Time restricted hPSC aggregate culture promotes maintenance of maximum growth kinetics

For bioreactor inoculation at ≈0.5 × 106 single cells/mL, hPSCs were initially expanded by 2D monolayer culture (schematic Fig. 1A). Applying perfusion feeding (including stepwise adaptation of the media throughput detailed in Methods), pH and the DO was feedback-controlled as recently established [13, 14]. On process day 7 (d7) about 35 × 106 cells/mL were observed for all three independent cell lines tested (Fig. 1B). A maximum specific growth rate of ≈1.0 d−1 was found on d2–d3, progressively declining to ≈0.76 d−1 from d4 onwards (Fig. 1C). Parallel cell cycle analysis revealed a low but progressive increase of cells in the G1 phase (indicative of non-proliferating resting cells) from ≈25% on d0 to ≈37% on d7 (Fig. 1D).

Fig. 1

Influence of cultivation interval on growth kinetics and cell cycle distribution. A Adherent 2D culture-derived human pluripotent stem cells (hPSCs; hHSC_1285) were once detached and seeded as single cells at 5 × 105 cells/mL to stirred tank bioreactors (STBRs). After 24 h, cultivation medium was exchanged via perfusion, while a porous glass filter was installed to retain hPSC aggregates inside the bioreactor. Throughout the culture, pH and dissolved oxygen concentration (DO) were controlled via triggered clock. The pH level was controlled to 7.1 initially by reduction of CO2 in the gas supply and afterwards by addition of 1 M NaHCO3. The DO was similarly regulated at 40% (of air saturation) by adaption of the O2 concentration in the supply gas. In two separate cultures, hPSC aggregates were dissociated inside the bioreactor after 3 or 4 process days and single cells were seeded back into the bioreactor at 5 × 105 cells/mL. This was repeated for a total of 5 passages, resulting in the consecutive passages p1 (grey), p2 (red), p3 (green), p4 (blue) and p5 (orange) (n = 3–11). B–D Exemplary viable cell density, specific growth rate µ and cell cycle analysis for the 7 day lasting historic process (D7) averaged for three independent cell lines. E–J Viable cell density, specific growth rate µ and cell cycle analysis (5 passage average) for a cultivation interval of 4 days (D4; E–G) and 3 days (D3; H–J). K Comparison of cell yield between the historic process (D7, black) and the shortened processes of 3 days (D3, grey) and 4 days (D4, red). Cumulative cell yield was calculated based on average fold expansions. L Cumulative cell yield for a total of 5 passages calculated based on average fold expansion for each individual passage in both approaches D3 and D4. Results are presented as mean ± standard error fo the mean (SEM)

This dataset indicates an optimal hPSC proliferation at our culture conditions until d3 – d4 only, suggesting further potential for process improvement by shortening the process duration from previously established 7 days (process designated as D7) [10, 11, 13] to 3 or 4 days (designated D3 and D4, respectively). Thus, we have next tested the dissociation of suspension-derived hPSC aggregates on day 3 or 4, respectively, followed by “cyclic suspension culture re-inoculation” of single cells seeded at ≈5 × 105 cells/mL.

Notably, in contrast to previously published enzymatic dissociation of suspension culture-derived hPSC aggregates [16, 17], we here established chemical hPSC aggregate dissociation directly in the bioreactor vessel under controlled temperature and impeller-based stirring at 120 rpm. For the sufficient dissociation of D3 (Additional file 1; Figure S2B) and D4 (Additional file 1; Figure S2C) aggregates into single cells, ≈12 min of treatment were required, resulting in cell suspensions equivalent to the enzymatic detachment of hPSCs from 2D monolayer (Additional file 1; Figure S2A).

