Graphical abstract

   graphic file with name fx1.jpg
   [47]Open in a new tab

Highlights

     * •
       Native and cultured RPESC-RPE monolayers are heterogeneous in form
       and function
     * •
       CITE-seq analysis has revealed cell surface markers of each RPE
       subpopulation
     * •
       RPE subpopulations have spatially distinct localizations in the eye
     * •
       RPE subpopulations demonstrate unique functional capabilities
     __________________________________________________________________

   Farjood et al. has set the stage for characterizing RPE heterogeneity
   by using single-cell sequencing technology to identify cell surface
   markers compatible with isolating each subpopulation for further study.
   Using these markers, select RPE subpopulations are shown to have unique
   localization within the eye and to have distinct functional
   capabilities.

Introduction

   The retinal pigment epithelium (RPE) supports neural retinal
   photoreceptor cells by providing nourishment, phagocytosing
   photoreceptor cell outer segments, and other vital functions. The RPE
   has traditionally been considered a homogenous monolayer composed of
   one cell type, but more recently accumulating evidence points toward
   significant RPE cell diversity in morphology, gene expression, and
   function ([48]Boulton et al., 1994; [49]Farjood et al., 2023;
   [50]Mullin et al., 2023; [51]Ortolan et al., 2022; [52]Xu et al.,
   2021). Understanding human RPE subpopulation diversity and associated
   functional and morphological specialization provides insights to
   improve understanding of retinal health and disease.

   Transcriptional, proteomic, and morphological analysis of human RPE
   cells in macular versus peripheral anatomical retinal regions has
   revealed significant spatial heterogeneity within the RPE monolayer
   ([53]Bhatia et al., 2016; [54]Ortolan et al., 2022; [55]Skeie and
   Mahajan, 2014; [56]Voigt et al., 2019; [57]Whitmore et al., 2014;
   [58]Xu et al., 2021). RPE cells localized in the macular region are
   narrower and express lower levels of BEST1 and higher levels of
   complement proteins ([59]Gao and Hollyfield, 1992; [60]Skeie and
   Mahajan, 2014; [61]Whitmore et al., 2014). Further evidence of RPE cell
   heterogeneity comes from our work demonstrating that a minor
   subpopulation of adult human RPE cells has the potential for extensive
   in vitro proliferation and self-renewal, defining the tissue-specific
   RPE stem cell (RPESC) ([62]Saini et al., 2016; [63]Salero et al.,
   2012).

   Single-cell RNA sequencing (scRNA-seq) has improved our ability to
   detect cellular heterogeneity in retinal tissue ([64]Li et al., 2023;
   [65]Yan et al., 2020). This approach enabled the discovery of new
   ganglion cell subtypes in the mouse retina, ([66]Tran et al., 2019)
   macroglial and photoreceptor cell subpopulations in the human retina
   ([67]Lukowski et al., 2019; [68]Menon et al., 2019), and the
   identification of diverse cell populations, including endothelial
   cells, macrophages, and RPE, from macular and peripheral regions of
   healthy and age-related macular degeneration (AMD) samples ([69]Voigt
   et al., 2019). Most studies of RPE cell subpopulations utilized both
   choroidal and RPE cells, and the RPE cells identified were often
   considered a single homogeneous cell population among a multitude of
   diverse cell types ([70]Collin et al., 2023; [71]Huang et al., 2023;
   [72]Monavarfeshani et al., 2023; [73]Voigt et al., 2019). A few studies
   focused on the native RPE cells isolated directly from the human eye
   ([74]Mullin et al., 2023; [75]Xu et al., 2021) that revealed diverse
   RPE subpopulations with unique transcriptional and chromatin profiles
   ([76]Mullin et al., 2023; [77]Xu et al., 2021). These prior studies did
   not assess whether transcriptomic heterogeneity is associated with
   functional specialization nor whether the diverse RPE cells represented
   distinct subtypes or different labile cell states.

   Cellular indexing of transcriptomes and epitopes by sequencing
   (CITE-seq) ([78]Stoeckius et al., 2017) has significantly advanced our
   ability to uncover cellular heterogeneity by enabling the simultaneous
   measurement of surface protein expression and transcriptomic profiles
   within individual cells. This technique allows for deeper understanding
   of cellular diversity and identification of previously unrecognized
   subpopulations ([79]England et al., 2020; [80]Liu et al., 2023b;
   [81]Pombo Antunes et al., 2021; [82]Wu et al., 2021). The
   identification of surface markers enables cell subtypes to be enriched
   by sorting for functional studies. Here, we describe the first CITE-seq
   study focused on the adult human RPE acutely isolated from the adult
   human eye (native RPE) to better characterize RPE heterogeneity. This
   approach allowed us to comprehensively dissect RPE subpopulations and
   identify surface proteins that enabled RPE subpopulation enrichment and
   functional assessment. Multimodal analysis revealed a remarkable
   heterogeneity of 12 RPE cell identities, each with distinct gene and
   surface protein expression signatures, indicating that multiple,
   previously unrecognized cell subpopulations are present in the native
   RPE layer. We then studied the spatial distribution and morphological
   heterogeneity of these subpopulations within the human eye. We further
   demonstrated that RPE cells derived from native adult RPE after several
   weeks in culture recapitulate the subpopulations found in native RPE
   tissue. This supports the use of cultured adult RPE cells for in vitro
   modeling and as a cell source for replacing the RPE cells that are lost
   in retinal degenerative diseases.

