Intercellular mRNA transfer alters the human pluripotent stem cell state

Significance

The presence of laterally transferred mRNA across different cell types is unclear in the context of mammalian biology. Here, we uncover extraordinary ability of episodic lateral (mouse to human) mRNA transfer that is sufficient to alter neighboring cellular phenotype. Our studies indicate that the mobile mRNA transfer is dependent on direct cell-to-cell contact. This also involves readily reprogrammable amount of transcription factor genes, such as Pou5f1, Nanog, Tfcp2l1, Tfap2c, and Klf4, which can be translated and directly affect the pluripotency state in the acceptor cells. This would indicate that mRNA transfer would allow for developing a genetically and chemically neutral reprogramming approach without nuclear transfer, cell fusion, transcription factor gene transduction, and small molecules.

Abstract

Intercellular transmission of messenger RNA (mRNA) is being explored in mammalian species using immortal cell lines. Here, we uncover an intercellular mRNA transfer phenomenon that allows for the adaptation and reprogramming of human primed pluripotent stem cells (hPSCs). This process is induced by the direct cell contact-mediated coculture with mouse embryonic stem cells under the condition impermissible for primed hPSC culture. Mouse-derived mRNA contents are transmitted into adapted hPSCs only in the coculture. Transfer-specific mRNA analysis shows the enrichment for divergent biological pathways involving transcription/translational machinery and stress-coping mechanisms, wherein such transfer is diminished when direct cell contacts are lost. After 5 d of coculture with mouse embryonic stem cells, surface marker analysis and global gene profiling confirmed that mRNA transfer-prone hPSC efficiently gains a naïve-like state. Furthermore, transfer-specific knockdown experiments targeting mouse-specific transcription factor-coding mRNAs in hPSC show that mouse-derived Tfcp2l1, Tfap2c, and Klf4 are indispensable for human naïve-like conversion. Thus, interspecies mRNA transfer triggers cellular reprogramming in mammalian cells. Our results support that episodic mRNA transfer can occur in cell cooperative and competitive processes, which provides a fresh perspective on understanding the roles of mRNA mobility for intra- and interspecies cellular communications.

In mammalians, there is an increasing evidence of transient mRNA transfers between cells in divergent biological contexts via extracellular vesicles (EVs) (1) and tubular connections (2). For example, EV-mediated transfer of mRNAs can encode reporter proteins, such as luciferase, Cre recombinase, and green fluorescent protein (GFP) in the in vitro conditions between HEK293T (3), mast cells (4), colon cancer cell lines (5), glioblastoma and endothelial cell coculture (6), and smooth muscle and endothelial cell coculture (7). The translated proteins from Cre mRNA-containing EVs can induce the functional recombination events in the mouse brain even in vivo (8, 9). However, most EV-enclosed mRNAs are fragmented and part of small-sized, full-length mRNAs are intact in the package of EVs (10–12).

Studies highlighted cell contact-dependent mechanisms of mRNA exchange that operate independently of EVs (13, 14). Intercellular tubular connections serve as potential subcellular structures that facilitate mRNA transport between cells. The tunneling nanotube (TNT), a thin, membranous bridge stabilized by F-actin, has been reported to mediate the intercellular mRNA transfer in the mammalian cells based on the in vitro culture system. Haimovich et al., used the immortalized mouse embryonic fibroblasts (MEFs) and several human cell lines to show that the endogenous mRNAs undergo transfer via TNT in the interspecies coculture (15). The mRNA transfer occurs globally in the transcriptome level (16), activated by extensive stressors such as oxidative stress and nutrient deprivation (15). Other tubular connections are also involved in the intercellular mRNA transfer such includes the microtubule-containing protrusions, which are likely to mediate the GFP and Gnat1 mRNA transfer from the transplanted photoreceptor cells to the recipient cells at the mouse retina (17). Despite advances in understanding direct intercellular mRNA transfer in various cellular contexts, its presence and significance in stem cells and their transcriptional coordination remain unexplored. Furthermore, the functional impact of donor-derived mRNAs on altering recipient cell phenotypes has yet to be established.

Here, by employing an interspecies experimental coculture model, we tested the nature and role of mRNA exchange between human pluripotent stem cells (hPSCs) and mouse embryonic stem cells (mESCs). The interspecies coculture enables us to distinguish endogenously expressed genes and laterally transferred mRNAs based on the sequence difference (16). We found the cell-contact-dependent transfer of divergent mRNAs between hPSCs and mESCs, containing pluripotency-regulating transcription factors. The membranous bridging structures mediate the mRNA transport. Upon receiving naïve mESC-derived mRNAs, hPSCs at the primed state, which resembles postimplantation epiblasts with more advanced properties than naïve state (18, 19), convert to naïve-like state. This process can be blocked by the inhibition of cell contacts or intercellular membrane protrusion formation and by mouse gene-specific knockdown of pioneer transcription factors in human cells. We propose interspecies mRNA transfer triggers cellular reprogramming in mammalian cells, which warrants further investigations in diverse physiological and pathological contexts to delineate the potential of episodic, lateral transcriptome exchange between neighboring cells.

Results

Interspecies mRNA Transfer between hPSC and mESC.

Unlike conventional hPSCs including induced pluripotent stem cells (hiPSCs) and ESCs, mESCs are maintained by the leukemia inhibitory factor (LIF) signaling activation and the inhibition of GSK3 and MEK kinases, referred to as 2i+LIF condition. Under 2i+LIF condition, primed hPSCs (H9 hESCs) alone culture lost the expression of pluripotency transcription factors Oct4 and Nanog (SI Appendix, Fig. S1 A–D) and failed to maintain their growth capacity (SI Appendix, Fig. S1E). However, we found that when cocultured with naïve mESCs, primed hESCs continued to thrive under 2i+LIF condition over 5 d with some human cells maintaining Oct4 and Nanog expression (SI Appendix, Fig. S1 A–D). mESCs-cocultured primed hESCs then expanded after human cell sorting and passaging (SI Appendix, Fig. S1F). To explore the possibility of interspecies mRNA transfer during cocultured hESC adaptation in 2i+LIF condition, we first investigated the presence of the mouse cell–derived reporter mRNA in human cells in the coculture system comprising naïve mESCs expressing enhanced GFP (EGFP) (G4-2) and primed hESCs (H9) without EGFP in situ by RNAscope (Fig. 1A and SI Appendix, Fig. S2A). By utilizing the human-specific antibody against a nuclear protein Ku80, we could readily distinguish the acceptor hESCs from the donor mESCs. The transfer of EGFP mRNA was detected in hESCs in the mESC coculture. At the same time, there were significantly less EGFP mRNA levels detected in the separate culture condition either with primed hPSC culture medium or with conditioned medium (CM) from mESC coculture (Fig. 1A). Quantification of EGFP mRNA spots revealed that an average of 1.56 EGFP probe-positive spots were present in a human cell in the coculture condition (Fig. 1B). The EGFP mRNA spots were observed in both peripheral and central regions of the colonies. To estimate the transferred amounts relative to the endogenous expression levels in mESCs, we unbiasedly processed images to count the RNAscope signal particles within the cells (SI Appendix, Fig. S2B). Quantification of over 400 cells revealed that the donor mESCs and acceptor hESCs contain 22.5 ± 3.8 and 1.6 ± 1.8 spots/cell of EGFP mRNA, respectively, indicating approximately 7.1% of donor-expressed transcripts transferred into acceptor cells.