The dissociation/re-inoculation cycle was repeated for 4 times, resulting in 5 sequential passages (including the first passage inoculated by 2D monolayer-derived hPSCs). In the first suspension passage (p1) of the D4 approach, an average cell density of 8.1 × 106 cells/mL was reached on d4, closely recapitulating the 9.1 × 106 cells/mL observed on d4 at previous D7 conditions. The following passages yielded similar cell densities of 6.6 (p2), 7.9 (p3), 9.7 (p4) and 9.2 × 106 cells/mL (p5), while the reduced value after p2 was noted (Fig. 1E). However, calculations of the specific growth rate revealed equal values and patterns of growth kinetics for all 5 passages tested, showing a maximum of 0.9–1.1 d−1 around d2–d3 (Fig. 1F). In accordance with D7 process data, a slight increase of a resting cell population in G1 from ≈20% (d1) to ≈26% (d4) was observed for all 5 passages tested (Fig. 1G), creating a striking “reset pattern” after single cell re-inoculation for all passages (Additional file 1; Figure S3B).

For D3 conditions, a maximum cell density of 5.2 × 106 cells/mL was found on d3 in p1, in accordance with the 4.6 × 106 cells/mL reached on d3 in D7. Similar but slightly lower cell densities of 4.2 (p2), 4.3 (p3), 3.7 (p4) and 4.0 × 106 cells/mL (p5) were observed in subsequent passages (Fig. 1H). In p1, the specific growth rate on d2–d3 was at 0.9–1.2 d−1 (equivalent to D4 and D7 conditions), remaining high at 0.9–1.0 d−1 on d2–d3 in following passages (Fig. 1I). The cell cycle pattern in D3 was similar to D4 and D7, while the population of resting cells in G1 notably increased from ≈18% on d1 to ≈24% on d3 (Fig. 1J), thereby again reflecting a cyclic “reset pattern” (Additional file 1; Figure S3A).

We subsequently compared the D3 and D4 output to our former D7 approach (Fig. 1K). Calculating the cumulative yield for 12 days, D3 yielded 4.37 × 1011 cells (i.e. in 4 passages), while D4 yielded 4.53 × 1011 cells (in 3 passages). To simplify the comparison to the D7 approach, the output of 2 passages was summated, resulting in 2.79 × 1011 cells in 14 days. An additional calculation for the 5 passages tested revealed that the D3 approach generated a cumulative cell yield of 3.76 × 1012 cells in 15 days, whereas the D4 strategy yielded 1.29 × 1014 cells in 5 passages equivalent to 20 days (Fig. 1L).

Overall, D3 and D4 share great similarity in maintaining the aspired maximum growth kinetics of hPSCs. This is supported by process-specific profiles of pH and DO revealing both high reproducibility across the 5 passages tested for each culture strategy and the cross-comparison of D3 versus D4 as well (Additional file 1; Figure S3C, D). Moreover, osmolality profiles of the culture medium, which increases as a direct result of base addition to maintain pH control (Additional file 1; Figure S3E, F), as well as the oxygen concentration pattern XO2 in the supply gas adapted for DO control (Additional file 1; Figure S3G, H) also share similar characteristics between the different passaging intervals tested.

hPSCs’ aggregate patterning and metabolic activity is maintained in matrix-free long-term suspension culture independent of the passaging interval

Continuous passaging of suspension-cultured hPSCs did not induce apparent changes in aggregate morphology and aggregate size patterning after repeated single cell inoculation. Throughout the 5 passages tested, aggregates maintained a smooth, round morphology in D3 and D4 (Fig. 2A, B). On d1, the mean aggregate diameter ranges at 80–120 µm for both conditions (Fig. 2C, D). Equivalently, on day 3, the mean aggregate diameter ranges at 185–195 µm in D3 and 170–200 µm in D4, indicating highly similar aggregate growth kinetics independent of the cultivation interval. Aggregates in D4 reached a maximum mean diameter of 220–260 µm on d4, which is notably below the critical 300 µm level reported to be potentially growth inhibiting due to diffusion limits [30].