Results

Native human RPE subpopulations have distinct surface protein expression

   The experimental approach is summarized in [83]Figure 1A. Three human
   donor eyes (two males and one female) were obtained from registered eye
   banks, and RPE cells were dissociated for analysis. Utilizing a
   TotalSeq-A antibody cocktail, acutely isolated RPE cells were labeled
   with 156 uniquely barcoded antibodies, followed by single-cell library
   preparation using the ICELL8 system. After sequencing, transcript reads
   were mapped to the human genome (hg38), and antibody-derived tag (ADT)
   reads were mapped to an antibody/barcode sequence list. The ADT counts
   were preprocessed by arcsinh transformation and normalization. The
   final dataset consisted of 3,706 cells with an average read depth of
   269,448 for transcriptomic data and 291,364 for the surfaceome profile.
   Dimensionality reduction, weighted nearest neighbor analysis
   (contributions of each modality can be found in [84]Figure S1), and
   clustering were performed using the Seurat package in R.

Figure 1.

   [85]Figure 1
   [86]Open in a new tab

   CITE-seq results revealed native RPE heterogeneity (n = 3, biological
   replicates)

   (A) Schematic of the CITE-seq study.

   (B) UMAP figure visualization of human RPE subpopulations clustered
   using WNN analysis.

   (C) Expression of RPE-specific markers in all clusters (log-normalized
   gene expression).

   (D) Gene markers of RPE clusters (z-normalized log1p gene expression).

   (E) Surface protein markers enriched in each cluster (z-normalized
   transformed ADT expression).

   (F) Selected parent GO categories from GO enrichment and semantic
   similarity analysis of RNA markers from each cluster.

   After clustering, seven distinct RPE subpopulations were identified and
   designated A0 through A6 ([87]Figure 1B). To verify the RPE identity of
   each cluster, we examined the expression of 171 genes designated to be
   RPE signature genes by bulk RNA sequencing (RNA-seq) analysis of
   acutely isolated RPE cells ([88]Bennis et al., 2015; [89]Strunnikova et
   al., 2010). One hundred forty of the signature genes were expressed by
   the primary RPE cells, with 95 shared between all clusters and each
   cluster expressing at least 101 signature genes ([90]Figure 1C;
   [91]Table S1), verifying the RPE identity of these subpopulations. Our
   results revealed unique gene and surface protein expression profiles or
   molecular signatures for each cluster ([92]Figures 1D and 1E;
   [93]Tables S2 and [94]S3). For example, cluster A1 exhibited higher
   expression of HTRA1, which has been associated with an increased risk
   of AMD ([95]Lin et al., 2018) and of the long noncoding RNA Linc01833,
   a biomarker of immune infiltration that is expressed in both cultured
   and native human RPE cells ([96]Liu et al., 2023a; [97]Postnikova et
   al., 2019). Clusters A2 and A3 showed higher expression of PLIN2, a
   gene responsive to oxidative stress and involved in the formation of
   lipid droplets by RPE cells, which is thought to contribute to drusen
   formation ([98]Chuang et al., 2022; [99]Grubaugh et al., 2023;
   [100]Meyer et al., 2019). Cluster A4 displayed higher expression of
   HNRNPH3, an RNA splicing protein that is linked to an increased risk of
   AMD ([101]Lin et al., 2018). Cluster A5 demonstrated enriched
   expression of a distinct set of genes, including SLC39A6, a zinc
   transporter, ([102]Chowanadisai et al., 2008) and SIX3, a regulator of
   Wnt/β-catenin signaling that maintains neuro-retinal progenitors
   ([103]Diacou et al., 2018). Cluster A6 exhibited higher levels of
   RB1CC1, essential for the autophagic activity of RPE cells ([104]Yao et
   al., 2015).

   We performed a Gene Ontology (GO) enrichment of the genes associated
   with each cluster that revealed specific functions were enriched
   ([105]Figure 1F; [106]Table S4). Clusters A0 and A4 did not demonstrate
   enrichment for any categories. Cluster A1 is enriched for metabolic
   processes and organization. Clusters A5 and A2 have the most enriched
   terms (937 and 598, respectively). While both clusters are enriched for
   the parent term “cell cycle,” cluster A2 is enriched for “negative
   regulation of cell cycle,” whereas A5 is enriched for the “positive
   regulation of cell cycle.” Cluster A5 is enriched for categories
   associated with “cell division,” whereas cluster A2 shows no such
   enrichment. These clusters also differ in categories related to
   “defense response” with cluster A2 showing enrichment in associated
   categories and cluster A5 demonstrating no enrichment. Like cluster A1,
   cluster A3 demonstrated enrichment for a variety of metabolic
   processes; however, A3 also showed enrichment for processes related to
   development and response to the environment. Finally, cluster A6 was
   primarily enriched for categories associated with “generation of
   precursor metabolites and energy.” Overall, the primary RPE clusters
   showed wide enrichment for different metabolic pathways; however,
   clusters A1, A2, A3, and A5 demonstrated enrichment for a broad array
   of non-metabolic functions.

