Using our biochips, we isolated different contact modes between MSCs and cardiomyocytes at the single-cell level and investigated MSC-cardiomyocyte interactions through junctional coupling, cell fusion, partial cell fusion, and nanotubular connections. Although directly extrapolating the results to actual tissues is unfeasible, these data demonstrate that the single-cell models, each with a defined contact mode, are significant in providing basic interactional data at the molecular level, such as the direction of mitochondria transfer and the relationship between the molecular transfer rate and the extent of cell fusion. These data cannot be obtained systematically under a complicated cell-cell interactional environment. Our results suggest that 1) cell fusion, observed in microwell studies between stem cells and cardiomyocytes, may be an additional mechanism by which grafted cells can improve the infarcted myocardium; and 2) partial cell fusion and tunneling nanotubes can facilitate transfer of rMSC mitochondria to cardiomyocytes. Mitochondrial transfer from stem cells has been reported to rescue aerobic respiration in mammalian cells and may also promote cardiomyocyte reprogramming back to a progenitor-like state.
Among different culture models, means of cell fusion have varied widely, and therefore extensive cell fusion as an in-vitro
artifact cannot be ruled out 
. Our concept of a defined microenvironment with single-cell resolution provides a tool to identify the cell-fusion phenomenon and quantitatively study its occurrence rate. Using a confocal fluorescence microscope, we found double-nuclei cells with mixed labels in contact-promotive biochips with a maximal percentage of 6.3%, which is higher than the average (0.7%) previously reported for cell fusion between stem cells and cardiomyocytes 
. Since the cell-fusion process requires that (at least) two interacting cells have a large contact area, there are two possible reasons for the high occurrence of cell fusion in our biochips: (1) Our 1
1 rMSC-to-cardiomyocyte ratio was much higher than the seeding ratio of 1
40 typical of conventional coculture 
; (2) Our contact-promotive microenvironment increased the probability of broad cellular contact formation up to 95% in the limited space of a microwell. The cell-fusion phenomenon was also observed by patterning one rMSC and one cardiomyocyte in a larger circular microwell (Figure S2
(C, D)), but the occurrence decreased to 4.2%.
Partial cell fusion is characterized as fusion of the heterocellular plasma membranes and resulted in direct cytoplasmic contact and organelle transfer between the cells. In vivo partial cell fusion has been confirmed in the border zone of a rabbit myocardial infarction, where close contact between cardiac fibroblasts and dedifferentiated cardiomyocytes is accompanied by disruption of the basal lamina 
. In our study, the partial cell fusion phenomenon was identified by mitochondrial transfer from rMSCs to contacted cardiomyocytes. Surprisingly, about 30% of heterotypic cell pairs with a broad contact area were found to have mitochondrial transfer; thus the occurrence of partial cell fusion was much higher than full cell fusion and approximately comparable to the rate of junction formation. This observation of partial cell fusion suggests that engrafted stem cells can contribute to the infarcted myocardium through partial cell fusion along with electrical and mechanical coupling through connexin 43 and N-cadherin. Partial cell fusion provides a unique opportunity for transfer of only cytoplasmic components without incorporation of the donor nucleus into the acceptor cells 
. Although the benefit of partial cell fusion to neonatal cardiomyocytes has not been fully elucidated, it has been observed that rMSCs and adult-cardiomyocyte fused cells look like a type of cardiac progenitor: They demonstrate proliferative potential and express early cardiac transcriptional factors, but they lack contractile proteins 
. Partial cell fusion opens a path for intercellular interactions to transport the cargo vesicle, mitochondria, growth factors and cytokines. It has been reported that MSCs synthesize and secrete a broad spectrum of growth factors, and cytokines such as VEGF, FGF, MCP-1, HGF, IGF-I, SDF-1, and thrombopoietin. Many of these factors have been demonstrated to produce beneficial effects on the heart, including neovascularization, attenuation of ventricular wall thinning and increased angiogenesis. Although partial cell fusion associated with mitochondrial transfer has been reported to facilitate cardiac reprogramming to progenitor-like cells, there might be other trafficking components that cross the fusion area that play key roles in the reprogramming procedure. It has been shown that a combination of three developmental transcription factors (i.e., Gata4, Mef2c, and Tbx5) rapidly and efficiently reprogram postnatal cardiac or dermal fibroblasts directly into differentiated cardiomyocyte-like cells 