Fig. 1.

Interspecies mRNA transfer between human and mouse PSCs. (A) Representative RNAscope images of EGFP RNA in separate cultures of EGFP-expressing mESCs, and hESCs either in primed maintenance medium or in CM of the coculture, and the coculture of mouse and human ESCs. The cells were also counterstained with anti-Ku80 antibody (human nuclei) and Hoechst 33342 (all nuclei). (Scale bar, 10 μm.) (B) Quantification of the number of EGFP RNA signals in A. The randomly selected fields (n = 6 to 10) of each sample were analyzed, and the RNAscope signals in approximately 400 to 600 cells were counted (the number of cells analyzed is indicated in parentheses on the graphs). The black horizontal lines indicate the mean of the distribution. (C) Representative, confocal z-sectioned RNAscope images of the coculture of hESCs (H9 mCh-Farnersyl5) and mESCs (MS2-tagged Oct4, OM) in 2i+LIF condition. The cells were also counterstained with Hoechst 33342 (nuclei). A confocal plane (xy) and orthogonal views (xz and yz) are shown. White arrowheads indicate the presence of mouse Pou5f1 mRNA inside the human cells. (Scale bar, 10 μm.) (D) Experimental design for primed hiPSCs (GFP-positive 317D6) and naïve mESCs (tdTomato-positive J1-tdT) coculture followed by FACS. Representative FACS analyses of mixed cells were shown either from day-5 coculture (mix with coculture) or from mix on ice prior to FACS (mix w/o coculture). GFP⁺/tdTomato⁻ and GFP⁻/tdTomato⁺ cells were sorted for the subsequent RT-qPCR analyses. (E and F) RT-qPCR analyses of mouse Actb and Neat1 RNAs in the sorted hiPSCs from mix w/o coculture, mix with coculture, or separate culture, or of human ACTB and NEAT1 RNAs in the sorted mESCs. Relative values normalized with endogenous β-actin (ACTB/Actb) expression are shown as an average ± SD from three independent experiments.

We also found mouse Pou5f1 (Oct4) transcripts were detected in the acceptor hESCs upon coculture with the donor mESCs. To detect endogenously transcribed Pou5f1 mRNAs in situ, we used Oct4 reporter mESCs (referred to as OM cell line), in which MS2 tag, a bacteriophage-derived stem-loop sequence, is knocked-in into the Pou5f1 locus (20). Confocal imaging of plasma membrane reporter hESCs (H9 mCherry-Farnesyl5) with serial z-sectioning revealed the presence of mESC-derived Pou5f1 mRNAs inside the human cells upon coculture (Fig. 1C). Quantification of the RNAscope spots revealed the acceptor hESCs contain 2.1 ± 1.4 spots/cell of mouse Pou5f1 mRNA, which was significantly higher than ones of hESC alone and negative control staining (DapB probe) (SI Appendix, Fig. S2 C and D).

To further validate the interspecies mRNA transfer at the endogenously expressed genes, we used the coculture system using primed hiPSCs expressing GFP (317D6) and naïve mESCs expressing tdTomato (J1-tdT) in 2i+LIF condition for 5 d, followed by FACS enrichment of each cell type (referred to hereafter as “mix with coculture”) (Fig. 1D and SI Appendix, Fig. S2E). To subtract the background signals potentially caused by cross contamination, we included the mixed samples of the two cell types that were grown separately, harvested, mixed on ice, and then immediately separated by the identical protocol to the coculture sample (referred to hereafter as “mix w/o coculture”) (Fig. 1D). Separate cultures of hiPSCs or mESCs were used as controls to determine endogenous gene expression levels before coculture. After FSC/SSC singlet cell sorting, we determined the purity of each enriched fraction by flow cytometry (SI Appendix, Fig. S2F). Among at least 5,000 counted cells of each fraction, the hiPSC-enriched fraction contained no tdTomato-positive cells and vice versa. Next, we designed RT-qPCR primers and probes specific for either human or mouse β-actin mRNA (ACTB or Actb) (SI Appendix, Fig. S2 G and H), which have been reported to undergo transfer between MEFs and human MCF7 breast cancer cells (16). Mouse Actb transcripts were more abundant in the sorted hiPSC fraction of mix with coculture condition than in the separate culture or mix without coculture condition (Fig. 1E). These patterns were confirmed in the other two hiPSC lines (SI Appendix, Fig. S2I). We also found the presence of mouse Nanog transcripts in hiPSCs of the coculture (SI Appendix, Fig. S2I). The detection of mouse Actb transcripts in hiPSCs upon coculture is cell-contact dependent since it was significantly less detected in CM of mESC and physically separated coculture condition via transwell (SI Appendix, Fig. S2J). We also found that human ACTB transcripts were present in the sorted mouse ESCs of mix with coculture condition (Fig. 1F). To test the transfer of a different type of RNAs, we evaluated a long-noncoding RNA, NEAT1/Neat1. Mouse Neat1 transcripts were not detected in the hiPSCs irrespective of conditions tested, and vice versa (Fig. 1 E and F).

Global Transcriptomic Profiling Revealed Divergent mRNA Transfer Involving Transcription Factors.

The presence of interspecies mRNA transfer prompted us to conduct comprehensive profiling by RNA-seq. We adapted the published informatics (16) to analyze the raw reads, including the length (>100 bp) and the type of fragment (paired-ends), by avoiding the mapping of mouse reads to the human genome and using data from both in silico simulation and sequencing experiment of the sorted cell samples. During the separation of human and mouse sequences in the obtained reads, we aligned the detected transcripts in human cells to mouse genome under the most stringent conditions to eliminate all ambiguous reads that may lead to reading misalignment. Compared to the separate culture condition, the coculture with mESCs increased the abundance of mouse-derived mRNAs found in the human cell fractions. Indeed, the percentage of mouse reads, relative to the total reads, in the hiPSC lines 317-12 and 317D6 increased from 0.05% (6,817 mouse reads) and 0.07% (8,186 mouse reads) to 0.3% (44,176 mouse reads) and 2.8% (311,859 mouse reads), respectively. On average, 9,953 transcripts of mouse genes were detected in the sorted hiPSC fractions derived from the coculture, with variable reads across genes (SI Appendix, Fig. S3A). Differential gene expression analysis (fold change >2, adjusted P values <0.01) revealed 4,881 genes that are highly detected in the sorted human cells after mESC coculture over separate culture of hiPSCs (Dataset S1). To evaluate the potential mouse cell contamination, we focused on the less transferred mRNAs within the coculture. Here, we included a long noncoding RNA Neat1, which could not transfer between hiPSCs and mESCs (Fig. 1E), as well as genes (Rc3h2, Ddit3, Sirt6, Ddc, Akt3, Cd82) characterized as less transferred mRNAs in the global analysis of contact-dependent transcriptome (16). Compared with highly transferred mRNAs such as house-keeping genes (Actb, Eef1b2, Rps11, Hspa8) and a transcription factor Pou5f1, these mRNAs were enriched neither in separate culture nor in sorted human cell fractions based on their raw counts (SI Appendix, Fig. S3B), supporting that the mRNAs directly transferred from mESCs significantly attribute to the mouse-derived transcriptome in the RNA-seq analysis.