   The CITE-seq data identified distinct surface protein markers for each
   cluster ([107]Figure 1E). While all clusters expressed RPE-specific
   markers ([108]Figure 1C), cluster A1 exhibited higher expression of
   PTPRC (CD45), a hematopoietic marker known to be expressed in RPE cells
   ([109]Limb et al., 1997), NCR1, a marker of natural killer cells
   ([110]Moretta et al., 2001), and CD24, a neural progenitor cell marker
   that regulates autophagy ([111]Lu et al., 2018; [112]Robson et al.,
   2019; [113]Sun et al., 2020). Clusters A2 and A3 had elevated
   expression of B3GAT1, a gene reduced in the RPE-choroid layers of AMD
   eyes ([114]Newman et al., 2012; [115]Wang et al., 2022). Cluster A3 had
   elevated levels of CD58, TNFRSF14, ITGA1, and CD55. CD58 and TNFRSF14
   are involved in the immune response, while CD55 has a role in
   complement inhibition in RPE cells ([116]Kim and Song, 2006; [117]Ma et
   al., 2010; [118]Yang et al., 2009). Cluster A3 expressed ITGA1, a gene
   involved in transforming growth factor beta (TGF-β)-induced
   epithelial-mesenchymal transition (EMT) ([119]Gharibi et al., 2017).
   Cluster A4 showed significantly higher levels of the chemokine
   receptor, CCR5, which is linked to geographic atrophy ([120]Krogh
   Nielsen et al., 2020; [121]Nagineni et al., 2015). Cluster A5 was
   enriched with transferrin receptor (TFRC), associated with iron uptake
   ([122]Hentze et al., 2004). Cluster A6 expressed a diverse range of
   surface proteins and uniquely exhibited high levels of the lectin
   SIGLEC7, an immune receptor, and KIR3DL1, which binds HLA (human
   leukocyte antigen) molecules ([123]Boudreau et al., 2016; [124]Stewart
   et al., 2024).

   Most of the surface markers detected in this CITE-seq analysis had not
   been previously reported in RPE cells. Hence, we evaluated their
   expression in human RPE tissue. We analyzed B3GAT1 (clusters A2, A3,
   and A6), CD24 (cluster A6), and TNFRSF14 (clusters A3 and A6).
   Immunostaining of human RPE tissue revealed positive expression for all
   three markers ([125]Figure 2). We then investigated the spatial
   distribution of these markers in the human RPE layer. We used an
   established anatomic classification system that categorizes RPE cells
   into five concentric regions, ranging from the macular (P1) to
   periphery (P5) ([126]Figures 2A and 2B) ([127]Ortolan et al., 2022).
   Quantitative analysis revealed that B3GAT1 was expressed in cells
   across all five regions with higher frequency of expression in P2 and
   P5. CD24 expression was detected in regions P3, P4, and P5, while
   TNFRSF14 was exclusively localized to region P5. These results confirm
   the expression of these CITE-seq markers in human RPE in situ with
   distinct regional distributions.

Figure 2.

   [128]Figure 2
   [129]Open in a new tab

   Immunofluorescence imaging of RPE flat mounts prepared from a human
   donor eye revealed subpopulations in human RPE monolayer

   (A) Schematics of human RPE showing spatial distribution of B3GAT1-,
   CD24-, and TNFRSF14-expressing cells in five concentric,
   morphologically different regions of the RPE described in a previous
   study ([130]Ortolan et al., 2022).

   (B) Immunofluorescence images showing expression of B3GAT1, CD24,
   TNFRSF14, and ITGB3 in subpopulations of adult RPE cells in different
   regions of the human RPE tissue. Scale bars, 50 µm.

   (C) Boxplots showing the area of B3GAT1-, CD24-, and
   TNFRSF14-expressing cells in different regions of the human RPE (n = 3,
   biological replicates, independent experiments).

   (D) Density plots showing the cell area distribution for RPE
   subpopulations. B3GAT1-expressing cells in P1 had a larger area than
   B3GAT1 cells, while in P5 B3GAT1+ cells were smaller than the negative
   population. CD24- and TNFRSF14 expressing cells were found to be
   smaller than their non-expressing counterparts (n = 3, biological
   replicates, independent experiments).

