Serum deprivation limits loss and promotes recovery of tenogenic phenotype in tendon cell culture systems

Abstract Current knowledge gaps on tendon tissue healing can partly be ascribed to the limited availability of physiologically relevant culture models. An unnatural extracellular matrix, high serum levels and random cell morphology in vitro mimic strong vascularization and lost cell elongation in pathology, and discord with a healthy, in vivo cell microenvironment. The thereby induced phenotypic drift in tendon‐derived cells (TDCs) compromises the validity of the research model. Therefore, this research quantified the extracellular matrix (ECM)‐, serum‐, and cell morphology‐guided phenotypic changes in tendon cells of whole tendon fascicle explants with intact ECM and TDCs cultured in a controlled microenvironmental niche. Explanted murine tail tendon fascicles were cultured in serum‐rich or serum‐free medium and phenotype was assessed using transcriptome analysis. Next, phenotypic marker gene expression was measured in in vitro expanded murine tail TDCs upon culture in serum‐rich or serum‐free medium on aligned or random collagen I patterns. Freshly isolated fascicles or TDCs served as native controls. In both systems, the majority of tendon‐specific genes were similarly attenuated in serum‐rich culture. Strikingly, 1‐week serum‐deprived culture—independent of cell morphology—converged TDC gene expression toward native levels. This study reveals a dynamic serum‐responsive tendon cell phenotype. Extracting fascicles or TDCs from their native environment causes large changes in cellular phenotype, which can be limited and even reversed by serum deprivation. We conclude that serum‐derived factors override matrix‐integrity and cell morphology cues and that serum‐deprivation stimulates a more physiological microenvironment for in vitro studies.


| INTRODUCTION
Healthy tendon tissue is poorly vascularized, 1-3 and characterized by a network of elongated fibroblasts, named tenocytes, embedded in an anisotropic extracellular matrix (ECM) mainly composed of collagen I. 4 In response to physiological loading, tenocytes remodel the tissue based on functional mechanical demands. However, tendon tissue overloading or underloading can induce remodeling toward a pathological state, 4,5 which is referred to as "tendinopathy." Tendinopathy is characterized by drastic structural changes: anisotropy in the ECM is lost, 3,6 the ratio of collagen I to collagen III decreases while proteoglycan content increases, [6][7][8] mechanical properties are compromised, 4 vascular and neuronal ingrowth potentially cause pain, 2 and cell numbers are increased, whereas cell alignment and elongated morphology are lost. 7 The concomitant shift of the predominantly tenogenic cell population toward other mesenchymal lineages [8][9][10][11] decreases the functional remodeling capacity of tendon and aggravates the disorder. 12,13 Tendinopathy accounts for 30%-50% of sports-related injuries 5 and nearly 30% of musculoskeletal issues-related medical visits. 14 Aside from physiotherapy and analgesic treatment, tendon disorders are hardly treatable due to gaps in knowledge concerning tissue healing and remodeling. 1 Tendon culture model systems such as tissue explantswith a native ECM as key tissue feature-and in vitro cells, are widely used due to their practicality and to gain a better understanding of the fundamental disease progression and healing mechanisms. 15,16 However, culture systems using tendon-derived cells (TDCs) with a reduced and therefore more controllable cell environment 15 [18][19][20] Limiting phenotypic drift of tendon cells in culture has been tried before 15 by varying biophysical (eg, impose cell morphology using anisotropic substrates 12,[21][22][23][24][25] or mechanical loading 20 ), biochemical (eg, partial oxygen pressure or temperature 17 ), and biological factors (eg, hormones, growth factors or cytokines 26 ), while the precise combination, dosage and time-dependent administration of various tenogenic stimulators remains to be determined. Nonetheless, blood serum, which provides a complex, poorly defined mixture of bioactive molecules, is widely used for in vitro TDC culture to increase cell proliferation and activity. This however, conflicts with the quiescent nature of healthy tendon cells, 1,27 and tendon-specific phenotype seems to be maintained in low rather than high serum concentrations. 18,26 However, the effect of serum-supplementation, tissue integrity cues (cell-matrix contacts) and elongated cell morphology on the differential cellular response, and phenotypic drift of ECM-embedded tendon cells and expanded TDCs has to our knowledge never been systematically quantified with respect to native tissue. 19 Examining the potential of these factors to reverse phenotypic drift in TDCs is key to increase the physiological relevance of tendon culture models.
We, therefore, investigated phenotypic drift in tendon cells away from freshly isolated tissue and cells by separately controlling for complexity of the ECM niche and tendinopathic features of the microenvironment (serum and cell morphology). First, we used transcriptome analysis to screen for tendon-relevant genes and quantify culture-driven phenotypic drift in whole tendon explants-with a complex and intact ECM-upon serum-free and serum-rich culture.
Subsequently, we validated serum-dependent phenotypic drift in expanded TDCs and tested the potential for reversibility, by culturing TDCs in serum-free or serum-rich medium on random and aligned collagen I substrates-with a limited ECM complexity and controlled cell morphology. Their phenotypic gene expression was assessed and compared to freshly isolated TDCs. Hereby we highlighted the influence of the culture environment on the dynamic tendon cell phenotype and contributed to the increased physiological relevance of different in vitro model systems.