We next asked whether the transfer of RNAs from mESCs into hiPSCs is dependent upon the endogenous expression level in the donor cells. To explore this, we performed linear regression analysis of the mouse read enrichment in the sorted human cells from coculture (log₂ fold change of the mouse genes of coculture over separate culture of hiPSCs) with the expression level in mESC separate culture (log₁₀ mean RPKM of mouse genes). We found the overall positive correlation between the endogenous expression level and the transferred amount (Pearson correlation coefficient 0.446) (SI Appendix, Fig. S3C): for example, translation factors (Eef1b2, Eif2s2), ribosomal proteins (Rps29, Rpl19, Rps11), and other metabolic and housekeeping genes (Fasn, Scd1, Paics, Hspa8). These indicate that the level of gene expression in donor cells reflects the intercellular RNA transfer as reported previously (15, 16).

Among the top 75 genes with the highest gene expression variance, all had higher expression in the coculture conditions (Fig. 2A). Gene ontology enrichment analysis revealed that a highly diverse group of genes are involved: e.g., protein targeting to membrane, symbiont process, translational initiation, and RNA processing and localization (SI Appendix, Fig. S3D). To analyze functional categories that are enriched in networks of the significantly abundant mouse genes, we performed overrepresentation analysis (ORA) based on the KEGG database (Fig. 2B). The resultant pathways included RNA processing (RNA transport, mRNA surveillance, spliceosome, and RNA degradation), protein translation and its quality control (ribosome, protein processing in endoplasmic reticulum, ribosome biogenesis in eukaryotes, ubiquitin-mediated proteolysis, and proteasome). Various signaling pathways, including pluripotency of stem cells (Oct4, Sox2, Nanog, and Stat3), metabolic reactions, cell proliferation/survival (cell cycle and survival-related growth factor signaling pathways), and cell death (apoptosis, ferroptosis, and senescence) were present. Interestingly, a series of stress-coping pathways are also enriched, including p53 signaling (35 genes including Tp53), DNA repair (38 genes), autophagy (71 genes), and signaling responsive to environment stresses (HIF-1 signaling, 40 genes; mTOR signaling, 73 genes; FoxO signaling, 60 genes), as well as metabolic pathways that may help stress adaptation response. In addition, Atf4, a transcription factor functioning as a master regulator of cellular stress responses such as endoplasmic reticulum stress and nutrient shortage, was listed as one of the most enriched transcription factor-coding mRNAs (Dataset S2). As hPSCs are exposed to 2i+LIF condition, cellular fitness might depend on such genes to thrive and adapt in an impermissible environment.

Fig. 2.

Global profiling revealed divergent mRNA transfer involving transcription factors. (A) Heatmap of differentially expressed mouse genes in hiPSCs cocultured with mESCs compared with the separate cultures of hiPSCs. The samples from two independent replicates were prepared from two hiPSC lines, 317-12 and 317D6. The top 75 genes with the highest gene expression variance across the samples were plotted. Colors represent gene expression of each sample relative to the mean. The coculture for the RNA-seq was performed in the absence of MEF feeder cells. (B) ORA of mouse genes enriched in the sorted hiPSCs after coculture with mESCs and the KEGG networks grouping these genes by the enrichment network tool NetworkAnalyst. The functionally grouped network was visualized based on the degree of connectivity between pathways and genes (nodes). The pathway nodes are colored according to P-value of ORA, while the size of the nodes corresponds to the number of genes from the analyzed gene list. The smaller nodes correspond to individual genes, colored according to their fold change. (C) Expression levels (read counts in RNA-seq) of the selected mouse transcripts related to pluripotency in hiPSCs after cocultured with mESCs. Average ± SD is shown. (D) RT-qPCR analyses of mouse Tfcp2l1, Nanog, Klf4, and Tfap2c mRNAs in the sorted hiPSCs from mix w/o coculture, mix with coculture, or separate culture. Relative values normalized with endogenous ACTB expression are shown as an average ± SD from three independent experiments. (E) Representative RNAscope images of MS2-tagged RNA in separate cultures of EGFP-expressing hiPSCs (317D6) in primed maintenance medium, and MS2-tagged Nanog (NM)- or MS2-tagged Oct4 (OM)-expressing mESCs in 2i+LIF, and in the coculture of hiPSCs and mESCs. The cells were also counterstained with anti-Ku80 antibody (human nuclei) and Hoechst 33342 (all nuclei). (Scale bar, 10 μm.) (F) Quantification of the number of MS2 probe-positive RNA spots in E. The randomly selected fields (n = 6 to 12) of each sample were analyzed for the RNAscope spots in over 500 cells. The black horizontal lines indicate the mean of the distribution.

Recently Hu et al reported that the RNA sensing pathway called retinoic acid-inducible gene I (RIG-I) pathway plays an important role in the interspecies PSC competition which determines the winner cell states at the primed pluripotency (21). They utilized the coculture of human and mouse primed PSCs and found the existence of cell contact-dependent human-to-mouse RNA transfer, which activates the RIG-I pathway in the surviving mouse cells. We asked whether the RIG-1 pathway is upregulated in hPSCs upon coculture with naïve mESCs by analyzing the RNA-seq of human cells sorted from the coculture. The surviving hiPSCs did not upregulate but rather downregulated the RNA receptors DDX58 (RIG-I) and IFIH1 (MDA5) without evident activation of the downstream genes such as IRF7, IFITM3, IFIT1, ISG15, and ISG20 (SI Appendix, Fig. S3E).