   Previous studies demonstrated a difference in RPE cell size based on
   location within the eye, with smaller cells in the P1 region and larger
   cells in the P5 region ([131]Ortolan et al., 2022). We found B3GAT1-,
   CD24-, or TNFRSF14-positive cells in the different zones of the native
   tissue had different sizes ([132]Figures 2C and 2D). B3GAT1+ cells
   within P1 or P5 show similar size distribution. However, B3GAT1+ cells
   were larger than their B3GAT1-neighbors in P1 but smaller than their
   B3GAT1-neighbors in P5. CD24^+ cells showed a similar size distribution
   in both P3 and P5; however, CD24^+ cells were smaller than the CD24^−
   cells. TNFRSF14+ cells in the P5 zone were smaller than their negative
   neighbors. These results underscore the unique distribution and
   morphology of cells bearing the three markers. Moreover, these
   observations indicate that the cluster A6 subpopulation (marked by
   TNFRSF14, B3GAT1, and CD24) is found in the P5 region and is smaller
   than the other cells in this region.

Extensively cultured adult RPE cells maintain distinct subpopulation
identities

   To investigate whether adult RPE subpopulations are recapitulated in
   RPE cells derived from the proliferative RPE subpopulation, we
   performed CITE-seq experiments on passage 2 (P2) adult human RPE cells
   ([133]Figure 3A). These cells had been cultured, passaged, frozen, then
   thawed to create P2 cells that were cultured for an additional 2, 4, or
   10 weeks. After sequencing and QC, we obtained 2,586, 3,354, and 2431
   cells in the 2W, 4W, and 10W cultures, respectively. The cultured cell
   CITE-seq data and the 3,706 primary RPE cell CITE-seq data were merged
   using canonical correlation analysis and clustered using the Seurat
   package in R. After merging the datasets, the average read depth was
   221,231 per cell for the transcriptomic profiles and 272,903 per cell
   for the surfaceome profile. Due to the increased number of cells in the
   combined dataset, we were able to identify 12 RPE subpopulations,
   including ones not in the primary cell data alone ([134]Figure 3B).

Figure 3.

   [135]Figure 3
   [136]Open in a new tab

   Analysis of CITE-seq results from both primary (n = 3, biological
   replicates) and adult cultured RPE (n = 3, biological replicates)
   reveals new RPE subpopulations

   (A) Schematic of the CITE-seq study.

   (B) UMAP figure visualization of combined native and cultured
   subpopulations clustered using WNN analysis.

   (C) River plot illustrating the contribution of the native RPE clusters
   to the integrated data clusters.

   (D) Surface protein markers of RPE clusters (z-normalized transformed
   ADT expression).

   (E) UMAP figure showing distribution of cells from different time
   points.

   (F) Proportional bar chart showing the distribution of RPE clusters
   across RPE source (native or cultured for 2,4,10 week).

   (G) Stacked bar plot illustrating the cell proportions by time points
   within RPE clusters.

   (H) Bar chart showing the portion of cells from each time within RPE
   clusters.

   (I) Surface protein markers enriched in different time points
   (z-normalized transformed ADT expression).

   (J) Changes in the expression of RPE-specific genes in native RPE and
   cultured RPE from 2W to 10W (z-normalized log1p gene expression).

   Each of the 12 RPE cell clusters exhibited expression of at least 110
   RPE signature genes, with 91 genes shared across all clusters and 140
   RPE signature genes in the combined dataset. Clusters 7, 8, and 11
   displayed the highest number of RPE signature genes (129, 128, and 126,
   respectively), while cluster 4 exhibited the lowest number ([137]Table
   S5). This analysis verified the RPE identity of all native and cultured
   populations. Most of the clusters found in primary RPE cells were
   preserved in the combined dataset, showing preservation native RPE
   subpopulations in cultured RPE preparations ([138]Figure 3C), albeit
   with differences in the individual subpopulation contributions to the
   whole. Analyzing the merged CITE-seq data identified unique gene and
   surface markers for each cluster ([139]Tables S6 and [140]S7) similar
   to that found in native RPE subpopulations ([141]Figures 1E and
   [142]3D).

   We quantified the contribution of each RPE source (native RPE and P2
   RPE cells cultured for 2, 4, or 10 weeks) to each cluster ([143]Figures
   3E–3H). Except for cluster 10, which contained mainly native cells, all
   clusters contained cells from both native and cultured RPE
   preparations, although the percentages varied. The predominant
   composition of clusters 1, 7, 8, 10, and 11 are native RPE cells, while
   clusters 0 and 4 are mainly comprised of cultured RPE cells (95%;
   [144]Figure 3G). Similar percentages of cells were present regardless
   of time in clusters 2, 7, and 5 ([145]Figure 3H). Overall, the cultured
   and native RPE demonstrate a similar subpopulation composition.