| Harvesting and culture of tendon fascicles
All experiments on tendon fascicle explants were ethically approved by the Cantonal Veterinary office of Zurich (permit number ZH265/14). In total, nine C57BL6/J mice, 11 to 12-week-old, were euthanized and tail tendon fascicles were isolated as described before. 12 Briefly, the tail tip was clamped with locking forceps and bent. The tendon fascicles were pulled out, hydrated in phosphate-buffered saline (PBS) and cut off.
Freshly isolated fascicles served as native control immediately after extraction. Explanted fascicles were cultured deprived from load at standard cell culture conditions (37°C, 5% CO 2 ) in high-glucose Dulbecco's Modified Eagle's Medium (HG-DMEM; Sigma D6429) with or without 10% fetal bovine serum (FBS) (Gibco 26140079) for 6 days (Figure 1), as previously described. 17 2.2 | RNA sequencing and data analysis RNA sequencing and data analysis were performed as previously described. 17 In short, approximately 20 fascicles per animal (representing n = 1) were snap-frozen in liquid nitrogen and pulverized  28 within the data analysis framework SUSHI. 29 Raw read counts were normalized using the quantile method, and differential expression analysis was performed using the DESeq2 package. 30 Only significantly and differentially expressed (upregulated/ downregulated) genes between a serum condition and native were considered and were defined at the cut-off values of |log 2 (fold change)| > 1 and P < .05 ( Figure 3C). Exclusively differentially expressed genes met these requirements in one experimental group and not in the other. Subsets of significantly and differentially expressed genes were functionally enriched within the gene ontology domain "Biological process" using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

| Harvesting and culture of TDCs
TDCs were harvested from the tails of thirteen 7 to 9-week-old C57BL/ 6J mice that were euthanized as untreated control for other, unrelated experimental studies. When all tissues, as described in the respective permits, had been harvested, the tails were collected, and tendon fascicles were isolated as described above. The ECM was dissolved using sterile-filtered 3 mg/mL collagenase IV (Gibco 17104019) in PBS for strates and grown to sub-confluency in complete growth medium (referred to as day 0; Figure 1). At day 0, the culture medium was switched to complete growth medium, with or without 10% FBS for 7 days. To differentiate the effect of serum-free medium from culture duration while preventing over-confluency when using FBS, cells were not F I G U R E 1 Schematic overview of experimental setup. Complete fascicles were cultured for 6 days in serum-rich (+FBS) or serum-free (−FBS) medium. mRNA was sequenced, cell viability was determined, and both readouts were compared to freshly isolated fascicles ("native fascicles"). TDCs were expanded in +FBS medium, and at sub-confluency in passage 4 (day 0) medium was switched to +FBS or −FBS medium. Gene expression, cell viability, and functional intercellular communication were monitored over time, and gene expression was compared to TDCs directly after fascicle digestion ("native TDCs"). 2D, two-dimensional; ECM, extracellular matrix; FBS, fetal bovine serum; mRNA, messenger RNA; qPCR, quantitative polymerase chain reaction; TDCs, tendon-derived cells [Color figure can be viewed at wileyonlinelibrary.com] VAN  consecutively cultured in serum-rich medium for 7 days, but medium was re-supplemented with FBS for 2 days after 7 days serum deprivation ( Figure 1).

