Tailor-made generation of insulin-producing cells from canine mesenchymal stem cells derived from bone marrow and adipose tissue

Trend of regenerative therapy for diabetes in human and veterinary practice has conceptually been proven according to Edmonton protocol and animal models. Establishing an alternative insulin-producing cell (IPC) resource is a challenge task for further clinical application. In this study, IPC generation from two practical canine mesenchymal stem cells (cMSCs), canine bone marrow-derived MSCs (cBM-MSCs) and canine adipose-derived MSCs (cAD-MSCs), was of interest. The results illustrated that cBM-MSCs and cAD-MSCs contained distinct pancreatic differentiation potential and required the tailor-made induction protocols. Effective generation of cBM-MSC-derived IPCs needed an integration of genetic and microenvironment manipulation using hanging-drop culture of PDX-1-transfected cBM-MSCs under three-step pancreatic induction protocol. However, this protocol was resourceand timeconsumed. Another study on cAD-MSC-derived IPC generation found that IPC colonies could be obtained by low attachment culture under three-step induction protocol. Further Notch signaling inhibition during pancreatic endoderm/progenitor induction yielded IPC colonies with trend of glucose-responsive C-peptide secretion. Thus, this study showed that IPCs could be obtained from cBM-MSCs and cAD-MSCs by different induction techniques, and further signaling manipulation study should be conducted to maximize the protocol efficiency.


Introduction
Diabetes is not only a major metabolic disease affecting people around the world, but also the companion animals, mostly dogs and cats [1][2][3] . By pathophysiological basis, it is classified into 2 main types, type I and II, as characterized by absenting or presenting of intact beta-cells, respectively 2,4 . Type I diabetes is referred to an immune-mediated beta-cell destruction causing endogenous insulin depletion, while type II is related to insulin secretion defect and/or insulin resistance 1,4 . Although, diabetes treatment seems well-established, adverse events and compromised clinical efficiency have been periodically reported 3,5 . Trend of regenerative treatment has been introduced for addressing these issues starting from cadaveric islet transplantation in diabetes type I patients, namely "Edmonton protocol", which resulted in long-term omitting of exogenous insulin administration [6][7][8] . However, two main obstacles have been suggested, donor shortages and immunosuppressants' side effects, making stem cell (SC)based regenerative approach be the potential clinical candidate [6][7][8][9] .
Concept of SC-derived insulin-producing cell (IPC) transplantation for treating diabetes has been conceptually approved in animal models. However, it comes with further challenges on finding potential candidate cell sources and establishing efficient IPC production platforms that are clinically applicable 10-12 . Although, the study on IPC production using human SCs has widely been studied and well-established, the knowledge in IPC generation aiming for veterinary application is still lacking. It has been a few reports suggesting the induction of canine somatic cells and canine mesenchymal MSCs (cMSCs) toward IPCs in vitro 13,14 . These generated IPCs were formed as cell aggregates attached to culture surface that might cause some difficulties during cell harvesting and processing for transplantation. To earn the clinical applicable IPCs, three-dimensional (3D) structure of IPCs floating or suspending in culture vessels would be required to ease the harvesting and encapsulating processes 15 . To address this issue, the integrative induction protocols aiming for the pancreatic differentiation of canine bone marrowderived MSCs (cBM-MSCs) and canine adipose-derived MSCs (cAD-MSCs) were established in this study. The protocols were aimed for the delivery of 3D colony structure of the generated IPCs. Notch signaling manipulation was additionally conducted in the potential protocol for maximizing the induction efficiency. The results will be the crucial platform supporting the IPC generation which eventually benefits the establishment of clinical protocols for both veterinary and human applications.

cBM-MSC and cAD-MSC characterization
The isolated cBM-MSCs ( Figure 1A and B) and cAD-MSCs ( Figure 1C and D) showed fibroblast-like appearance upon 2D culture. mRNA expression of stemness-related markers (Rex1 and Oct4) and proliferation marker (Ki67) were detected ( Figure 1E and F). MSC-related surface marker analysis by flow cytometry showed that both cells contained high proportion of Cd90 + cells, while the proportion of Cd73 + cells was relatively low. The expression of hematopoietic surface marker (Cd45) was considered absent in both cells ( Figure 1G and H).
Both cells illustrated the in vitro osteogenic differentiation potential upon the 14-day induction protocol regarding ECM mineralization as demonstrated by Alizarin Red S and Von Kossa staining ( Figure 1I and J) and osteogenic mRNA marker expression (Alp, Runx2, Osx, Opn, Ocn, and Col1a1) ( Figure 1K and L).
The results revealed the MSC-related characteristics of the isolated cBM-MSCs and cAD-MSCs.