RPE exhibits unique surfaceome and morphological profiles between
subpopulations

   Cultured and native RPE showed a distinct surface protein signature by
   CITE-seq ([146]Figure 3I). Forty-three of the probed surface proteins
   showed differential expression across native and P2 cultured cells
   ([147]Table S8). Both native and cultured RPE cells had a higher
   expression of different subsets of proteins when compared ([148]Figure
   3I). RPE-4W cells had a unique gene and surface protein signature
   ([149]Tables S8 and [150]S9) with elevated expression of CD47, CD58,
   NT5E, B3GAT1, ITGA1, CD44, CD55, and TNFRSF14 ([151]Figure 3I).
   Interestingly, the expression levels of all enriched RPE-4W markers
   decreased at 10 weeks as the cells matured, reaching levels like those
   observed in primary cells ([152]Figure 3I). We also observed an
   increase in the expression of RPE signature genes from 2 to 10 weeks in
   cultured RPE cells ([153]Figure 3J), further suggesting that in vitro
   maturation of RPE cells occurs during time in culture. We next
   identified surface markers that could be used to enrich RPE
   subpopulations, for example, using magnetic or FACS-based isolation
   methods ([154]Figure 3D). For instance, CSF1R or ITGB3 can be used to
   isolate cluster 6 or 9 subpopulations, respectively ([155]Figure S2).

   Given our finding of surface markers that distinguish different
   clusters, our next goal was to confirm these findings through
   immunofluorescence microscopy of cultured RPE cells. Immunostaining was
   performed on P2-cultured RPE-4W cells, targeting B3GAT1, CD24, and
   TNFRSF14, and the cluster 9 marker, ITGB3 ([156]Figure 4A). CITE-seq
   showed B3GAT1, CD24, TNFRSF14, and ITGB3 were expressed in 77.8%,
   91.6%, 50.7%, and 4.7% of the RPE-4W cells, respectively. Analysis
   revealed similar proportions of positive cells for TNFRSF14 and ITGB3
   populations (55.6% and 6.8%, respectively), with fewer B3GAT1 and
   CD24-expressing cells than predicted by the CITE-seq data (36% and 52%,
   respectively), possibly due to a lower sensitivity of
   immunocytochemistry ([157]Stoeckius et al., 2017).

Figure 4.

   [158]Figure 4
   [159]Open in a new tab

   Immunofluorescence imaging of RPE-4W cultures shows expression of
   cultured RPE surface markers

   (A) Subpopulations of RPE cells were detected in RPE-4W cells
   expressing B3GAT1, CD24, and TNFRSF14 surface proteins that were
   observed to be upregulated in RPE-4W cells according to CITE-seq
   analysis ([160]Figure 1F) (n = 3, biological replicates, concurrent
   experiments). ITGB3, a marker of cluster 9, was also found to be
   expressed in RPE-4W cultures. Scale bar, 50 µm.

   (B) Morphological examination of cells expressing B3GAT1, CD24, and
   ITGB3 revealed distinct characteristics. CD24^+ and ITGB3+ cells
   exhibited larger cellular areas, whereas TNFRSF14+ cells displayed a
   smaller area compared to their negative counterparts. ITGB3-positive
   cells demonstrated the largest cellular area among the four
   subpopulations.

   Morphological analysis revealed that CD24^+ and ITGB3+ cells were
   larger, while TNFRSF14+ cells were smaller than their negative
   counterparts ([161]Figure 4B). In RPE-4W cultures, B3GAT1+ and
   TNFRSF14+ cells were smaller compared to native RPE cells, whereas
   CD24^+ and ITGB3+ cells were larger. Expression frequency comparisons
   showed that B3GAT1 expression increased from 36% in native RPE to 78%
   in RPE-4W cells, CD24 increased from 23% to 92%, and TNFRSF14 increased
   from 17% to 51%. ITGB3 expression remained stable, with 4% in native
   RPE and 4.7% in RPE-4W cells. Immunofluorescence imaging revealed a
   small ITGB3+ subpopulation (∼1.6%) within the P3 region of native RPE
   tissue ([162]Figure 5A).

Figure 5.

   [163]Figure 5
   [164]Open in a new tab

   Gene Ontology (GO) enrichment analysis reveals RPE subpopulation
   functional specialization

   (A) Expression of ITGB3 in human RPE tissue was confirmed through
   immunofluorescence imaging of a human RPE flat mount. The spatial
   distribution of ITGB3+ cells was analyzed as described in [165]Figure
   2. (n = 3, biological replicates, independent experiments) (B) GO
   enrichment and semantic similarity analysis of RPE subpopulations. The
   dot size reflects the percentage of genes in each category expressed by
   the cluster, and the y axis represents the log of the false discovery
   rate (FDR) for the enrichment of each category.

   (C) Representative images of automated cell tracking results showing
   cells with low- and high-migratory capacity.

   (D) Density plots showing the distribution of migration distance for
   cells expressing markers of clusters 6 and 9, CSF1R and ITGB3,
   respectively, in native RPE and RPE-4W cells. CSF1R-expressing cells
   had a lower migration distance compared to CSF1R− cells in both native
   and RPE-4W cultures. ITGB3+ cells of the native and RPE-4W cultures
   traveled a longer distance compared to their negative counterparts. (n
   = 3, biological replicates, independent experiments).

