We successfully fabricated 168 nichoids directly onto 12-mm diameter standard microscope glass slides, covering 70 % of the available culture surface (Fig. 1a). The scan speed and the laser power (1.5 mm/s and 12 mW, respectively) were optimized for the best mechanical integrity of the nichoids and to reduce the amount of microfabrication time (i.e., 11–12 hours to pattern 70 % of the available culture surface) (Fig. 1b, c).

Our first aim was to confine the mES cells within the nichoids. We expected that, upon cell seeding, a major fraction of the cells would fall by sedimentation driven by gravity inside the nichoids and that the confinement walls would prevent these cells from leaving the nichoids during culture. However, a small fraction of cells anchored themselves onto the 80-μm gap flat glass surface in between the nichoids. Interestingly, mES cells adhered to the nichoid substrates in the absence of a feeder layer, thus demonstrating that these nichoids provide favorable conditions for cell adhesion (Fig. 

2b

). mES cells, maintained in culture with LIF conditioning up to day 3, and with neither a feeder layer nor LIF conditioning from days 4 to 14, formed EBs. While EB configuration was immediately lost for those cells cultured on kidney ECM substrates, EBs cultured on both the nichoid and the 2-D glass substrates were preserved up to day 3 (Fig. 

2a

). While EBs on 2-D glass substrates greatly increased in size and spread during the culture, EBs in the nichoids maintained their spherical morphology and dimension (Fig. 

2a

). This feature was also confirmed by SEM analysis (Fig. 

2b

,

c

) which showed EBs adhered to the mid-plane of the nichoid preserving their round configuration. We attribute such behavior to the physical and geometrical constraints provided by the nichoid architecture.

Fig. 2

Morphology of the embryoid bodies formed by mES cells cultured in the nichoid substrates, compared to flat glass and to kidney ECM. Cells were cultured in the absence of a feeder layer and with LIF up to day 3, then without either a feeder layer or LIF from days 4 to 14. a Phase contrast images at days 3, 7, and 14. The scale bar is 100 μm. b, c SEM images of an embryoid body adhering to the nichoid. d Embryoid body diameter; n = 15, *p < 0.01. The red line shows the dimension of an individual niche. e Number of cells per embryoid body; n = 15, *p < 0.01

To quantify the containment effect, we measured the EB Feret diameter (Fig. 2d). While the sizes of both EBs in the nichoids and 2-D glass at day 3 were comparable (82.50 ± 7.8 μm and 100 ± 15.45 μm, respectively), the EB diameter in the nichoids was systematically lower at day 7 (120.50 ± 40.12 μm, in nichoids, 248.40 ± 68.45 μm on 2-D glass, n = 15, p value = 0.01) and day 14 (250.01 ± 52.35 μm, in nichoids, 325.40 ± 125.30 μm on 2-D glass). The average EB diameter in nichoids at days 3 and 7 was comparable to the characteristic length (i.e., 90 μm) of the repetitive niche units composing the nichoid substrate (Fig. 1c, Fig. 2d).

These measurements prove that there is a containment effect due to the 3-D nichoid architecture (Fig. 2d). In addition, the average cell number per EB in nichoids was significantly lower than that calculated on 2-D glass substrates at day 3 (19.70 ± 5.20 cells/EB and 28.88 ± 6.08 cells/EB, respectively, n = 15, p value = 0.01) and day 7 (29.55 ± 3.96 cells/EB and 37.04 ± 4.83, respectively, n = 15, p value = 0.01). Conversely, no statistical differences were found for the number of cells normalized by EB at day 14 (31.08 ± 4.05 cells/EB in the nichoids and 35.88 ± 4.08, n = 15, p value = 0.01) (Fig. 2e).

The number of cells per EB in the nichoids demonstrated, not only the confinement effect due to the nichoid architecture (Fig. 2d), but also a higher proliferation rate compared to the 2-D glass culture substrates (Fig. 2e). At days 7 and 14 compared to day 3, the proliferation rate measured was 50 % greater in average in nichoids, compared to 2-D flat glass substrates. However, this confinement effect diminished slightly with culture time. In fact, as the cell number increased, together with the size of the EBs over time, the available nichoid internal volume decreased. In addition, since there was no available space where they could grow, the EBs spread out from the 3-D architecture and no longer experienced the physical containment provided by the nichoid.