Fig. 2

Influence of cultivation interval on aggregate size distribution and morpholoy. A, B Exemplary light microscopic pictures of process-derived aggregate samples on all process days of passage 1 and passage 5 for the 3 day (D3) and the 4 day (D4) cultivation interval (scale bar = 200 µm). C, D Aggregate diameter distribution over the cultivation time for D3 and D4 in consecutive passages p1 (grey), p2 (red), p3 (green), p4 (blue) and p5 (orange) (n = 3–11)

Sufficient glucose supply for avoiding cell starvation has been reported critical for maximizing the cell yield [13]. Starting with a glucose concentration of 16–17 mM measured after fresh E8 addition, the values readily decrease on d1 to 8.5–11 mM in D3 (Additional file 1; Figure S4A) and 10–13 mM in D4, respectively (Additional file 1; Figure S4B). Subsequently, perfusion with glucose-enriched culture medium was applied, resulting in stabilized glucose concentrations ranging at 14–19 mM (Additional file 1; Figure S4A, B). Correspondingly, lactate concentrations constantly increased to 22.5–25.5 mM on d3 in D3 (Additional file 1; Figure S4C) and 21–26.5 mM on d4 in D4 (Additional file 1; Figure S4D), showing comparable patterns for all 5 passages.

We further calculated the cell-specific metabolite conversion rates to better monitor cells’ metabolism along the process. In both D3 and D4 (Additional file 1; Figure S4E-F), qGlc was ranging at 9–19 pmol/(cell × d) while peaking on d2, showing similar, almost overlapping profiles for all passages. This was closely matched by qLac profiles for D3 and D4, ranging at 11–25 pmol/(cell × d) with a maximum on d2. Finally, the yield coefficient calculated as a ratio of qLac and qGlc, was stably ranging at 1.3–1.5 on d2–d3 and d2–d4 (Additional file 1; Figure S4G-H), respectively, further confirming that hPSCs’ metabolism remained highly stable at least for the 5 passages of continuous, matrix-free suspension culture investigated.

Gene expression patterning suggests 2D-to-3D-culture adaptation, but some differences among 3D samples are observable as well

Previous studies indicated that the transition from 2D adherent to 3D suspension culture may result in molecular changes in hPSCs [31, 32]. To investigate our approach, bulk gene expression patterning was performed on hHSC_1285 cells cultured in: 2D monolayer, harvested on d3 prior to bioreactor inoculation (sample termed: ML); 3D suspension, harvested from both the D3 and the D4 approach after earliest (p1) and latest (p5) passages on respective passage end points d3 (samples D3/p1 and D3/p5) or d4 (samples: D4/p1 and D4/p5).

Gene sets/pathways of DEGs comparing ML vs. 3D samples (D3/p1, D3/p5, D4/p1 and D4/p5; Additional file 1; Figure S5) with the web-based tool Enrichr [26] are listed in Table S11 (Additional file 1), revealing that pathways upregulated in ML correspond to TGF-β signaling (LEFTY1, ID1, PITX2, THBS1, LEFTY2, BMP7, NODAL, SMAD7), PI3K/AKT signaling (PDGFRB, CDKN1A, ITGB5, ANGPT1, TNC, PDGFB, THBS1, FGF4, TCL1B, FGF8, LPAR6, COL4A3, TEK) and heavy metal-binding metallothioneins (MT2A, MT1F, MT1G, MT1X, MT1H, MMP9, MT1E). In contrast, DEGs upregulated in 3D are predicted to be involved in cell adhesion (CLDN11, NTNG2, MADCAM1, HLA-DQB1), AMPK signaling (FBP1, CFTR, CREB5) and insulin secretion in response to glucose stimulus (ADRA2A, CFTR, FOXA2).

In more detail, a comparative analysis of ML versus 3D according to D3 and D4 origin revealed 256 significant (p < 0.05) DEGs (> twofold change; clustered in Additional file 1; Figure S6A), while 229 DEGs are shared between D3 and D4 (Fig. 3A). This high similarity of D3 and D4 is emphasized by their close clustering in the principle component analysis (PCA) compared to ML (Fig. 3B). Focused comparison of 3D according to the differential passaging intervals (D3 vs. D4) revealed 50 significant (p < 0.05) DEGs, 15 of which were beyond the twofold change threshold (clustered in Fig. 3C). DEGs upregulated in D3 are predicted to be involved in cell adhesion (CDH5, NTNG2, MADCAM1), JAK/STAT signaling (PDGFRB, CDKN1A, THPO) and kinase activity (PDGFRB, PKDCC, TEK; Figure S7A). On the other hand, DEGs upregulated in D4 correspond to TGF-β signaling and SMAD protein phosphorylation for regulation of pluripotency (LEFTY1, LEFTY2, BMP7, NODAL, SMAD7, ID1, WNT3; Additional file 1; Figure S7A).