Functional diversity among RPE subpopulations

   To gain insight into the functional diversity among molecularly
   distinct RPE subpopulations, we identified the gene expression
   associated with each cluster using Seurat’s FindAllMarkers function
   ([166]Table S6) and performed GO pathway enrichment analysis utilizing
   the identified genes ([167]Figure 5B). We employed a semantic
   similarity approach provided by the “screp” package
   ([168]https://github.com/neural-stem-cell-institute/screp) in R
   ([169]Table S10). Our analysis revealed significant diversity in
   pathway enrichment across RPE subpopulations. Clusters 0, 3, and 9
   expressed more in cultured cells exhibited highly significant
   enrichment for pathways associated with cell-cycle regulation and cell
   division. Despite the lower expression levels of RPE-specific markers
   within these subpopulations, likely indicative of a less mature
   differentiation state, the subpopulations remained enriched for
   pathways specific to fundamental RPE functions, such as those involved
   in pigmentation and cell polarity ([170]Figure 5B).

   Clusters 1 and 10 exhibited enrichment in pathways associated with
   metabolic processes, stress response, and homeostasis pathways.
   Clusters 2 and 3 had enrichment for cell cycle, migration, cytoskeleton
   organization, and response to stress. Cluster 5 was enriched for
   pathways “response to endogenous stimulation” and “homeostatic
   processes.” Clusters 7, 8, and 11 showed enrichment for “sensory
   perception of light stimulus” and “vascular process in circulatory
   system.” Cluster 9, marked by expression of ITGB3, exhibited
   predominant and substantial enrichment in pathways associated with cell
   morphogenesis, migration, locomotion, and regulation of organelle
   organization, suggesting a potential emphasis on cell motility within
   this cluster. ITGB3+ RPE cell motility was confirmed using an in vitro
   functional assay.

ITGB3+ cells are more migratory than CSFR+ RPE cells

   Due to autofluorescence of native RPE cells, we employed
   magnetic-activated cell sorting (MACS) to separate ITGB3+ (cluster 9)
   and CSF1R+ (cluster 6) cells in native RPE and RPE-4W cells. Each
   subpopulation was separately cultured under time-lapse microscopy over
   24 h. Time-lapse videos were analyzed to quantify migration using the
   Bayesian Tracker (btrack) package in Python ([171]Figure 5C,
   supplementary information 1, [172]Video S1) ([173]Ulicna et al., 2021).
   Video S1. Example cell tracking results using the btrack package
   [174]Download video file^ (858.9KB, mp4)

   CSF1R+ cells exhibited a slightly lower migratory capacity and a slower
   migratory rate compared to CSF1R− cells in both native RPE and cultured
   RPE-4W. However, native CSF1R+ RPE cells migrated longer distances
   compared to their cultured counterparts ([175]Figure 5D). The motility
   of ITGB3+ cells in both native and RPE-4W showed a bimodal
   distribution, indicating two distinct migratory potentials within the
   cluster 9 subpopulation. The larger population demonstrated high
   motility, with a peak migration distance of approximately 150 μm, while
   the smaller population exhibited lower motility, with a peak migration
   distance of around 12 μm. Hence the cluster 9 cells, which were
   enriched for markers of cell motility, demonstrated increased
   migration. These data for the first time demonstrates a specialized
   cellular function associated with a distinct isolated RPE
   subpopulation.

Discussion

   RPE cell heterogeneity is increasingly recognized as a critical aspect
   of retinal physiology and pathology. Recent studies have identified
   multiple RPE subpopulations by transcriptome analysis, but a full
   understanding of the functionally distinct RPE subpopulations is still
   lacking ([176]Collin et al., 2023; [177]Huang et al., 2023; [178]Mullin
   et al., 2023; [179]Ortolan et al., 2022; [180]Voigt et al., 2019;
   [181]Xu et al., 2021). Our approach builds on previous work identifying
   cell surface proteins along with the transcriptional signatures of
   individual RPE cells. This enables us to identify and enrich different
   subpopulations using surface marker expression to assess morphological
   and functional specialization. We established seven subpopulations of
   native RPE cells with diverse morphology and function that are
   distributed in different regions of the RPE layer. Notably, these
   differences are propagated over long-term culture, consistent with the
   RPE cells being true subpopulations rather than labile states. Our
   findings highlight the value of cultured adult RPESC-RPE cells as an in
   vitro model of the native RPE tissue and as a source of cells to
   replace native RPE upon transplantation.

   Prior studies demonstrated that RPE cells show morphological
   differences in the diseased state, during aging, and across different
   regions of the eye ([182]Ortolan et al., 2022; [183]Rashid et al.,
   2016). Previous studies associated variations in cell size with
   differences in mitochondrial function and metabolic activity
   ([184]Miettinen and Bjorklund, 2017; [185]Seel et al., 2023). Our
   morphological analysis revealed size differences among subpopulations
   in both native and cultured RPE cells. We identified an ITGB3+
   subpopulation (cluster 9) characterized by a larger cell size and
   greater migratory capacity compared to ITGB3− cells. Notably, the
   migratory capacity of ITGB3+ cells remained consistent from native to
   long-term cultured RPE, demonstrating the stability of this functional
   specialization in vitro.