Our second aim was to assess the nichoid confinement effect on EBs on pluripotency maintenance and differentiation toward the three germ layers in feeder-free and LIF-free culture conditions. To evaluate stemness promotion and inhibition to differentiation, we stained and evaluated the co-occurrence of OCT4 and DAPI. The OCT4 pluripotent marker was highly expressed in cells cultured on both EBs in the nichoids, 2-D glass and kidney substrates in the presence of LIF conditioning at day 3 (78.80 ± 11.65 %, 76.16 ± 12.52 % and 64.94 ± 22.24 %, respectively). Once the LIF was devoid, OCT4 expression turned out to be significantly greater in nichoids compared to 2-D glass and kidney ECM substrate at day 7 (64.41 ± 15.51 %, 31.65 ± 20.75 % and 39.39 ± 15.00 %, respectively) (Fig. 

3a

,

b

,

n

 = 15,

p

value = 0.01). However, OCT4 was greatly diminished in all of the substrates tested at day 14 (20.42 ± 17.30 % in nichoids, 11.50 ± 7.86 % on glass and 4.34 ± 3.46 % on kidney ECM).

Fig. 3

Maintenance of pluripotency of mES cells cultured in the nichoid substrates, compared to flat glass and to kidney ECM. Cells were cultured in the absence of a feeder layer and with LIF up to day 3, then without either a feeder layer or LIF from days 4 to 14. a Immunofluorescence for OCT4 (red) and DAPI (blue) on the nichoid (gray), the flat glass and the kidney ECM at days 3, 7, and 14. The scale bar is 50 μm. b Quantification of OCT4 expression by image processing; n = 15, *p < 0.01

While such observations were to be expected for cells on kidney ECM, where EBs were rapidly lost (Fig. 2a), the pluripotency maintenance on EBs in the nichoids may have arisen from the physical constraints provided by their 3-D microarchitecture. In fact, OCT4 expression on 2-D glass substrates, in which EBs spread considerably and increased in size (Fig. 2c), greatly decreased at day 7 (Fig. 3b).

Note that at day 14, OCT4+ EBs were still greater than the EBs on 2-D glass substrates, but not significantly because EBs managed to escape from the nichoids, as can be observed from the measurement of their average diameter (Fig. 2d) and the number of cells per EB (Fig. 2e). In fact, at day 14 the EB size exceeded the repetitive niche unit characteristic length (i.e., 90 μm, red line in Fig. 2d), thus losing the physical constraint provided by the 3-D architecture. This means that EBs were not completely confined in the nichoids, at least not in the uppermost part. Thus, day 7 was the most representative time point to evaluate the effect of the nichoids on mES cell pluripotency because of the absence of LIF medium conditioning and the presence of physical containment due to the nichoids.

To further examine this effect, we plan to increase the nichoid height to prolong the time in which the EBs are constrained. However, a good compromise will be needed between the manufacturing times, the optical accessibility of the system (i.e., the focal depth of confocal or, if necessary, a two-photon microscope) and the necessity to easily detach and collect cells by trypsin at the end of the culture.

We thus assessed the differentiation potential toward the endoderm germ layer by staining and quantifying the co-occurrence of GATA-4 and DAPI in feeder-free layer culture conditions (Fig. 

4a

,

c

). This endoderm marker was highly expressed in cells cultured on both EBs in the nichoids and 2-D glass substrates (75.12 ± 12.33 %, 79.57 ± 11.31 %, respectively). On the other hand, it was negligible on kidney ECM (3.42 ± 2.86 %) (Fig. 

4c, n

 = 15,

p

value = 0.01) in the presence of LIF conditioning at day 3. This could be explained by the dimethyl sulfoxide (DMSO) used in cell freezing. DMSO has been reported as an induction factor for endodermal differentiation [

22

]. In the absence of LIF, GATA4 expression in nichoids thus slowed down significantly compared to the 2-D glass (5.24 ± 3.97 %, 24.42 ± 17.07 %, respectively.

n

 = 15,

p

value = 0.05) at day 7, while it was negligible on kidney ECM up to day 14. GATA4 in the nichoids and in the 2-D glass increased at day 14 (20.39 ± 19.06 %, 42.21 ± 22.15 %, respectively.

n

 = 15,

p

value = 0.05), resulting in an up-down-up expression that is well documented in the literature [

7

] (Fig. 

4a

,

c

). However, this behavior could also be due to a possible paracrine signaling effect (Fig. 

4b

). In fact, GATA4

+

-EB were mostly localized at the outer boundaries of the nichoids, in particular close to the GATA4

+

cells grown on the 80-μm gap flat glass surface in between the nichoids. Therefore, such cells experiencing the 2-D environment could have affected the mES cell differentiation in the peripheral nichoids.