Generation of IPCs from cBM-MSCs requires 3D culture condition
To generate IPC colonies from cBM-MSCs, three different culture techniques were investigated (Figure 2A-C). In all culture techniques, three pancreatic induction media were used as a microenvironmental manipulating/small molecule inducing approach. The results as illustrated in Figure 2D showed that suspending the cells in low attachment culture dish (Protocol I) was unable to deliver IPC colonies, while maintaining the cells using hanging-drop technique (Protocol II) could successfully generate IPC colonies with 50-200 µm in diameter.
However, the colonies seemed loose cell aggregates. Further investigation was performed by maintaining the colonies collected from hanging-drop culture in Matrigel ® -embedded culture condition (Protocol III). Although, the generated colonies were dense and compact, they could not maintain colony structure after gel digestion using Cell Recovery Solution ® making them unable to be harvested for further functional testing.
Comparison of the pancreatic mRNA markers of the generated IPC colonies revealed that colonies from Protocol II expressed high pancreatic endoderm marker (Pdx1), but low pancreatic beta-cell markers (Nkx-6.1, Isl-1, Glut-2, and Insulin), comparing with those from Protocol III ( Figure 2E and F). However, the mRNA expression of pancreatic-relating markers (Glucagon and Glp1r) was not detected in Protocol III ( Figure 2G).
Further functional testing showed that IPC colonies collected from Protocol II secreted C-peptide under basal condition but could not produce a significant response upon low (5.5mM) and high (22mM) glucose stimulation. There was only trend of increased C-peptide secretion compared to basal control ( Figure 2H). 6 Thus, generating IPCs from cBM-MSCs by microenvironmental manipulating/small molecule inducing approach required 3D culture condition. However, the generated IPCs showed limited function and maturity.

Overexpression of PDX1 fails to generate IPCs from cBM-MSCs
Further generating IPCs from cBM-MSCs using genetic manipulating approach was The results suggested that overexpression of PDX1 could not successfully generate IPC colonies from cBM-MSCs in terms of pancreatic islet morphology and genotype.

MSCs
In order to effectively generate IPCs from cBM-MSCs, combination of genetic and microenvironmental manipulating approaches was used. Cells were transfected with lentiviral vector carrying human PDX1 at MOI 20 then maintained with three-step induction protocol under 3D culture condition (hanging-drop technique) ( Figure 4A). The results illustrated that IPC colonies started forming since day 5 of the induction, and size of colonies at day 12 was approximately 100-200 µm ( Figure 4B).
Pancreatic mRNA analysis showed that pancreatic endoderm marker (Pdx1) and pancreatic beta-cell markers (Isl-1, Maf-A, Glut-2, and Insulin) were significantly upregulated ( Figure 4C and D). However, alpha-cell hormonal marker (Glucagon) was highly expressed ( Figure 4E), while Glp1r was not detected. Functional testing also showed that IPC colonies secreted C-peptide under basal condition, but they could not produce a dose-dependent response upon low (5.5mM) and high (22mM) glucose stimulation ( Figure 4F).
Thus, combination of genetic and microenvironmental manipulating approaches effectively generated IPCs from cBM-MSCs with high pancreatic mRNA marker expression along with the ideal islet morphology. However, their functional property was still limited.

Low attachment culture is efficient to generate IPCs from cAD-MSCs
To generate IPCs from cAD-MSCs, microenvironmental manipulating approach was used by suspending the cells onto low attachment culture dishes and maintaining in three-step induction media ( Figure 5A). It was quite interesting that cells formed colony-like structure since day 3 of the induction, and the colonies became denser and bigger along the culture period 8 ( Figure 5B). At day 12, approximately 834 colonies (median) were obtained from 1x10 6 seeding cells ( Figure 5C), and the colony size was varied from <50 µm to > 700 µm ( Figure 5D).
However, it was not statistically significant compared to basal secretion ( Figure 5G).
The results suggested that microenvironmental manipulating approach using low attachment culture was efficient to generate IPCs from cAD-MSCs in term of pancreatic islet characteristics. However, their functional property was still limited.