   Our mapping of diverse RPE subpopulations to different regions of the
   RPE layer, combined with previous studies showing heterogeneous
   responses of RPE cells to disease processes ([186]Ach et al., 2014;
   [187]Ortolan et al., 2022; [188]Rashid et al., 2016), suggests that
   specific RPE subpopulations have different susceptibilities to retinal
   disorders. We found that in native RPE subpopulations cluster A1
   expresses multiple AMD-related genes, cluster A3 expresses genes
   involved in lipid droplet accumulation in the RPE ([189]Hara et al.,
   2023), and cluster A6 expresses genes involved in autophagy ([190]Yao
   et al., 2015). Clusters A6 and A3 are both associated with the outer
   RPE layer. We speculate that defects in autophagy and increases in
   lipid accumulation due to changes in the A3 and A6 subpopulations in
   the central retina may be due to subpopulation-specific pathologies
   that contribute to AMD pathogenesis ([191]Wei et al., 2023). Hence, RPE
   cell subpopulation distribution may explain why retinal disease
   progression occurs at variable rates in different regions of the RPE
   layer, such as the macular region.

   Promising vision improvements have been obtained in preclinical and
   early clinical studies using human RPE cell products to replace
   degenerated RPE tissue ([192]Cho et al., 2019; [193]Kajita et al.,
   2024; [194]Liu et al., 2021; [195]Mandai et al., 2017; [196]Mehat et
   al., 2018; [197]Qiu, 2019; [198]Sharma et al., 2019; [199]Song et al.,
   2015; [200]Takagi et al., 2019; [201]Thomas et al., 2016). While
   several studies have used pluripotent stem cells ([202]Cho et al.,
   2019; [203]Kajita et al., 2024; [204]Mandai et al., 2017; [205]Mehat et
   al., 2018; [206]Qiu, 2019; [207]Sharma et al., 2019; [208]Song et al.,
   2015; [209]Takagi et al., 2019; [210]Thomas et al., 2016) as RPE
   sources, our previous work demonstrated that adult RPESC can serve as a
   practical source of RPE progeny for both modeling and RPE cell
   replacement therapy ([211]Blenkinsop et al., 2015; [212]Boles et al.,
   2020; [213]Davis et al., 2017; [214]Rabin et al., 2013). In this study,
   we show that cultured adult RPESCs recapitulate the major RPE
   subpopulations observed in acutely isolated, native RPE cells, with
   only minor differences in composition and gene expression. Our current
   findings support the use of cultured RPESC-RPE cells to replace RPE
   cells lost in diseases such as AMD. Moreover, these findings underscore
   the value of RPESC-RPE to model the native RPE layer in vitro
   ([215]Fortress et al., 2023).

   While the surface expression profiles of cultured and native RPE were
   similar, native RPE showed enrichment of immune-associated surface
   markers, including CD74 (regulates macrophage activation)
   ([216]Schroder, 2016), CD200 (inhibiting microglial activation)
   ([217]Zhang et al., 2011), and CX3CR1 (role in immune cell migration)
   ([218]Lee et al., 2018). In addition to their role on immune cells,
   CD74, CD200, and CX3CR1 can also be expressed by endothelial cells to
   interact with immune cells. The expression of these markers indicates
   dynamic immune cross-talk in native RPE, contrasting with cultured
   cells where corresponding immune interactions are not present. It is
   possible that following transplantation, cultured RPE may adopt these
   native immune-related expression patterns upon integration with the
   host tissue.

   In conclusion, our study provides a comprehensive characterization of
   adult human RPE subpopulations within native tissue and in culture.
   This characterization deepens our understanding of RPE biology and
   pathology and underscores the utility of RPESC-RPE.

Methods

Experimental model and subject details

   CITE-seq was performed on native RPE cells from three adult human donor
   eyes and three cultured adult RPE cell lines at passage 2. Donor eyes
   were obtained from approved eye banks with consent for research use.
   The donor details are listed in [219]Table S11. RPE cells were
   dissociated from donor tissues, and CITE-seq was performed on native
   and cultured cells. Cultured cells were examined for cobblestone
   morphology prior to use.

Method details

Eye dissection and RPE dissociation

   Human eyes were cut at the ora serrata, and the RPE layer was exposed
   by removing the anterior segment, the vitreous, and retina. Whole RPE
   dissection was performed with care taken to avoid the edge of the
   posterior eyecup or to puncture Bruch’s membrane ([220]Blenkinsop et
   al., 2013). RPE cells were dissociated by incubating the RPE with
   collagenase IV (Worthington Biochemicals) at 37°C in a humidified
   incubator. Cells were either immediately utilized or passaged as
   RPESC-RPE (additional details in [221]supplemental methods).