Fig. 4

Spontaneous endodermal and mesodermal differentiation of mES cells cultured in the nichoid substrates, compared to flat glass and to kidney ECM. Cells were cultured in the absence of a feeder layer and with LIF up to day 3, then without neither a feeder layer nor LIF from day 4 to day 14. a Immunofluorescence for GATA4 (green) and DAPI (blue) in the nichoid (gray), the flat glass and the kidney ECM at day 3, 7, and 14. The scale bar is 50 μm. b Detail of a flat region surrounding a nichoid block, showing a possible paracrine effect generated by GATA4+ cells on the expression of the GATA4 marker by cells of peripheral nichoids. The scale bar is 20 μm. c Quantification of GATA4 expression by image processing; n = 15, *p < 0.01 **p < 0.05. d Immunofluorescence for α-SMA (red) and DAPI (blue) in the nichoid (gray), the flat glass and the kidney ECM at day 3, 7, and 14. The scale bar is 50 μm. e Immunofluorescence for NKX2.5 (green) and DAPI (blue) in the nichoid (gray), the flat glass and the kidney matrix substrate at day 3, 7, and 14. The scale bar is 50 μm. f Quantification of α-SMA expression by image processing; n = 15, *p < 0.01 **p < 0.05. g Quantification of NKX2.5 expression by image processing; n = 15, *p < .0.01 **p < 0.05

We also evaluated the EB differentiation potential toward the mesoderm germ layer by staining and quantifying α-SMA, as well as the co-occurrence of NKX2-5 and DAPI in feeder-free layer culture conditions (Fig. 4dg). The expression of α-SMA was negligible at all time points for EBs cultured in the nichoids as well as in the kidney ECM. Conversely, EBs cultured on the 2-D glass substrate were significantly upregulated with respect to the nichoids both at day 7 (1117.15 ± 425.06 pixel2 on 2-D glass, 36.73 ± 35.63 pixel2 in nichoids, n = 15, p value = 0.01) and day 14 (4045.80 ± 716.15 pixel2 on 2-D glass, 550.42 ± 398.02 pixel2 in nichoids, n = 15, p value = 0.01) (Fig. 4d, f). NKX2-5 was weakly expressed in all the culture substrates tested (Fig. 4e, g). However, NKX2-5 expression slightly increased on 2-D glass substrates at day 14. Both cardiac mesodermal differentiation markers were localized on glass substrates and therefore, this effect might be related to the stiffness of the glass substrate (of the order of GPa).

To assess the pluripotent nature of the cells used in this study and to adequately characterize their response in our control conditions, we investigated the spontaneous differentiation toward the three germ layers in long-term culture on 2-D glass substrates, without nichoid microstructures. This experiment was conducted in the absence of a feeder layer and with LIF conditioning up to day 3, while without feeder layer and LIF conditioning from day 4 to day 21 (Fig. 

5a

). Several markers were tested. As mentioned, we observed OCT4

+

cells at day 3 (i.e., LIF conditioning and no feeder layer) and OCT

EBs at day 21 (i.e., no LIF conditioning and no feeder layer). Concerning the endodermal differentiation, GATA4 showed an up-down behavior at day 3 and day 21, respectively, which is also reported in [

7

]. As previously mentioned, a possible explanation for this outcome may be due to DMSO reported as an induction factor for endodermal differentiation [

22

].

Fig. 5

Spontaneous differentiation of mES cells cultured on flat glass substrates. Cells were cultured in the absence of a feeder layer and with LIF up to day 3, then without either a feeder layer or LIF from days 4 to 21. a Immunofluorescence for OCT4 (red), GATA4 (green), SOX-17 (green), NKX2.5 (green), α-SMA (red), collagen type I (red), osteocalcin (red), βIII-tubulin (green) and DAPI (blue) at days 3 and 21. The scale bar is 50 μm. b Phase contrast image of differentiated cells at day 19, showing a morphology resembling pacemaker-like cells (arrows). The scale bar is 100 μm. c Phase contrast image of beating cells (black arrows) surrounding two embryoid bodies at day 19 (cell beating is shown in the Additional files 2: Video 1 and Additional file 3: Video 2). The scale bar is 100 μm. d Quantification of the beating frequency by image processing

