were high in three tumors (Supplementary Fig. S4C). 

Expression of Exosome-Specific Protein CD9 and RB1 in RB Tumors and Control Retina

The control retina showed diffuse and marked membranous and cytoplasmic CD9 expression in the ganglion cell layer and mild expression in the inner and outer plexiform layer. Two primary enucleated specimens showed variable expression of protein within the tumor tissue. Higher cytoplasmic expression of CD9 was evident in the poorly and undifferentiated tumor areas, areas of gliosis, apoptotic cells, retinocytomatous areas, and at the invasion site. Optic nerve and endothelial cells also showed diffuse and marked expression of the CD9. The two TR-RB cases were positive for both CD9 and RB1 expression. One TR-RB1 specimen showed patchy and marked expression for CD9 and RB1 antibodies. However, the TR-RB2 case showed no expression (Fig. 4). In this case, genetic testing revealed the presence of a pathogenic variant in the RB1 gene. 

Figure 4.

 

CD9 expression in RB. (A) Control (skin). (B) Control retina. (C–E) RB tumor cells showing marked, moderate, and mild expression (original magnification, 30×). DAB, diaminobenzidine chromogen. (F) Retinocytoma area showing marked membranous expression. (G, H) Optic nerve invasion; membranous expression of CD9 in tumor cells infiltrating laminar cribrosa (DAB; original magnification, 10× and 30×). (I) Choroidal invasion by tumor cells along with few choroidal melanocytes; RB1 expression in RB. (J) Control (brain). (K) Nuclear expression in tumor (RB1 retained). (L) Negative expression in tumor (loss of expression) (DAB; original magnification, 30×).

Discussion

Treatment resistance remains a persistent challenge for RB management. In the past, several groups have explored the chemoresistance mechanisms in RB by developing drug-resistant clones from original RB cell lines and primary cell cultures established from enucleated specimens.^22–25 However, only a few gene expression studies have used either tumor or liquid biopsy samples from patients with RB. As EVs reflect the cell of origin at particular physiological and pathological conditions, they offer real-time monitoring of the tumor burden and response to treatment.^26–28 The present study comprehensively analyzed the serum sEVs from two patients with recurrent RB (TR-RB) and one patient who had been previously treated with completely regressed RB and a history of good treatment response (CR-RB). It was observed that higher transcriptomic reprogramming in the serum sEVs of TR-RB (492 DEG vs. 184 DEG genes). Neuroactive ligand-receptor interaction signaling pathway, Huntington's disease, and miRNAs in cancer are enriched only for DEGs of TR-RB. In addition, the number of DEG genes enriched for cAMP, PI3K-AKT, cGMP-PKG, and calcium signaling KEGG pathways in TR-RB are high compared to the sEVs of CR-RB. Also, TR-RB sEVs are functionally enriched with biological processes related to negative regulation of apoptotic process, GO:0007186∼G-protein coupled receptor signaling pathway and GO:0006351∼transcription. 

Previous microarray and bioinformatics analysis on RB tumor specimens revealed that the above enriched terms were associated with RB tumor occurrence, development, and chemo-resistance by regulating cell growth, cell survival, and proliferation.^22,29,30 Several studies have demonstrated that the neuroactive ligand–receptor interaction pathway is linked to a poor prognosis in various cancers, and breast cancer patients with low expression of cannabinoid receptor 1 (CNR1, part of this pathway) may benefit from chemotherapy.³¹ Bioinformatics analysis of gene expression data of RB tumor tissues compared to normal retina revealed that the aberrantly expressed protein coding genes and target genes for DE lncRNAs are enriched for neuroactive ligand–receptor interaction.^29,30 Moreover, this is one of the 50 pathways significantly enriched for DEGs between chemoresistance Y79 versus parental Y79 cells.³² MicroRNAs in cancer were enriched only for DE genes of TR-RB. Consistent with our data, numerous studies have pointed out that abnormal expression of miRNAs is linked to invasiveness and the drug resistance of tumor cells, and regulation of these RNAs sensitizes the drug resistance of cancer stem cells to chemotherapy.³³ In RB, several tumor suppressors (miR-214-3p, miR-34A) and oncogenic miRNAs (miR-222) have been identified. They affect the chemoresistance of RB cells by regulating MRPs, autophagy, and apoptotic genes.^34–36 Similar to our findings, the downregulated miRNAs mir124 and miRlet-7 in RB sEVs are also found to be underexpressed in RB tumors.³⁷ 