CITE-seq

   After dissociation of adult RPE cells from human donor eyes or cultured
   cells from transwell inserts, the resulting single RPE cells were
   tagged with the TotalSeq-A human universal cocktail, V1.0 (BioLegend),
   containing 154 unique cell surface antigens and 9 isoform control
   antibodies. Cells were then washed three times with DPBS and stained
   with 1 μL SYTO64 dye (Invitrogen, Carlsbad, CA) in 1mL DPBS for 20 min
   at room temperature, then washed twice in fresh DPBS. Next, cells were
   diluted to 25,000 cells/mL and dispensed into ICELL8 3′ DE chips
   (Takara Bio, CA) using an MSND device (Takara Bio). Cell dispensing and
   in-chip reverse-transcription PCR were performed using a 3′ DE Chip and
   Reagent kit (Takara Bio) according to the manufacturer’s instructions.
   Libraries were prepared using our in-house protocols, then sequenced on
   a NovaSeq 6000 high-output flow cell, generating 2 × 150-bp read
   lengths (GeneWiz). Raw reads were processed and mapped to human hg38
   genome and count matrices generated for further analysis (additional
   details in [222]supplemental methods). Full analysis code is available
   at [223]https://github.com/neural-stem-cell-institute/RPE-CITE-Seq.

RPE flat mount preparation and immunofluorescence staining

   Human donor eyes were dissected as described above. After removing the
   retina, the eyecup was filled with 4% paraformaldehyde to fix the
   RPE-choroid for 1 h at room temperature. RPE-choroid was flattened by
   cutting the RPE-choroid-sclera from the edge toward the optic nerve.
   Flat mount pieces containing peripheral and central RPE were moved to
   glass slides and immunostained with conjugated primary antibodies
   diluted in 2% BSA according to the manufacturer instructions. Tissues
   were then washed with DBPS three times and covered with glass
   coverslips using Fluoromount-G (Thermo Fisher) mounting media and
   imaged using a fluorescence microscope.

Cell sorting

   Magnetic activated cell sorting (MACS) and florescence-activated cell
   sorting (FACS) were performed for positive selection of RPE
   subpopulations. For MACS sorting, RPE cells were immunostained using
   phycoerythrin (PE)- or allophycocyanin (APC)-conjugated antibodies, and
   MACS anti-PE of anti-APC magnetic microbeads were used to capture
   positively labeled cells using MACS MS Columns (Miltenyi Biotec).
   Briefly, cells were incubated with PE- or APC-conjugated primary
   antibodies for 20 min, rinsed twice with DPBS+2% BSA and captured in MS
   columns attached to an MACS separator magnet (Miltenyi Biotec).
   Unlabeled cells were collected by washing the column twice with RPE
   medium containing 10% FBS, and labeled cells were released by removing
   the column from the magnetic stand and washing the column with fresh
   RPE medium with 10% FBS. Flow cytometry and FACS sorting were performed
   using an ARIA I system (BD Biosciences).

Time-lapse imaging and cell migration analysis

   After cell sorting, labeled and unlabeled cells were cultured in
   12-well plates at ∼10,000 cells per well. Time-lapse microscopy was
   performed on cultured cells using a Zeiss Axio Observer Z1 microscope
   equipped with a humidified incubator at 37°C and 5% CO2. Time-lapse
   videos were analyzed in Python (V 3.12.2) using pyclesperanto and
   napari-pyclesperanto-assistant plugin for segmentation, and btrack for
   cell tracking. The analysis of tracking data and figure generation was
   performed in R (V 4.3.2).

Image processing

   Morphological analysis was performed by analyzing phase contrast and
   phalloidin staining images from human RPE or cultured RPE. Images were
   converted to binary, and cell-cell junctions were detected using the
   Ridge Detection plugin for ImageJ
   ([224]https://ieeexplore.ieee.org/abstract/document/659930). Object
   identification and morphological analysis were performed using the
   EBImage package in R ([225]Pau et al., 2010).

Quantification and statistical analysis

   Analysis of single-cell transcriptomic data was carried out in R
   utilizing the following packages: Seurat (V4 and V5), ggplot2, hypeR,
   rrvgo, GOfuncR, igraph, and riverplot packages. Differences in the
   distribution of cell area in morphological analysis and migration
   distance in cell tracking were analyzed with a two-sample test based on
   Wasserstein’s distance using the two-sample package (V 2.0.1) in R.

Resource availability

Lead contact

   Further information and request for resources and reagents should be
   directed to the lead contact, Nathan Boles (nathanboles@neuralsci.org).

Materials availability

   This study did not generate new reagents.

Data and code availability

   The accession number for the raw single-cell RNA-seq (scRNA-seq) data
   reported in this paper is GEO: [226]GSE244572
   (GEO:
   [227]https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE244572).
   Code written for this study is available at GitHub (GitHub:
   [228]https://github.com/neural-stem-cell-institute/RPE-CITE-Seq). FAIR
   and CARE data management principles were followed.

Acknowledgments