Among the mouse mRNA detected in the coculture conditions, we found 491 genes related to transcriptional regulator activities (Dataset S2). The pathway enrichment analysis revealed that signaling pathways regulating pluripotency of stem cells were among the overrepresented pathways (ranked sixth) (SI Appendix, Fig. S3F). These factors regulating pluripotency are clustered into the core network (Nanog, Pou5f1, Esrrb, and Sox2), and the downstream signaling of SMAD family protein (Smad2, Smad3, and Smad5) and Wnt signaling (Ctnnb1 and Tcf3) (SI Appendix, Fig. S3G). This is consistent with the finding that the detected set of mouse-derived transcription factor mRNAs includes Pou5f1, Esrrb, Tfcp2l1, Nanog, Sox2, and Klf4 as transcripts with abundant counts in the RNA-seq (Fig. 2C). We validated the RNA-seq results by RT-qPCR and RNAscope analyses. Compared with the mix without coculture condition, the direct contact–dependent coculture with mESCs (mix with coculture) led to the pronounced transfer of mouse-derived Tfcp2l1, Nanog, Klf4, and Tfap2c mRNAs into human iPSCs as determined by RT-qPCR of FACS sorted hiPSCs from the coculture (Fig. 2D and SI Appendix, Figs. S2I and S4 A and B). To detect the transfer of endogenous transcription factor-coding mRNAs in situ, we cocultured hiPSCs (317D6 line) with MS2-tagged Nanog (NM) or Oct4 (OM) reporter mESCs (20, 22) (SI Appendix, Fig. S4C). Using the MS2-specific RNAscope probe, the transfer of mouse Nanog and Oct4 mRNAs were detected within hiPSCs in the mouse ESC coculture, which displayed significantly more abundant RNAscope signals than those under separate culture condition, and no signals were detectable in the negative control probe staining (Figs. 1C and 2 E and F and SI Appendix, Fig. S4 D and E). These results indicate that transcription factor-coding mRNAs are transferred from mouse ESCs into human PSCs during the coculture.

mRNA-Containing Membrane Protrusions Interconnecting Human and Mouse Cells Are Required for mRNA Transfer.

Given the interspecies transfer of various mRNAs in PSCs in a cell-contact-dependent manner, we examined the physical contact sites in the coculture of primed hESCs and naïve mESCs. Consistent with previous observations of cell contact-mediated mRNA exchange (13, 15–17), we observed thin protrusions that appeared to connect hPSCs and mESCs in 2i+LIF condition (Fig. 3A). The structures were positive for both membrane-specific dye and F-actin, most of which were extended from hESCs as revealed by the expression of membrane reporter mCherry-Farnesyl5. These membrane structures containing F-actin did not touch the substratum (Fig. 3B), indicating the formation of TNT-like structures (23). In RNAscope assay, we found the presence of EGFP mRNA in the protrusion of hESCs when cocultured with EGFP reporter mESCs (Fig. 3C). We also found that hESCs extend their membranes toward the mESCs near the bottom of coculture over the distance corresponding to 2 to 3 cells (Figs. 1C and 3D), and that these extensions contain mESC-derived mRNAs as assessed by visualization of MS2-tagged Pou5f1 mouse RNAs (Fig. 3D).

Fig. 3.

mRNA-containing membrane protrusions from mouse to human cells are required for mRNA transfer. (A) Coculture of H9 mCherry-Farnesyl5 (red) hESCs and G4-2 GFP (green)-expressing mESCs stained with F-actin (cyan) and membrane (yellow) specific dyes. Top images are maximum projection of z-stacked images (Scale bar, 10 μm.) Bottom images are from a confocal plane of a magnified region containing membrane protrusions (Scale bar, 10 μm.) Arrows indicate representative membrane protrusions connecting hESCs and mESCs. (B) Representative, enlarged image from Fig. 3A showing the interconnected membrane bridges that were formed above the substratum. Top image is a maximum projection of z-stacked images while bottom orthogonal views (xy). (Scale bar, 10 μm.) (C) Representative confocal images of EGFP RNA (white) in the coculture of H9 mCherry-Farnesyl5 (red) hESCs and G4-2 GFP (green)-expressing mESCs visualized by RNAscope (Scale bar, 10 μm.) The area indicated by the dashed box is also magnified in the Right panels. (D) Representative confocal images of MS2-tagged Pou5f1 RNA (green) in the coculture of H9 mCherry-Farnesyl5 (red) hESCs and Oct4-MS2 mESCs visualized by RNAscope (Scale bar, 10 μm.) The images are captured near the bottom of the culture. The area indicated by the dashed box is also magnified in the Right panels. (E and E’) Maximum projection images of the coculture of H9 mCherry-Farnesyl5 (red) hESCs and G4-2 GFP (green)-expressing mESCs treated either with control or with L-778123 (0.5 μM). The cells were counterstained with F-actin-specific dye (blue) (E, Scale bar, 10 μm.) The area indicated by the dashed box is also magnified in E’ as a confocal plane (Scale bar, 5 μm.) (F) Quantification of the number (Left) and the length (Right) of membrane connections observed between hESCs and mESCs in the presence or absence of L-778123. (G) Experimental design for primed H9 hESCs (mCherry-positive) and naïve G4-2 mESCs (GFP-positive) coculture in the presence of L-778123 followed by FACS sorting of mCherry-positive fraction (Left, also see SI Appendix, Fig. S5B). RT-qPCR analyses of mouse Actb and Nanog RNAs in the sorted hESCs from mix with coculture conditions treated with the indicated concentrations of L-778123. The separate culture as well as mix w/o coculture was set as control. Relative values normalized with endogenous human ACTB expression are shown as an average ± SD from three to four independent experiments (right bar graphs).

Intercellular TNT-like structures have been implicated in mediating the transfer of various cellular materials including RNAs and organelles (24–28). The membrane remodeling associated with F-actin is regulated by GTPase family proteins, in which the prenyl group (farnesyl or geranylgeranyl) modifications are essential for their membrane association and biological activities. To target the membrane protrusions observed in the coculture of hESCs and mESCs, we treated cells with L-778123, a type 1 farnesyltransferase and geranylgeranyltransferase inhibitor (29). Upon L-778123 treatment the mCherry-Farnesyl5 reporter, which contains the farnesylation site, lost its membrane localization and diffused into cytoplasm in hESCs (SI Appendix, Fig. S5A). L-778123 significantly reduced the formation of membrane protrusions and their length observed in the contact site between hESCs and mESCs (Fig. 3 E and F). To investigate the effect of L-778123 on the intercellular mRNA transfer, we isolated the human cells based on mCherry expression from L-778123-treated, mESC coculture followed by RT-qPCR analysis of mouse Actb and Nanog genes (SI Appendix, Fig. S5B). L-778123 treatment upon coculture decreased the detected levels of mouse Actb and Nanog transcripts in the sorted hESCs (Fig. 3G). In this situation, the expression of the two genes as well as pluripotency markers in mESCs remain unchanged (SI Appendix, Fig. S5C). As control, L-778123 alone cannot alter pluripotent stem cell state of hiPSC in a concentration where Actb and Nanog reduction was noted (SI Appendix, Fig. S5 D–G). Collectively, these results indicate that the mESC-derived mRNAs can be delivered into the neighboring hESCs at least in part via intercellular membrane protrusions.