Notch signaling optimization generates potential cAD-MSC-derived IPCs
According to the results of IPC induction protocol efficiency, it has been suggested that generation of cAD-MSC-derived IPCs using microenvironmental manipulating approach seemed the most efficient protocol in terms of 1) morphological appearance and colony number, 2) pancreatic mRNA marker expression, and 3) functional property. In this regard, Notch signaling optimization was performed for generating the potential cAD-MSC-derived IPCs using protocol mentioned in our previous report 16 .
cAD-MSC-derived IPCs were generated using optimized three-step induction protocol ( Figure 6A) with Notch signaling manipulation using gamma-secretase inhibitor, DAPT, during definitive endoderm induction (DAPT-A) ( Figure 6B) or pancreatic endoderm/progenitor induction (DAPT-B) ( Figure 6C). The results showed that, in all conditions, cells started colony 9 formation since day 3 post-induction, then colony size and number were increased during the induction period ( Figure 6D). Total colony counts (median) were 834, 691.5, and 504 colonies per batch (1x10 6 seeding cells) for control, DAPT-A, and DAPT-B, respectively ( Figure 6E). It seemed that DAPT-B delivered more proportion of small-size colony (<50 µm and 50-100 µm), but statistical difference was not recognized due to variation among groups ( Figure 6F).
Pancreatic mRNA analysis illustrated that cAD-MSC-derived IPCs from DAPT-B condition significantly showed lesser degree of pancreatic endoderm marker (Pdx1) and pancreatic beta-cell markers (Isl-1, Maf-A, Glut-2, and Insulin), comparing with those from DAPT-A condition ( Figure 7A and B). However, alpha-cell hormonal marker (Glucagon) of DAPT-B group was much lower than that in DAPT-A group. Glp1r was downregulated in all conditions ( Figure 7C). Interestingly, analysis of Notch target genes, Hes-1 and Hey-1, showed that DAPT-B group showed significant upregulation of both genes comparing with others ( Figure 7D). Functional testing showed that cAD-MSC-derived IPCs from DAPT-B condition yielded highest basal C-peptide secretion as well as the higher glucose-responsive C-peptide secretion upon low (5.5 mM) and high (22mM) glucose stimulation, comparing with control and DAPT-A groups. It should be noted that, due to variation within group, statistical difference within each group was not found ( Figure 7E).
Taken together, the results suggested that cAD-MSC-derived IPCs could be efficiently generated using microenvironmental manipulating approach with Notch optimization. The obtained IPCs from Notch inhibition during pancreatic endoderm/progenitor induction showed pancreatic islet/beta-cell characteristics and positive trend of functional property.