Fig. 3

Gene expression analysis of 3D suspension culture-derived cell samples. A Venn diagram based on differentially expressed genes (DEGs) found between samples of ML and D3 or D4, respectively. B Principle component analysis (PCA) clustering samples according to D3 (grey), D4 (red) and ML (blue) origin. C Clustering and heat map analysis of significant (p < .05) DEGs beyond a twofold change threshold for D3 and D4-derived samples. D Venn diagram based on DEGs found between samples of ML and p1 or p5, respectively. E PCA clustering samples according to p1 (grey), p5 (red) and ML (blue) origin. F Clustering and heat map analysis of significant (p < .05) DEGs beyond a twofold change threshold for p1 and p5-derived samples

A similar comparison of ML with 3D according to early (p1) and late passage (p5) origin revealed 275 significant (p < 0.05) DEGs (> twofold change; clustered in Additional file 1; Figure S6B), with 225 DEGs shared between p1 and p5, respectively (Fig. 3D); the p1 vs. p5 similarity is further emphasized by their close PCA clustering compared to ML samples (Fig. 3E). Among the 94 significant (p < 0.05) DEGs between p1 and p5, 28 genes were differentially expressed beyond a twofold change threshold (Fig. 3F). Gene set enrichment analysis for these DEG sets (Additional file 1; Figure S7B) reflects upregulation of DNA deamination-related pathways (APOBEC3H, APOBEC3B) in p5, whereas the pathways identified for p1 correspond to negative regulation of receptor signaling via the JAK/STAT pathway (SOCS3, CAV1, MT1X), metallothioneins (MT2A, MT1F, MT1X) and fluid shear stress (CAV1, CAV2, PDGFB, CCL2).

Multiple passage suspension culture is fully compatible with hPSCs’ pluripotency, differentiation potential and karyotype stability

Pluripotency and unrestricted differentiation potential is a key feature of hPSCs. Analysis of markers associated with an undifferentiated state on the protein level revealed > 95% positivity for the transcription factors OCT-3/4, NANOG and the proliferation-associated marker KI67 as well as > 99% positivity for the surface markers TRA-1-60, SSEA-3 and SSEA-4 (Fig. 4A). These results were obtained after 5 passages of continuous suspension culture for both D3 (15 days) and D4 (20 days), thereby matching patterns of hPSCs that were initially detached from matrix-supported monolayer culture used for suspension culture inoculation. Aggregate-derived single cell dissociated hPSCs that were exemplarily seeded in monolayer after 5 passages of D4, showed robust nucleus-restricted expression of OCT-3/4 and SOX2 as well as surface markers TRA-1-60 and SSEA-4 in respective immunocytological staining (Fig. 4B).

Fig. 4

Influence of long-term, matrix-free suspension culture on pluripotency and genetic stability. A Exemplary flow cytometry analysis plots for surface markers associated with an undifferentiated state TRA-1-60, SSEA-3 and SSEA-4 as well as transcriptions factors OCT-3/4 and NANOG and proliferation marker KI67. Cells harvested at process endpoints of D3 (5 passages; 15 days) and D4 (5 passages; 20 days) were compared to monolayer-derived cells used for inoculation of the processes. B Representative immunocytological stainings of single cell-dissociated, seeded hPSC aggregates derived after 5 passages in D4 (20 days) stained for markers TRA-1-60, OCT-3/4, SSEA-4 and SOX2 (scale bar = 100 µm). C Undirected differentiation of D4 process-derived aggregates after 5 passages revealed expression of markers representing the three germ layers ectoderm (based on TUBB3), endoderm (based on SOX17 and FOXA2) and mesoderm (based on Vimentin). Scale bar = 100 µm. D Exemplary flow cytometry analysis plots of cardiomyocyte-specific markers NKX2.5, MHC, SA and cTNT after directed differentiation of D4-derived cells at the end of passage 5. E Karyotype of cells cultivated for 5 passages in D4 approach (20 days)