To explore the possibility that the donor cell-derived mRNAs are translated in the acceptor cells, we measured the reporter fluorescence protein signals of acceptor human cells in the coculture of GFP-expressing primed hiPSCs and tdTomato-expressing naïve mESCs described in Fig. 1D. While we did not capture significant tdTomato signals inside human cells under confocal microscope, which might be due to the detection limit in our microscope setup, in flow cytometry analysis we found the increased shift of mean tdTomato intensity in mix with coculture compared to separate culture and mix without coculture conditions (SI Appendix, Fig. S6A). In the membrane connections between hiPSCs and mESCs, we did not observe the EGFP signals in mCherry-Farnesyl5-positive protrusions (Fig. 3A), likely precluding direct EGFP protein transfer. We also assessed whether endogenous genes are detected in protein levels in human cells upon coculture with mESCs. We used mouse-specific antibodies against Oct4 and Nanog. Using separate culture of human iPSCs and mouse ESCs, we validated the specificity of these antibodies (SI Appendix, Fig. S6B), which never cross-reacted with human orthologs. Upon coculture with mESCs, the human cells in close proximity to mESCs were dim-positive for mouse Oct4 and Nanog as revealed by the mouse-specific antibody immunostaining (SI Appendix, Fig. S6 C and D). These results indicate the presence of mouse transcription factor proteins inside human cells upon coculture with mESCs.

Interspecies Coculture with mESCs Enables Naïve-Like Conversion of Human PSCs.

Cellular reprogramming can be achieved by the combinatorial introduction of mRNA subsets (30) with broad transcription control ability (31, 32). Given that mRNA transfer involves transcription factors, we hypothesized that coculture might influence cellular phenotype. Relative to mESC in a naïve state, conventional hPSCs, either ESCs or iPSCs, represent a more developmentally advanced, morphologically distinct, primed pluripotent state. We found that hPSCs after 5 d of naïve mESCs coculture in 2i+LIF condition underwent naïve-like morphological change (Fig. 4 A and B and SI Appendix, Fig. S1F). This naïve-like morphology became more prominent after the sorting human cells from the coculture and the subsequent culture in PXGL, a medium for culturing naïve hPSCs (33), confirmed by the presence of hESCs (H9) with dome-shaped morphology (Fig. 4A). Several other hiPSC lines (317-12, 317D6, and TkDA) also reproduced naïve-like morphological conversion after the coculture with mESCs (SI Appendix, Fig. S7A). RT-qPCR analyses of the hiPSCs after human cell isolation revealed the upregulation of naïve pluripotency markers (DPPA3, TFCP2L1, DNMT3L, KLF4, and KLF17) and downregulation of primed marker (DUSP6) compared with the separately cultured cells in primed states (SI Appendix, Fig. S7B). We also found with flow cytometry that 5 to 6 d after sorting following the coculture with mESCs, a proportion of human H9 cells (3 to 4%) were positive for SUSD2, a specific surface marker of human naïve PSCs (34, 35), as was seen in positive control chemically reset (cR) cells (33, 34) established from the same cell line (Fig. 4C). In contrast, most H9 cells cultured in 2i+LIF without m ESCs were mostly SUSD2-negative, similar to conventional, primed cells. We detected other cell surface markers of naïve hPSCs such as CD130 and CD77 in hiPSCs cocultured with mESCs (SI Appendix, Fig. S7C). In contrast, the expression of CD90 and HLABC reported as primed hPSCs markers (36) decreased.

Fig. 4.

Interspecies coculture with mouse ESCs enables naïve-like conversion of human PSCs. (A) Protocols for coculture of primed hPSCs (H9 human ESCs expressing a membrane reporter mCherry-Farnesyl5) and naïve mESCs (G4-2 mouse ESCs expressing GFP). Both cell types (50% each) were cocultured in 2i+LIF condition for 5 d followed by sorting human cells. The sorted human cells were further cultured in PXGL condition, representing the morphological conversion into dome-shaped colonies (white arrowheads). (Scale bar, 100 μm.) (B) Representative images captured at the indicated time points during the coculture of primed hPSCs (H9, red) and naïve mESCs (G4-2, green). (Scale bar, 100 μm.) (C) Flow cytometry analysis using naïve-specific SUSD2 antibody in H9 hESCs (expressing mCherry) that were separately cultured in a primed state (primed) or 2i+LIF followed by culture in PXGL (w/o mESC coculture), or cocultured with mouse ESCs in 2i+LIF followed by human cell sorting and culture in PXGL (with mESC coculture). The cR cells derived from the same H9 cells were also included. (D) Phase-contrast images in H9 hESCs that were converted into cR cells (Left) or expanded after sorting SUSD2⁺ cells from the cocultured H9 cells with mESCs (Right). (Scale bar, 100 μm.) (E) Immunostaining images of human naïve pluripotency markers KLF4, KLF17, and TFCP2L1 together with human-specific nuclear antigen (HuNu) in parental primed hiPSCs (Upper), cR cells (Middle), and SUSD2⁺-sorted human cells cocultured with mESCs (Bottom). (Scale bar, 20 μm.) (F) Principal-component analysis (PCA) was performed on RNA-seq measurement obtained from hiPSCs, including primed cells (parental), cR cells, and converted cells via mESC coculture and subsequent expansion. The data were clustered with previous, independently generated RNA-seq of hESCs in primed and cR states (both on-feeder and feeder-free) (33). (G) Unbiased hierarchical clustering of samples represented in F. The expression levels of selected naïve and primed pluripotency marker genes are shown. (H) Expression heatmap of selected naïve and primed pluripotency markers in hiPSCs, including direct coculture with mESC, CM from mESCs, transwell coculture, and primed cells. The representative images of each condition are shown (Top, Scale bar, 100 μm.) RT-qPCR analyzed the relative expression levels of each gene from three independent experiments.

We next asked how native gene expression of hPSCs is impacted by coculture with mESCs after human cell sorting and subsequent culture in PXGL. We performed RNA-seq at different timepoints during mESC coculture-dependent phenotypic conversion. Pattern clustering revealed seven distinct clusters of genes, which include downregulation of genes related to cell adhesion molecules and neuroactive ligand–receptor interaction toward the converted state as well as upregulation of genes related to signaling pathways regulating pluripotency of stem cells (SI Appendix, Fig. S7D). We observed the transient upregulation of genes during the coculture with mESCs that function in calcium signaling pathway and focal adhesion. Since cell–cell and cell–matrix adhesion interactions are dynamically rearranged between naïve and primed pluripotency and intracellular calcium contributes to these regulations (35, 37, 38), the observed transitioning transcriptome may influence the human cells adapting to the culture environment for naïve PSCs. While a portion of human cells maintained the expression of Oct4 and Nanog upon coculture with mESCs in 2i+LIF (SI Appendix, Fig. S1 A–D), we found a subset of naïve pluripotency signature genes, such as TBX3, IL6ST, TFAP2C, and DMNT3L, being expressed in the coculture period at the intermediate levels between primed and fully converted states (SI Appendix, Fig. S7E). Expression levels of some genes of pluripotency signaling including KLF2 and LEFTY2 are comparable with those in converted cells and cR cells (SI Appendix, Fig. S7E). These data indicate that hPSCs adopt to the state close to the naïve pluripotency after sorting from coculture with mESCs.