KAT2B, NR2F2, CX3CL1, VWAI, CD248, CYP19A1, and KLK3 are some of the top aberrantly expressed protein-coding genes, and USP46-AS1, miR6724-4, LINC01257, ZNRD1-AS1, BRE-AS1, MGC12916, and LINC01164 are the non-coding RNAs in TR-RB. However, these genes remained the same in the CR-RB versus control groups. Some of these genes were found to play a key role in epithelial-to-mesenchymal transition, tumor cell invasion, migration, and angiogenesis in various cancers.^38–40 Consistent with our findings on serum sEV expression profiles, the levels of KAT2B, WEE1, and C1ORF226 were elevated and CYLD was downregulated in RB tumors compared to normal retina.⁴¹ The oncogenic role of ZNRDS1-S1 was studied in RB. It acts as a “sponge” for miR-128-3p, increasing BMI1 levels and promoting RB progression.⁴² 

Metabolic reprogramming, one of the known hallmarks of cancer,⁴³ is clearly evident in RB sEVs, and it is more prominent in TR-RB as compared to CR-RB sEVs. Genes involved in glucose metabolism (G6PD, ALDH1L2) and fatty acid metabolism (ELOVL5, MLYCD) were aberrantly expressed in TR-RB sEVs. However, the expression status of these genes remained unaltered in CR-RB. In addition to the sorting and transmembrane functions, the oncogenic role of exosome marker proteins CD9, CD63, and CD81 has been documented in various cancers.⁴⁴ However, their expression profile in RB tumors has not yet been deciphered. The qRT-PCR results revealed that mRNA levels of CD9, CD63, and CD81 were significantly higher compared to the control retina, but varied mRNA levels were noted across the tumors. Nonetheless, to determine their individual functions in RB progression, more tumor samples should be tested for their expression patterns. The study also measured the CD9 and RB1 protein expression for two TR-RB tumor samples. TR-RB1 (unilateral RB) showed marked expression of CD9 and partial RB1 expression. In TR-RB2 (bilateral RB), the RB1 genetic testing report revealed the presence of a pathogenic variant at exon 17 of the RB1 gene (chr13:g.48955432G>A). For this case, both RB1 and CD9 proteins showed no expression. The pathogenic variant c.(607+1_608-1)_(1695+1_1696-1)del (exon 7-17 deletion in the RB1 gene) was identified for a bilateral RB3 (CR-RB) adult. Previous studies demonstrated that RB risk is associated with the germline pathogenic variant and with maintenance of RB protein.^45,46 However, from our own experience, we found that the pathogenicity of a genetic variant and clinical response of the RB are not correlated.⁴⁶ 

In general, the clinical response of patients with RB varies greatly. Several factors such as patient age, laterality, presence of pathogenic RB1 genetic variants, and individual responses to therapy influence the treatment outcome. It is quite challenging to predict the tumor relapse during or after RB management. This proof-of-concept study could provide a basis for analyzing large numbers of patient-derived EV samples to understand TR mechanisms in RB. 

The limitations of the present study include small sample size, possible individual variability of EV profiles leading to differences between two samples, and the possible influence of chemotherapy on EV profiles. 

Conclusions

Apparent differences in the serum sEV transcriptome profiles of TR-RB and CR-RB require further investigation in larger samples to arrive at a definitive conclusion. Regulation of apoptosis, cell growth, proton transport genes, neuroactive ligand–receptor interactions, and miRNAs in cancer are enriched for DEGs of TR-RB, suggesting their role in TR mechanisms. We speculate that serum sEVs could serve as a potential liquid biopsy source for understanding TR mechanisms in RB and could possibly pave the way for prognostication and responsiveness to treatment. 

Acknowledgments

The authors thank Roli Budhwar, PhD, and Jeffrey Godwin of Bionivid Technology Private Limited for help with the RNA sequencing and bioinformatics analysis. 

Supported by Hyderabad Eye Research Foundation, LV Prasad Eye Institute, and CSIR-Government of India (SERB grant CRG/2021/003236). 

Disclosure: R. Manukonda, None; S. Jakati, None; J. Attem, None; D.K. Mishra, None; T.R. Mocherla, None; M.M. Reddy, None; K. Gulati, None; K.M. Poluri, None; G.K. Vemuganti, None; S. Kaliki, None