Importantly, undirected differentiation of p5 aggregates from D4 resulted in the formation of cell lineages representing the three layers endoderm (SOX17, FOXA2), mesoderm (Vimentin) and ectoderm (TUBB3) shown in Fig. 4C. Moreover, the directed differentiation of such p5/D4 aggregates by an established protocol [24] resulted in the induction of functional cardiomyocytes at ≈70% purity (Fig. 4D). Lastly, in agreement with quality control guidelines [33], we analyzed the karyotype to exclude abnormalities in our matrix-free 3D approach, as exemplarily shown for the D4 process after 20 days of culture (Fig. 4E).

The D4 approach was repeated with the genetically independent hPSC line GMPDU_8 for 5 passages, closely reflecting hHSC_1285 results (Additional file 1: Figures S8, S9).

Process upscaling and high density cryopreservation of suspension-derived hPSCs is compliant with direct bioreactor inoculation with cryopreserved cells

A major challenge for adapting our previously established suspension culture protocol [13, 14] to full GMP is the initially required 2D pre-culture, which is poorly controlled, manual handling-dependent and, in particular, incompatible with a closed system approach, which would be highly favorable for the full GMP bioprocessing of hPSCs in a clean room setting [34, 35].

As outlined in the Fig. 5A scheme, cryopreservation of suspension-derived hPSCs may provide an expedient solution to these challenges by ousting the need for 2D monolayer cultivation. Such strategy aims at cryopreserved cell banking for the direct inoculation of 3D suspension culture in STBRs by the thawed hPSC cells.

Fig. 5

Cryopreservation of suspension-derived hPSCs. A Intermediate cryopreservation of suspension culture-derived cells was established to fully enable matrix-free cultivation of hPSCs in stirred tank bioreactors. After 3 passages of each 4 days, cells were harvested after dissociation inside the bioreactor and cryopreserved in different cell densities as single cells. After thawing, these cells were directly transferred to bioreactors for an additional 4 days of cultivation. The cryopreservation cell densities (CPDs) tested were 20 (CPD20, yellow), 40 (CPD40, purple), 60 (CPD60, dark green), 80 (CPD80, dark orange) and 100 × 106 cells/mL (CPD100, brown) (n = 3). B The recovery after thawing was calculated as a ratio of the amount of cryopreserved cells and the viable cell count after thawing. C Recovery on d1 was calculated as a ratio of the amount of the starting cell number on process day 0 (aimed to be at 0.5 × 106 cells/mL) and the viable cell count on process day 1. D Viable cell density (bar chart) and viability (line graph) throughout 4 process days of cultivation after thawing cells cryopreserved at different cell densities. E Viability of cells directly after thawing. F Viability of cells on process day 1 (corresponding to line graph in D). G Specific growth rate µ calculated based on viable cell densities. Results are presented as mean ± SEM

For this purpose, larger amounts of cells have to be cryopreserved in a comparably small volume; we have therefore tested the cryopreservation cell densities (CPD) of 20, 40, 60, 80 and 100 × 106 cells/mL (termed CPD20–CPD100, respectively) in cryogenic vials. To produce the required amount of cells for these experiments, the cells were expanded in the DASbox Mini Bioreactor System for two passages at 150 mL culture scale followed by process upscaling to the 500 mL scale for an additional third passage in the DASGIP Bioblock, thereby reflecting a typical seed train step. Thereafter, aggregates were dissociated (as described before) in the Bioblock system to obtain a single cell solution. For cryopreservation a relatively simple medium composition consisting of 90% E8 medium and 10% DMSO supplemented with the non-ionic surfactant Pluronic F-68 (0.1%) and RI (10 µM) was used.