To further characterize the converted cell phenotype induced by the coculture, the SUSD2-positive sorted cultures of human cells were maintained for several passages (Fig. 4D and SI Appendix, Fig. S1F). We carried out immunofluorescence analysis in the propagated putative naïve hiPSCs. Consistent with cR cells, coculture-derived hiPSCs expressed naïve markers KLF4, KLF17, and TFCP2L1 with human specific nuclear antigen, whereas parental, separately cultured primed hiPSCs did not (Fig. 4E). Unbiased clustering and dimension reduced plot analysis of transcriptional profiles of RNA-seq data demonstrated that cR cells and the propagated cells via mESC coculture share nearly indistinguishable gene expression patterns, while both profiles were significantly different from those of the primed hiPSCs (Fig. 4F). The transcriptome of mESC-cocultured cells showed high correlation with that of cR cells both in-house and in previously published, extended passage cell datasets (33) (SI Appendix, Fig. S8A). The naïve markers including key transcription factors (KLF2, KLF4, KLF5, KLF17, TFCP2L1, and TFAP2C), epigenetic regulators (DPPA3, DPPA5, and DNMT3L), and transposable element regulator (ZFP57) were highly expressed in cR and mESC-cocultured cells (Fig. 4G). Conversely, these cells downregulated primed pluripotency markers including THY1, DNMT3B, and CD24 as well as lineage commitment genes like SALL4, ZIC2/ZIC3, and PODXL compared to the primed hiPSCs. Compared with the primed hiPSCs, the cR and mESC-cocultured cells displayed 5,195 and 5,467 differentially expressed genes, respectively, with a high level of shared genes between the two conditions (SI Appendix, Fig. S8 B and C). CM from mESCs and transwell coculture assays could not induce dome-shaped hiPSC phenotype and resulted in negligible induction of naïve pluripotency markers (Fig. 4H). The observed downregulation of DUSP6, a primed marker gene that is induced by FGF-stimulated Erk signaling, can be attributed to an inhibition of Erk by PD0325901, a component of 2i+LIF and PGXL media.

We also studied the loss of imprinting of mESC coculture-induced converted hPSCs, a hallmark of naïve, but not primed, PSCs, by studying biallelic expression genes in RNA-seq data (see SI Appendix). By calling single-nucleotide variants in certain exons, we listed imprinting genes which were expressed biallelically in either our sample while referring to the previous, large-scale analysis of imprinting in human PSCs (39, 40). In line with these studies our cR naïve hESCs as well as mESC coculture-induced converted cells represented loss of imprinting in ZDBF2, IGF2, DLK1, and ZNF597 while primed cells did not (SI Appendix, Fig. S8D), indicating the similar pattern of loss of imprinting in hPSCs along the mESC coculture-induced conversion as conventional chemical reset naïve state. Collectively, these results indicate that direct cell contact with mESCs induced the phenotypic conversion of primed hPSCs into naïve state.

To extend our observation to other interspecies conditions, we used the rat naïve iPSC line (riPSC), which has been generated in our previous reports (41, 42). Similarly to the coculture with naïve mESCs, the riPSCs constitutively expressing EGFP were cocultured with primed H9 hESCs expressing mCherry-Farnesyl5 under 2i+LIF condition for 5 d (SI Appendix, Fig. S9A), followed by FACS sorting of mCherry⁺ EGFP^- and mCherry^- EGFP⁺ cells (SI Appendix, Fig. S9B). RT-qPCR of the sorted mCherry⁺ EGFP^- human cell fractions using rat-specific primers and probes revealed the significant amount of rat-derived transcripts, such as Actb, Klf4, Nanog, Tfap2c, and Tfcp2l1, detected in the mix with coculture condition compared with the separate culture or mix without coculture condition (SI Appendix, Fig. S9 C–E). The human ACTB was present in riPSCs upon mix with coculture, suggesting the intercellular, bidirectional RNA exchange between these cells (SI Appendix, Fig. S9E). After the human cell sorting and subsequent culture in PXGL, the SUSD2⁺ human cells emerged from both conditions of coculture with mESCs and riPSCs, as revealed by flow cytometry (SI Appendix, Fig. S9F). We further expanded the SUSD2⁺ cells in PXGL that displayed dome-shaped morphology (SI Appendix, Fig. S9G). RT-qPCR of the expanded cells showed the upregulation of naïve marker genes and the downregulation of a primed marker gene DUSP2 (SI Appendix, Fig. S9H). These results suggest that the interspecies mRNA transfer and subsequent naïve-like conversion of hPSCs also occur in the other context of coculture with rat naïve pluripotent stem cells.

mRNA Transfer-Prone Human PSCs Efficiently Convert into Naïve-Like State.

Next, we examined whether the human cells that receive RNAs from the donor mESCs transition to a naïve-like state. In order to track the distribution of donor-derived RNAs during coculture, we first transfected Cy5-labeled GFP mRNA as a tracer mRNA into naïve mESCs (SI Appendix, Fig. S10A) and then cocultured them with primed hESCs (H9 mCh-Farnesyl5) in 2i+LIF for 5 d (Fig. 5A). As a control, the Cy5-labeled GFP mRNA transfer was not detected when the CM from the transfected mESCs was used for hESC culture (SI Appendix, Fig. S10B). Flow cytometry of the day 5 coculture cells revealed that mCherry-negative mESCs contained both Cy5 and GFP signals (Fig. 5B). A portion of hESCs also contained Cy5 signals upon coculture with Cy5-mRNA transfected mESCs, indicating the direct intercellular transfer of the tracer mRNA from mESCs into hESCs (Fig. 5B).