| Microcontact printing
Collagen I patterns for controlling cell morphology and orientation were created using microcontact printing as described before. 31 Briefly, 15 × 15 mm polydimethylsiloxane (PDMS; Dow Corning Sylgard 184) stamps were made, with either 10 µm wide lines and 10 µm spacing, or a "fishing net" structure with 5 µm wide lines and 10 µm spacing ( Figure 2A1-B1). These corresponded to aligned and random cell morphologies, respectively, with similar collagen I-covered surface areas for cell attachment in both patterns (±50%).
Substrates were prepared by spin-coating PDMS on coverslips, and curing overnight at 65°C. In all, 50 µg/ml collagen I in PBS was adsorbed onto the stamps for 1 hour, and meanwhile the substrates were hydrophilized with UV-Ozone for 8 minutes. The stamps were blow-dried with compressed air, gently pressed onto the substrates, and collagen was transferred for 15 minutes. Stamps were removed, substrates were washed 3× with PBS, and nonprinted areas were blocked with 10 mg/mL Pluronic-F127 (Sigma P2443) in PBS for 5 minutes. Finally, the substrates were washed for 5 minutes with PBS and stored in PBS at 4°C until further use.

| Quantitative polymerase chain reaction
TDCs (pooled from two mice) seeded on the microcontact printed patterns were grown to sub-confluency in complete medium for 3 days. This moment of sub-confluency is referred to as day 0 after which the culture medium was switched to serum-free or serum-rich HG-DMEM or low-glucose DMEM (Gibco 22320-022). TDCs were lysed in RLT buffer directly after tissue digestion ("native TDCs") and at days 0, 1, 7, or 9 ( Figure 1), pooling up to four substrates to one sample. Total RNA was isolated using the RNeasy mini Kit (Qiagen 74106) and cDNA was synthesized using Moloney Murine Leukemia Virus reverse transcriptase (Invitrogen 28025-013). Quantitative polymerase chain reaction (qPCR) primer sequences were obtained from literature or Primer-BLAST, 32 checked for specificity using Primer-BLAST, and for proper efficiency using positive control mRNA dilution series. Sequences of primer sets (Sigma-Aldrich) that passed all tests are listed in Table 1, and corresponding details are given in Supporting Information Table S1.
For gene expression analysis, cDNA samples were diluted 60× in ddH 2 O, and cDNA amplification was measured in the CFX 384 Thermal Cycler (BioRad) for 40 cycles, using corresponding iQ SYBR Green Supermix (BioRad 1708886). C t -values were normalized to the mean C t -values of housekeeping genes Rpl4 and Rps29 and native TDCs ( Figure 1).

| Cell viability
Cell viability was assessed from fascicles of six different mice as previously described, 17 and in adherent TDCs (pooled from nine mice). A detailed description of both methods is provided in the Supporting Information.