Firstly, we wanted to investigate cell loss due to necrosis, which may be caused by ice crystallization or osmotic shock and is directly observable after thawing [36, 37]. The post-thawing recovery was calculated as a ratio of the cell count before and after cryopreservation, revealing a recovery rate of 80 – 100% for all conditions tested (Fig. 5B). Cells that did recover from thawing were showing high cell viability of > 97% tested by conventional trypan blue staining (Fig. 5E). However, the major cause of cell loss after cryopreservation is reportedly apoptosis, more precisely anoikis, occurring in the first 12–24 h post-thawing [38,39,40]. We therefore calculated recovery on d1 (i.e. 24 h post suspension culture inoculation) that is the ratio of the viable cell density on d1 compared to the inoculation density on d0 (Fig. 5C). This assessment revealed a substantial cell loss of up to 45% at CPD20. Interestingly, the cell loss was substantially reduced to 15–30% for the higher freezing densities; these findings were backed-up by respective viability values ranging at 75–85% on d1 (Fig. 5F). However, subsequent cell sample analysis from day 3 onwards notably revealed > 93% cell viability (Fig. 5D), closely reflecting values observed for the conventional (cryopreservation-independent) hPSC cultivation (data not shown).

It must be noted though, that the differential amount of viable cells recovered on d1 post inoculation, substantially impacted on the final cell yields observed on d4 (Fig. 5D) that is: 2.4 (for CPD20-based inoculation), 2.8 (CPD40), 4.7 (CPD60), 6.0 (CPD80) and 4.7 × 106 cells/mL (CPD100). However, despite this early cell loss-dependent reduction in cell yields on d4, the specific growth rate calculated for the recovered cells was stable and reached a maximum of 0.8–1.0 d−1 from d2 onwards (Fig. 5G). This underscores that the cryopreservation step, per se, did not affect the growth kinetics of the viable hPSCs. Moreover, our finding that the cryopreservation at higher cell densities promotes viable cell recovery may facilitate the generation of respective cell banks for direct suspension culture inoculation, without the need for excessive cryopreservation capacities.

Intermediate high-density cryopreservation of suspension-derived cells fosters long-term matrix-free hPSC bioprocessing

We next combined our newly established protocol for the long-term matrix-free suspension culture with intermediate cryopreservation, to test the hypothesis that matrix-dependent 2D culture is entirely dispensable for hPSC bioprocessing. The cell line hHSC_1285 was cultivated for 3 passages at D4 conditions (equivalent to results shown in Fig. 1E) before aggregates were harvested, dissociated and cells were cryopreserved at 100 × 106 cells/mL, representing a suitable CPD. Subsequently, bioreactors were directly inoculated with these cells after thawing.

Reflecting previous results, the growth kinetic for the first passage after thawing (p3 + 1) was lower compared to the pre-cryopreservation results, yielding an average cell density of 1.8 × 106 cells/mL on d4 (Fig. 6A). Accordingly, the cell viability was at 78% on d1, but stabilized at 95% on d4. After aggregate dissociation and re-inoculation, the cell yield in p3 + 2 was higher compared to the previous passage i.e. 4.4 × 106 cells/mL, but was still lower compared to values before cryopreservation. This was notably associated with a decline in cell viability down to 83% on d1, potentially suggesting a higher sensitivity of the cells to the applied dissociation protocol in consequence to the intermediate cryopreservation step. Nonetheless, cell densities obtained in later passages successively increased to 5.7 (p3 + 3), 6.4 (p3 + 4) and 7.7 × 106 cells/mL (p3 + 5) on d4, respectively, and cell viability remained high and stable at > 95%.