We also assessed the SOX17 expression, which turned out to be negative for the whole culture, including in the nichoids (data not shown). Interestingly, α-SMA and NKX2-5 in EBs cultured on flat substrates were negligible up to day 3, and greatly expressed at day 21. Such markers are involved in mesodermal cardiac differentiation and were consistent with the observation of beating cells surrounding the EBs at day 19. This highlighted the particular spherical morphology typical of pacemaker-like cells (Fig. 5b, arrows), together with spontaneous beating (Fig. 5c, arrows, Additional files 2: Video 1, Additional file 3: Video 2, Additional file 4: Video 3, Additional file 5: Video 4 and Additional file 6: Video 6). The measured beating frequency was 42 beats/minute (Fig. 5d). Collagen I, a nonspecific matrix differentiation marker and osteocalcin, a tardive osteogenic differentiation marker, were also expressed in EBs in long-term culture on 2-D glass. Finally, we evaluated βIII-tubulin expression to assess the ectodermal differentiation on 2-D glass substrates. As expected, the differentiation was negative at day 3 (i.e., with LIF conditioning), but surprisingly it was upregulated at day 21 (i.e., without LIF conditioning). This could be due to the FBS in culture that might contain inductive factors [27].

Finally, our third aim was to assess the pluripotency maintenance and differentiation toward the three germ layers in feeder-free layer culture conditions in reused nichoids. We cultured mES cells in the absence of a feeder layer and with LIF conditioning up to day 3, then without either a feeder layer or LIF conditioning from day 4 to day 14. To evaluate stemness promotion and inhibition to differentiation, we stained and evaluated the co-occurrence of OCT4 and DAPI, while for the differentiation potential toward the endoderm and mesoderm germ layer, we stained and quantified the co-occurrence of GATA-4 and NKX2-5 on DAPI (Fig. 

6a

,

c

). Surprisingly, EBs were OCT4

+

throughout the culture period (74.35 ± 2.70 on average), whereas both GATA4 and NKX2-5 expression were negligible. EB-GATA4

+

was observed at day 3. As previously mentioned, a possible explanation could be the DMSO which was reported as an induction factor for endodermal differentiation [

22

]. The high OCT4 expression in the reused nichoids could depend on the protocol for cell detachment by trypsin. As shown in Fig. 

6b

(right), F-actin (red), and therefore fragments of plasma membranes were observed with no nuclei. Despite trypsin, cell residues (and/or other secreted proteins) may have remained anchored to the substrate, favoring cell adhesion, EB growth in nichoids (Fig. 

6b

, left), and pluripotency maintenance.

Fig. 6

Spontaneous differentiation of mES cells cultured on reused nichoid substrates. Cells were cultured in the absence of a feeder layer and with LIF up to day 3, then without either a feeder layer or LIF from days 4 to 14. a Immunofluorescence for OCT4 (red), GATA4 (green), NKX2.5 (green) and DAPI (blue) at days 3, 7, and 14. The scale bar is 50 μm. b Immunofluorescence for F-actin (red) and DAPI (blue) showing a few residual embryoid bodies anchored both to the nichoid (left) and the flat glass (right) after trypsinization. The scale bar is 30 μm. c Quantification of OCT4, GATA4 and NKX2.5 expression by image processing; n = 15, *p < 0.01

Our study revealed that the nichoid allowed expansion and pluripotency maintenance without feeder cells in the absence of exogenous soluble factors (i.e., LIF) by providing biophysical signals for pluripotency in prolonged culture of mES cells. The observation that mES cells lost EB configuration on 2-D flat glass substrates suggests that this 3-D expansion behavior was mediated by the physical microenvironment provided by the nichoid alone. Compared to 2-D glass substrates, the nichoid substrate provides cells with an increased surface-to-volume ratio and space to adhere and proliferate. In addition, the nichoid 3-D architecture, as well as its physical/geometrical constraint, may be the primary feature controlling the ES cell fate.

Our results are in agreement with the most recent literature on the effect of the biophysical environment on ES long-term pluripotency maintenance [1, 21, 23]. For example, various functionalized polymer substrates have been demonstrated to facilitate the long-term maintenance of human pluripotent stem cells (hPSCs) in a xeno-free culture and feeder-free culture system [23]. In addition, in [1] human induced pluripotent stem cells (hIPSCs) and human embryonic stem cells (hESCs) grown in a 3-D nanofiber environment maintained their pluripotency as long as they were kept on nanofibers. In contrast to these studies, we did not use a soluble conditioning medium for mES pluripotency [1, 21, 23, 28, 30].