Fig. 5.

mRNA transfer-prone human PSCs efficiently convert into naïve-like state. (A) Experimental design for tracing a reporter mRNA transfected in mESCs during the coculture in 2i+LIF condition. In vitro transcribed, Cy5-labeled GFP mRNA was transfected in mESCs 1 d before the start of coculture with primed hESCs (mCherry-positive). Day 5 post coculture, mCherry-positive human cells were sorted into either Cy5-positive or -negative fraction. (B) Representative FACS analyses of the coculture at day 5. In mCherry-negative fraction, both Cy5 and GFP signals were higher in the coculture with Cy5-GFP mRNA transfected mESCs compared to that with nontransfected mESCs (Top). To gate the Cy5-positive human cells, coculture condition with nontransfected mESCs as well as one without mESCs (hESC only) was set as controls (Bottom). (C) Expansion of Cy5-positive and -negative hESCs (mCherry-positive) in PXGL medium after isolation from coculture with Cy5-GFP mRNA transfected mESCs. (Scale bar, 200 μm.) (D) RT-qPCR analysis of a proliferation marker MKI67 of the cells in C. Average ± SD is shown from three independent experiments. (E) Flow cytometry analysis of a naïve PSC surface marker SUSD2 of the cells in C. Primed hESCs were set as a control. Average of SUSD2 positivity ± SD is shown from three independent experiments. (F) RT-qPCR analyses of naïve or primed marker genes expressed in the cells in C. Primed hESCs were set as a control. Relative values normalized with endogenous ACTB expression are shown as an average ± SD from three independent experiments. (G) Experimental design for primed H9 hESCs (mCherry-positive) and naïve G4-2 mESCs (GFP-positive) coculture in the presence of L-778123 followed by FACS sorting of mCherry-positive fraction and expansion in PXGL medium (Top). RT-qPCR analyses of NANOG, KLF4, and TFAP2C expression in the expanded hESCs from mESC coculture. Relative values normalized with endogenous ACTB expression are shown as an average ± SD from three independent experiments (Bottom).

In addition, we tested the correlation between the Cy5 signal intensity and EGFP mRNA levels following the Cy5-based cell sorting experiments (SI Appendix, Fig. S10C). From the coculture with Cy5 GFP mRNA-transfected mESCs, we isolated Cy5-positive and -negative human cell fractions with the same gating as Fig. 5B and SI Appendix, Fig. S10D. Cy5-positive, but not -negative, fraction contained a large proportion of GFP mRNAs as assessed by RT-qPCR. In the sorted human cells from the coculture with Cy5 GFP mRNA transfected mESCs, Cy5-positive fraction contained more Actb mRNAs than Cy5-negative one (SI Appendix, Fig. S10D). The sorted Cy5-positive human cells (mCherry-positive) more expanded in the PXGL medium with the shape of domed-like colonies than Cy5-negative cells (Fig. 5C). Consistent with this, the Cy5-positive human cells expressed higher levels of a proliferation marker gene MKI67 than Cy5-negative cells (Fig. 5D). To assess the efficiency of conversion into naïve-like state, we performed flow cytometry of the naïve PSC surface marker SUSD2 in the expanded human cells. Compared to the Cy5-negative cells, the Cy5-positive human cells consisted of higher proportion of SUSD2 expressing cells 5 d after the human cell sorting (day 10 after the start of coculture) (Fig. 5E). RT-qPCR analysis verified the more naïve-like induction in Cy5-positive human cell fraction which expressed higher levels of naïve markers NANOG, KLF4, TFCP2L1, and TFAP2C as well as lower expression of the primed marker DUSP6 (Fig. 5F). The treatment with L-778123 during the coculture inhibited the induction of naïve marker genes NANOG, KLF4, and TFAP2C following isolation of human cells and subsequent culture in PXGL (Fig. 5G). To validate whether L-778123 does not directly influence the process of naïve-like conversion, we performed chemical resetting of H9 hESCs in the presence of L-778123 (SI Appendix, Fig. S5D). While the highest concentration (1.0 μM) reduced the emergence of dome-shaped colonies potentially due to the exposure at the vulnerable phase of cR protocol (PDL+VPA), the concentration of 0.1 μM of L-778123 produced the comparable, dome-shaped colonies along the cR (SI Appendix, Fig. S5E). RT-qPCR of the obtained human cell colonies showed no inhibitory effects of L-778123 addition in terms of naïve marker induction (NANOG, KLF4, TFCP2L1, TFAP2C) and primed marker downregulation (DUSP6) (SI Appendix, Fig. S5F). Surface expression of SUSD2 was also not affected by L-778123 (SI Appendix, Fig. S5G), suggesting that L-778123 itself does not inhibit the induction into naïve-like state in human primed PSCs. Overall our results support the notion that the hPSCs receiving mRNAs from mESCs are prone to convert into naïve-like state.

Pioneer Transcription Factor mRNA Transfer Is Indispensable for Naïve-Like Conversion of HiPSCs.

To examine the causal relationship of mESC-derived transferred mRNA on hPSC conversion into naïve-like state, we established a loss-of-function system in which mouse-derived transcripts in hiPSCs are specifically downregulated via shRNAs targeting mouse transcription factors (Fig. 6A). Among the list of mouse-derived transcription factors identified in the RNA-seq (Fig. 2C), we focused on Tfcp2l1, KLF4, and Tfap2c genes, since they are essential for the maintenance of human naïve pluripotency while self-renewal in the primed state is not affected when these genes are inactivated (43, 44). Based on the alignment of human and mouse transcripts for each gene, we designed shRNA sequences to target mouse, not human, genes (SI Appendix, Fig. S11A), followed by packaging them into lentivirus. To validate the specificity and efficacy of the designed shRNAs, we transduced these lentivirus-packaged shRNAs into independently propagated mESCs and hiPSCs in a separate culture and selected shRNAs that specifically down-regulated mouse Tfcp2l1/Klf4/Tfap2c genes (Fig. 6B; the shRNAs used are shKlf4 #1, shKlf4#3, shTfcp2l1 #1, shTfcp2l1 #2, shTfcp2l1 #3, and shTfap2c #3). We next generated hiPSCs (Venus-positive) stably expressing the selected shRNAs and confirmed that shRNA expression affected neither Oct4 expression in primed states nor naïve conversion efficiency via chemical resetting (SI Appendix, Fig. S11 B and C). After the coculture of shRNA-expressing hiPSCs with mESCs in 2i+LIF condition, we passaged the cells and continued the culture in the presence of puromycin to select human cells (Fig. 6A). As a control, luciferase-targeting shRNA expression resulted in the emergence of domed shape colonies after the puromycin selection while knockdown of mouse-derived Klf4, Tfcp2l1, and Tfap2c in hiPSCs induced no or few colonies (Fig. 6C). Immunostaining a naïve specific marker KLF17 in the remnant colonies after puromycin selection demonstrated that, unlike control shRNA, all tested shRNAs targeting mouse Tfcp2l1, Klf4, and Tfap2c did not produce KLF17-positive cells (Fig. 6D). Together, inactivation of mouse-derived Tfcp2l1, Klf4, and Tfap2c transcription factor-coding mRNAs in hPSCs abolishes the naïve reprogramming potential induced by mouse naïve PSC coculture.

Fig. 6.