| Fluorescence recovery after photobleaching
Intercellular communication was examined on randomly oriented TDCs (pooled from two mice) using fluorescence recovery after photobleaching (FRAP). 33 Details are provided in the Supporting Information.
F I G U R E 2 TDC morphology and orientation were controlled using microcontact printed collagen I patterns. Schematic of aligned (A) and random (B) microcontact printing patterns (1), with collagen I printed surfaces in gray and Pluronic-F127-coated surfaces in black. Microcontact printed substrates with collagen I fluorescently stained in green (2) and resulting phase contrast microscopy images of seeded TDCs (3). All sizes are given in micrometers. TDCs, tendon-derived cells [Color figure can be viewed at wileyonlinelibrary.com] 3 | RESULTS

| Phenotypic drift of cells in tendon explants is more pronounced in serum-rich medium
To evaluate the impact of serum on the phenotypic drift of tendon cells within their native ECM, murine tail tendon explants were cultured in serum-free or serum-rich medium. Differential gene expression of tendon explants was assessed between native, freshly isolated, and cultured tail fascicles ( Figure 1). The number of differentially expressed genes compared to native was higher in tendon fascicles cultured with serum than in the serum-free condition at all cut-off values ( Figure 3A). Based on expression of the 100 most active genes in native tendon tissue (Supporting Information Table S2), drastic phenotypic drift was observed in both the serum-free and serum-rich condition ( Figure 3B). Strikingly, most of the tendon-specific markers (Tnmd, Scx, Col1a1, Col1a2) were downregulated exclusively in the serumrich condition ( Figure 3D and Supporting Information Table S3).
This suggests a strong phenotypic drift away from the tenogenic lineage due to the serum-rich environment. Genes upregulated exclusively in the serum-rich condition comprised several vasculature-(eg, Wnt7b, Hif1a), immune system-and inflammatoryassociated markers (eg, Nfkb2, Il1a, Il20, Cxcl1, Hmgb2, Mmp1) and interestingly also the tendon marker Tenascin C (Tnc) (Supporting Information Table S3). Gene ontology analysis of these genes further indicated activation of processes related to proliferation, angiogenesis, inflammation, and immune response (Supporting Information Table S4). Genes regulated exclusively in the serum-free condition included ECM (modulating) proteins (eg, Col3a1, Col6a5, Acan, Lox, Postn) and growth factors (eg, Tgfa, Fgf2) (Supporting Information Table S5) involved in different biosynthetic processes (Supporting Information Table S6). Overall, 1428 genes were significantly regulated, exclusively in the serum-rich condition and 952 genes exclusively in the serum-free condition, compared to native, but the majority of the genes significantly regulated relative to native (3775) were not exclusive, and thus in both the serum-rich and serum-free conditions. However, genes deviating less from native in serum-free compared to serum-rich conditions comprised tendon-specific (eg, Dcn, Fmod) and several inflammationassociated (eg, multiple Mmps, Il1b, Il11, Ccl2, Cxcl12) genes ( Figure 3E and Supporting Information Table S7). Cells showed comparable viability in native fascicles and serum-free conditions, whereas cell number and viability were increased after 6 days of culture in the serum-rich group (Supporting Information Figure S1A).
In summary, tendon tissue explant culture drastically changes cellular phenotype, and a serum-rich environment exacerbates this phenotypic drift.
3.2 | Serum deprivation for 7 days nearly restores native levels of phenotype marker genes in TDCs We next aimed at assessing the phenotypic drift in cultured TDCs  Figure S3). Remarkably, Tnmd gene expression increased almost 100-fold upon 7 days serum deprivation.
Strikingly, gene expression levels of Tnc, Dcn, and Acta2 nearly approached levels of day 0 cells within 2 days after switching the medium back to serum-rich in serum-deprived samples. In the same time span, Tnmd dropped 10-fold, compared to day 7 ( Figure 4).

| The effects of morphology and glucose level on TDC phenotype are negligible compared to serum concentration
To assess whether confining TDCs to an elongated morphology further rescues gene expression toward native levels, murine TDCs were seeded on random and aligned microcontact printed collagen I substrates. Figure 2A3-B3 confirms that TDCs seeded on these substrates adapted their morphology to the patterns. All changes in gene expression were shown to be independent of cell orientation (Supporting Information Figure S4).
To rule out any gene modulatory effect of the high glucose level in the culture medium, gene expression levels of in vitro TDCs were assessed in high-and low-glucose DMEM. No phenotypic recovery in low glucose levels was detected (Supporting Information Figure S4).