Tunneling nanotubes, a novel biological phenomenon in cell-to-cell communication over long distance, allow for transfer of cargo vesicles, mitochondria, and small molecules such as calcium 
. Actin filaments and microtubules are known to support the nanotubes between rMSCs and cardiomyocytes. In vivo nanotubular connection has been confirmed in mouse-heart tissue, in which fibroblasts extended membrane nanotubes between adjacent muscle bundles 
. To study this long-distance communication phenomenon, we designed contact-preventive biochips to separate one rMSC and one cardiomyocyte inside two compartments connected by one narrow channel. In our study, the nanotubes were extended only from rMSCs. The other type of long-distance connection from cardiomyocytes resembled as filopodium-like structures protruding from the membrane. The overall amount of long-distance connections, including nanotubes and filopodia, was about 30% in our contact-preventive biochips, lower than that typically observed in conventional culture. A similar conclusion was made by Dr. Rustom's group, who found that nanotube formation among neuronal cells on a microstructured surface was 6% lower than on a conventional glass substrate 
We found that mitochondria from rMSCs can be transferred into cardiomyocyte cytoplasm either through rMSC-origin nanotubes or cardiomyocyte-origin filopodia. The mitochondria from rMSCs accumulated around the cardiomyocyte nucleus after transfer, which may modulate transcription-factor activities in cardiomyocytes through mitochondria-to-nucleus retrograde signaling. Recently, it has been reported that, after being transferred into cardiomyocytes, stem-cell mitochondria may cause somatic reprogramming 
. In this report by Acquistapace and coworkers, although correlation between reprogramming and mitochondria transfer was confirmed using mitochondrial-depleted human MSCs, a genetic mechanism of reprogramming was not investigated. However, mitochondrial-retrograde signaling has been hypothesized to participate in the physiology and pathology of multiple cell types. For example, in skeletal myoblasts, mitochondrial stress caused upregulation of a number of genes involved in Ca2+
transport and storage, including Ryanodine Receptor I or II (RyR1 or RyR2), calreticulin, and calsequestrin 
. Furthermore, mitochondrial-retrograde signaling has been hypothesized to induce phenotypic changes and progression in tumors. Increased expression of cathepsin L, a target gene involved in retrograde signaling, is an important factor in the invasive behavior of tumor cells 
. Based on these facts, it could be hypothesized that, after being transferred into the cardiomyocytes, the MSC mitochondria would activate a mitochondrial-retrograde signaling pathway to induce dedifferentiation of the cardiomyocytes because of higher cytoplasmic Ca2+
concentration. According to the study on Zebrafish heart regeneration, it has been proposed that heart regeneration can be mediated by cardiomyocyte dedifferentiation and proliferation 
. These findings suggest a regenerative pathway that is an alternative to well-accepted candidates, including differentiation, cell fusion, and paracrine effects. In particular, transfer of stem-cell mitochondria may explain why functional cardiac improvements are observed despite the fact that few of the donor cells are engrafted long-term.
In our contact-promotive biochips, connexin 43 and N-cadherin expression between two interacting cells was analyzed. The diffusive distribution of these two proteins was frequently observed in our biochips, but junctional distribution remained at one-third of the entire sample. Similar research has been conducted by Dr. Bursac's group 
, who found upregulation of expression of connexin 43 and N-cadherin in skeletal muscle cells and mesenchymal stem cells that were cocultured with cardiomyocytes. However, their results were obtained on only homotypic/heterotypic cell pairs with broad cellular contact. In the study we reported here, connexin 43 and N-cadherin were also expressed on noncontact rMSC-cardiomyocyte cell pairs in the contact-preventive biochips at a rate of approximately 50% but only in a diffusive manner (Figure S2
(B)). These results indicate that cellular contact not only upregulated protein expression, but also relocated expressed proteins to the contact area to facilitate cellular interaction through the junction mode. Although a diffusive distribution of connexin 43 and N-cadherin has been reported in the infarct zone 
and in regions with implanted cells, further studies are necessary to investigate the precise distribution of junctional proteins at heterocellular interfaces within the heart.