Pioneer transcription factor mRNA transfer is indispensable for naïve-like conversion of hiPSCs. (A) Protocols for coculture of primed hiPSCs expressing Venus and shRNA specifically targeting each mouse transcription factor (puromycin-resistant) with naïve mESCs (puromycin-sensitive) followed by the induction of naïve-like conversion. After the coculture of both cell types (50% each) in 2i+LIF condition for 5 d, they were passed onto the puromycin-resistant feeder cells in the presence of puromycin in PXGL medium to eliminate mESCs. After the puromycin selection, the residual human cells were subsequently analyzed. (B) RT-qPCR analysis to validate the target species specific of the designed shRNA. The graphs show endogenous KLF4, TFCP2L1, TFAP2C, NANOG, and POU5F1 gene expression levels in hiPSCs (Right) as well as Klf4, Tfcp2l1, Tfap2c, Nanog, and Pou5f1 gene expression levels in mESCs (Left) in which lentiviral shRNAs targeting mouse Klf4, Tfcp2l1, Tfap2c, Nanog, or Pou5f1 designed in SI Appendix, Fig. S11A were expressed. Average ± SD (n = 3) is shown. (C) The emerging colonies of dome-shaped, human cells expressing control luciferase shRNA (shLuc) emerging after puromycin selection are shown (white arrowheads in the left image; Scale bar, 100 μm.) The number of Venus-positive, dome-shaped colonies emerging after the puromycin selection in each shRNA expression condition was shown in the box and whiskers plot (Right; bars represent min to max of the counts). The randomly selected fields (in 10 mm² area, n = 32) of each sample from two independent experiments were analyzed. (D) Immunostaining images of human naïve pluripotency marker KLF17 in Venus-positive hiPSCs expressing shRNAs targeting luciferase (shLuc) or mouse Klf4, Tfcp2l1, and Tfap2c after the coculture with mESCs followed by puromycin selection. (Scale bar, 50 μm.)

Discussion

Intercellular transfer of mRNAs via cell-contact-dependent (16, 45) or independent (46, 47) mechanism is gaining attention. Our experimental human and mouse pluripotent stem cell culture system uncovered the presence of extraordinary mRNA transfer that can influence neighboring cellular phenotype. In the transcriptome-wide trend, the transfer efficiency of mRNAs correlates with the endogenous expression in the donor cells and the variety of transfer-prone genes is not linked with specific, limited pathways, which is consistent with a previous study (16). Transfer-specific global profiling indicates that the mobile mRNA transfer i) is dependent on direct cell-to-cell contact, ii) contains divergent stress-coping genes, including ER stress protection genes, such as ATF4 and apoptosis inducing-genes such as p53, and pioneer transcription factors regulating pluripotency, such as KLF4 and TFAP2C (48, 49), iii) elicits both cellular “cooperative” and “competitive” events in the receivers, the part of which is found in interspecies PSC coculture (50), and iv) involves a readily reprogrammable amount of transcription factor genes, which results in alternations of the acceptor hPSCs from primed into naïve-like state.

The intercellular transfer of mRNAs derived from distinct cell types could provide a source of novel proteins in the recipient cells. EVs are the best-studied example delivering mRNAs across the cells, but most of the packaged mRNAs are fragmented with the untranslated regions being more enriched than coding sequences. As EVs typically carry both mRNA and its encoded protein, it is difficult to distinguish de novo translation in the recipient cells (10–12). Retroviral-like capsids that are translated and formed from endogenously encoded ARC and PEG10 mRNAs package their own mRNAs for the transfer to the neighboring cells (51–53), but are not able to deliver a variety of mRNAs expressed in the donor cells. As an alternative route, tubular connections including TNT have been reported to allow the intercellular transfer of divergent, full-length mRNAs as well as long noncoding RNAs (15, 16, 21, 54). In contrast to secreted EVs, the tubular connections can provide a direct, nonselective transfer of intact mRNAs from donor cells without the packaging process. This is probably due to the open-ended contact between cells in the proximity, which can mediate the transfer of bulk cytoplasmic components such as mitochondria and lysosomes (25). Recently, tubular membrane connections formed by an endogenous fusogen syncytin-2 has been reported to transfer a large cargo comprising CRISPR-Cas9 RNA–protein complex between the neighboring HEK293T cells (45). Together with these findings, our results supports that the cell contact-dependent, tubular connections enable the global transfer of mRNAs between distinct cell types, which can directly coordinate the genetic information within the local cellular community.

A previous study demonstrated that in response to the coculture with human breast cancer cells, MEF upregulates the gene expression related to the cancer-associated fibroblasts. However, the detailed phenotypic alternation and its causality of the human cell-derived mRNA transfer for such has been ill-defined (16). In our study, we developed the loss of function by using mouse gene-specific shRNA expressed in the human cells, revealing the requirement of mouse cell–derived transcription factors (Tfcp2l1, Klf4, and Tfap2c) for naïve-like conversion of hPSCs upon coculture with mESCs. Given the multiple transcription factor-encoding mRNAs transfer from mESCs, divergent mRNA can cooperatively drive the pluripotent state transition inside the acceptor hPSCs.

A major limitation of this study is the use of in vitro interspecies coculture system, which could not be directly translated to more biologically relevant contexts. Recently, interspecies chimeras containing hPSCs have gained much attention in the field of regenerative medicine, where human and other mammalian PSCs coexist in the reconstituted early embryos and compete with each other (21, 50). Future investigation will be of interest to determine the relevance and significance of mRNA transfer in the interspecies chimera conditions. Another limitation is the incomplete ultrastructural understanding of how much intercellular membrane protrusions including TNT between human and mouse PSCs contribute to the intercellular mRNA transfer. We used L-778123 to block the intercellular bridging of these structures, but the precise molecular machinery involved is still unknown. Understanding the intercellular mRNA transfer within a single species (e.g., mouse, human) and the mechanism of cell contact-dependent mRNA transfer will be key to further explore the physiological importance.

These findings indicate that mRNA transfer would permit the development of genetically and chemically neutral reprogramming approaches without nuclear transfer (55–57), cell fusion (58–60), transcription factor gene transduction (31, 61, 62), or small molecules (63, 64). The biological significance of the mRNA transfer mechanism sets the stage for further investigation in broadening the view of intra- and interspecies cell communication.

Methods

Primed human PSCs and naïve mouse ESCs/rat iPSCs were cocultured under 5% O₂ condition with or without feeder cells. On day 5, mixed cells were treated with Accutase into single cells followed by sorting human cells either by FACS or by MACS. After naïve conversion, SUSD2-positive cells with the labeled fluorescent reporter expression were sorted either with FACS or MACS for further characterization. Detailed medium, culture, cell isolation, and characterization protocols can be found in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

RNA seq data have been deposited in Gene Expression Omnibus (GE) of NCBI under accession number GSE216348 (65), GSE216338 (66), and GSE272112 (67). All other data are included in the article and/or supporting information.