Fig. 6

Influence of intermediate cryopreservation of hPSCs on long-term, matrix-free suspension culture in STBRs. A–G After 3 passages of D4 cultivation (12 days), hHSC_1285 hiPS cells were cryopreserved at 100 × 106 cells/mL for direct inoculation of bioreactors with these cells. Depicted here are the viable cell density (A), cell cycle analysis (averaged over 5 passages; B), metabolic conversion rates (qGlc as continuous line, qLac as dashed line; C), specific growth rate µ (D), aggregate diameter distribution (E), metabolic yield coefficient (F) and the recovery on d1 (G) for hHSC_1285 in a D4 cultivation approach over 3 passages (not shown) pre- and 5 passages post-cryopreservation. This resulted in the consecutive passages p3 + 1 (grey), p3 + 2 (red), p3 + 3 (green), p3 + 4 (blue) and p3 + 5 (orange) (n = 3). Results are presented as mean ± SEM. H Representative immunocytological stainings of cryosections of hPSC aggregates derived after 3 + 5 passages in D4 (32 days) stained for markers of an undifferentiated state TRA-1-60, OCT-3/4, SSEA-4 and SOX2 (scale bar = 100 µm). (I) Exemplary flow cytometry analysis plots for surface markers associated with an undifferentiated state TRA-1-60 and SSEA-3 as well as transcriptions factors OCT-3/4 and NANOG and proliferation marker KI67. Cells were harvested at process endpoint (3 + 5 Passages; 32 days). J Karyotype of cells cultivated for 3 + 5 passages in D4 approach (32 days). K Undirected differentiation of D4 process-derived aggregates after 5 passages revealed expression of markers representing the three germ layers ectoderm (based on TUBB3), endoderm (based on SOX17 and FOXA2) and mesoderm (based on Vimentin). Scale bar = 100 µm. (L) Exemplary flow cytometry analysis plots of cardiomyocyte-specific markers NKX2.5, MHC, SA and cTNT after directed differentiation of cells at the end of passage 3 + 5

Importantly, the specific growth rate remained high for all passages tested after cryopreservation, reaching the expected maximum between d2 and d4 (Fig. 6D). In p3 + 1, the maximum specific growth rate ranged around ≈0.8 d−1, raising to 0.9–1.0 d−1 at later passages similar to data observed before cryopreservation. Moreover, cell cycle distribution patterns were also highly consistent with observations before cryopreservation (Fig. 6B; Additional file 1: Figure S10A).

The recovery of cells on d1 (depending on cell survival after re-inoculation) continuously increased from the first to the fifth passage after cryopreservation: 34% (p3 + 1), 44% (p3 + 2), 65% (p3 + 3), 84% (p3 + 4) and 89% (p3 + 5). Notably, this passage-dependent viability on d1 correlated to the cell yield on d4 (Fig. 6G), clearly indicating key importance of early cell survival and aggregation for the subsequent process performance, as previously noted by us [13].

The increase in cell yield from p3 + 1 to p3 + 5 was also reflected by the respective process control parameters i.e. of glucose/lactate (Additional file 1: Figure S10B, C), pH and DO (Additional file 1: Figure S10D), the oxygen concentration in the gas supply (Additional file 1: Figure S10E), and the medium osmolality (Additional file 1: Figure S10F). However, profiles of the cell-specific metabolic conversion rates were matching with those before cryopreservation (Fig. 6C) and the yield coefficient calculated on basis of these conversion rates was ranging from 1.1 to 2.0 for d2–d4 and did not reveal any passage-dependent changes (Fig. 6F).

Finally, we found that the overall aggregate morphology and size distribution patterns were generally retained after cryopreservation (Additional file 1: Figure S10G), but the average aggregate diameter somewhat increased consecutively from ≈210 µm (p3 + 1) to ≈280 µm (p3 + 5) on d4 (Fig. 6E), which similarly observed for GMPDU_8 cells (Additional file 1; Figure S8E).

When adding the 3 passages before and 5 passages after cryopreservation, the hPSCs were cultivated for 32 days in matrix-free suspension culture without indication of pluripotency loss, as demonstrated by immunocytological staining on sections of process-derived hPSC aggregates (Fig. 6H) and by flow cytometry (Fig. 6I). Furthermore, maintenance of the full differentiation potential was validated by undirected differentiation resulting into derivatives representative of all three germ layers (Fig. 6K) and the formation of > 75% cardiomyocytes by our directed differentiation approach (Fig. 6L). Importantly, an unaltered karyotype was revealed (Fig. 6J), together highlighting the great potential of this advanced hPSC-bioprocessing strategy.