Discussion
As the proof-of-concept evidences for treating diabetes by regenerative therapy have been reported in human and animal models 9,17-20 , MSCs have been proposed as one of the promising resources for generating clinical applicable IPCs [21][22][23][24] . In this study, the pancreatic differentiation potential of cBM-MSCs and cAD-MSCs was evaluated aiming for determining the feasibility of IPC formation in vitro and the potential of their clinical application. The cBM-MSCs and cAD-MSCs were isolated, cultured, and expanded using previous published protocols [25][26][27][28] . Their characteristics were similar as described in previous reports including fibroblast-like structure, mRNA expression related to stemness and proliferation, MSC-related surface marker expression, and osteogenic differentiation potential 26-30 . It should be noted that the expression of Cd73 in both MSCs was relatively low as mentioned in previous report 31 . This evidence supported the consistency of the cMSCs' properties used in this report.
Here, we illustrated that cBM-MSCs and cAD-MSCs could be differentiated toward pancreatic lineage in vitro. However, each cell type contained different pancreatic differentiation potential and required a tailor-made induction technique. For IPC generation by cBM-MSCs, it has been shown that microenvironmental manipulating approach with low attachment culture (2D culture) could not produce an islet-like cell aggregate in vitro, but it required 3D culture technique for generating and maintaining the colony-like structure of IPCs. By using hangingdrop culture technique, cBM-MSCs formed cell aggregates since day 3 post-induction, then size of the colony was increased along with the expression of pancreatic mRNA markers. Further experiment showed that Matrigel ® -embedded culture of the colonies derived from hanging-drop culture could give a dense colony structure and higher levels of pancreatic marker expression.
Previous publications reported that small molecule induction could imitate the environment during pancreatic endocrine development 10,16,43,51-56 . Generally, an in vitro pancreatic differentiation from SCs could be categorized into 6 differentiation stages: pluripotent/multipotent SCs, mesendoderm, definitive endoderm, pancreatic endoderm, pancreatic endocrine, and pancreatic beta-cells/IPCs 15,57 . In this study, activin A was used to mimic the effects of endogenous noggin for shortcutting the definitive endoderm establishing step as described in previous reports 32, [57][58][59][60][61][62] . It was quite interesting that maintaining cBM-MSCs with pancreatic induction media in low attachment culture was unable to form colony-like structure which is the natural pancreatic islet topology and crucial for an in vitro pancreatic differentiation 16,32,59,63-65 . Therefore, the 3D culture condition using hanging-drop and Matrigel ® -embedded culture techniques were used for generating the cBM-MSC-derived IPC colony. It has been shown that hanging-drop culture was an efficient technique for embryoid body/cell colony formation in vitro [66][67][68][69] along with the natural/synthetic hydrogel-embedded culture that was one of the effective culture techniques used for organoid formation and expansion [70][71][72][73][74] . In this study we demonstrated the successful IPC colony formation by these two culture techniques. However, it was quite difficult to collect and expand the IPC colonies since colony maintaining and medium changing for hanging-drop culture were time-consumed. In addition, treating the Matrigel ® -embedded colonies with hydrogel digesting solution (Cell Recovery Solution ® ) caused colony dissociation. Further functional assay could only be performed for IPC colonies derived from hanging-drop culture and found that the obtained IPC colonies could basally secrete C-peptide but not a significant response to glucose stimulation.
Additional genetic manipulating approach was performed and showed that overexpression of PDX1 at MOI 20 could enhance pancreatic beta-cell marker expression but was unable to produce 3D IPC colony. It has been suggested that the promising regenerative therapy for diabetes relies on the availability and the potential of stem cells used for generating PPs or IPCs, the efficiency of induction protocol, and the potential application on further established transplantation platform.
One of the potential transplantation platforms is cell or colony encapsulation which requires the 3D colony structure of the IPCs that can be harvested after an in vitro production. This encapsulation platform can support and immobilize IPC colonies with the immunoisolating property against host immunity 15 . By comparing the potential clinical application, it seemed that 14 cBM-MSC-derived IPCs showed less potential due to the complicated and time/labor-consumed induction protocol. Therefore, cAD-MSC-derived IPCs were further optimized.
Various factors and signaling have been studied for the potential effects on IPC generation in vitro. In this regard, Notch signaling was of interest due to its significant effect during pancreatogenesis both in vivo and in vitro [82][83][84][85] . cAD-MSC-derived IPCs were generated using optimized three-step induction protocol with Notch signaling manipulation by gammasecretase inhibitor, DAPT, during definitive endoderm or pancreatic endoderm/progenitor induction. We found that Notch inhibition during pancreatic endoderm/progenitor induction benefited the cAD-MSC-derived IPC production in terms of high basal C-peptide secretion and positive trend of glucose-responsive C-peptide secretion. These findings were correlated with previous studies that Notch signaling played a biphasic role in pancreatogenesis during embryonic development. Downregulation of Notch is required for pancreatic endoderm commitment and Pdx1-postive pancreatic precursor expansion, while Notch upregulation is crucial for late-state pancreatic maturation [84][85][86] . Our previous study also showed that inhibition of Notch during pancreatic endoderm induction by human dental pulp stem cells (hDPSCs) resulted in high number of IPC colony production with high expression of PDX1, whereas the inhibition during maturation stage caused the impairment of glucose-responsive C-peptide secretion 16 .
During pancreatogenesis, endocrine precursors formed clusters which allowed cell-to-cell contact and the interaction so called "lateral inhibition". This led to the activation of Notch signaling and the regulation of endocrine fate descended from Pdx1-positive progenitors [87][88][89] .
Previous studies have confirmed the involvement of Notch signaling during endocrine progenitor fate commitment toward one of pancreatic endocrine subtypes (beta-or alpha-cells) [89][90][91] . The inhibition of Notch by HES1 shRNA could induce the redifferentiation of expanded human betacell-derived cells following with the significant expansion of beta-cell in vitro and the upregulation of beta-cell-related genes 92 . However, overactivation of Notch could limit the differentiation capability toward fully matured IPCs by inhibiting the expression of "predifferentiation" gene by pancreatic progenitors 93 . These evidences also supported our findings that the cAD-MSC-derived IPCs could be generated in vitro, and the selective Notch signaling manipulation played the beneficial roles on colony production, pancreatic marker expression, and functional property.

Conclusion
In conclusion, we illustrated that cMSC-derived IPCs could be generated from cBM-MSCs and cAD-MSCs in vitro. However, these two cMSCs contained different pancreatic differentiation potential and required specific induction techniques. Further studies focusing on maturation and transplantation platform will fulfill the production of clinical applicable cMSCderived IPCs.

Cell isolation, culture, and expansion
All protocols were conducted in accordance with guidelines and regulations approved by

Characterization of cBM-MSCs and cAD-MSCs
The

IPC induction by microenvironmental manipulation
In this regard, three-step induction protocol modified from previous published reports was In this regard, 100-150 µL of hydrogel and induction medium mixture (1:1) was used for forming a dome-like structure onto each well of 24-well culture plate (Corning). Cell Recovery Solution ® was used for gel digestion.

IPC induction by genetic manipulation
Overexpression of PDX1 by lentiviral vector was used for genetic manipulating approach.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)
RT-qPCR was used for mRNA analysis. The total RNA was collected using TRIzol-RNA  Supplementary Table S1. 20

Functional analysis for IPCs
Glucose-stimulated C-peptide secretion (GSCS) was used for functional analysis of IPCs.

Statistical analysis
The results were illustrated as whisker and box plot (N=4). Statistical analysis was determined using SPSS statistics 22 software (IBM Corporation, USA). Mann-Whitney U test was used for comparing two independent samples, while Kruskal Wallis test and pairwise comparison were used for three or more group comparison. The significant difference was considered when p-value < 0.05.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.