| Serum deprivation slightly affects cell viability and does not affect intercellular communication
TDC viability in serum-rich and serum-free conditions was assessed.
Number and percentage of live cells dropped approximately 25% upon 7 days serum deprivation, resulting in a similar distribution as in native and serum-deprived fascicles. Remarkably, many cells died when serum was added after 7 days serum deprivation (Supporting Information Figure S1B). | 1567 To rule out differences in cell connectivity, we examined gap junction functionality-an important functional hallmark of tenocytes -by FRAP. We compared TDCs expanded in serum-rich medium (day 0) and after 7 days of serum deprivation on random substrates.
Serum deprivation did not abolish fluorescence recovery in serumdeprived TDCs, indicating that functional intercellular communication was maintained (Supporting Information Figure S5).
Altogether, in these 2D experiments, serum-rich medium induced strong deviations from native gene expression levels, whereas 7 days of serum deprivation appeared to reverse this phenotypic drift, particularly for the tenogenic marker and tendon ECM genes (as classified in Table 1). No large adverse effects on TDC viability or functionality were found. The impact of cell morphology and glucose level appeared negligible in this system (Table 2)

| DISCUSSION
The relevance of in vitro tendon culture models is compromised by phenotypic drift of tendon-derived cells, in response to high serum levels and random cell morphology. 18,[21][22][23] While previously phenotypic properties were compared between cultured cells in different passages 19,21 or various culture conditions within the same passage number, 18,22,23,25 quantification and comparison of phenotypic drift in tendon culture models with respect to native tissue and TDCs was scarcely described. 34 In this study, we explored the phenotypic drift of tendon cells by stepwise reduction of model system complexity using first explanted tendon fascicles with preserved native threedimensional (3D) ECM and then expanded TDCs on 2D substrates.
Using these model systems, we investigated whether native cell F I G U R E 4 Gene expression of expanded TDCs (p4) cultured in serum-rich and serum-free medium on random substrates for 7 days after the moment of sub-confluency (day 0), compared to freshly isolated TDCs (fold change = 1). After 24 h, serum deprivation had a "rescuing" effect on phenotype marker genes (Tnmd, Tnc, Dcn, Acta2) in TDCs that were expanded in serum-rich medium. Tendon-specific phenotypic markers were almost restored to native levels in p4 TDCs after 7 days of serum deprivation, and this effect was counteracted by serum supplementation for 2 days. n = 2.  In both culture models (explants or in vitro TDCs) native gene expression levels were highly similar, and all measured genes showed the same differential gene expression patterns for the serum-free and serumrich conditions compared to the native controls, except for Scx, Acta2, Col1a1, and Fabp4. This indicates that cues from serum-derived factors override matrix-integrity cues for these genes. The fascicle explant is expected to provide lower substrate stiffness than in vivo tendon, whereas the thin PDMS film on glass in vitro is expected to be stiffer.
Proposedly, this relates to the mechanosensitive nature of certain differentially regulated genes in the explant vs the TDC culture, like Scx, 35 which was downregulated in the free-floating fascicle and constantly expressed in vitro. Similarly, Acta2 equilibrates cytoskeletal tension to substrate stiffness, 36  We speculate that high serum levels in vitro, resembling neovascularization in vivo, may induce a switch in tenocytes toward a more active, metabolic, inflammatory phenotype (supported by our ontology analysis), analogous to tendinopathy, and similar to cells that crossed the "metabolic tipping point," which recruit extrinsic tissue compartments to heal the damaged tissue. 1 Strikingly, Tnmd-the gene that was most affected by serum-has an anti-angiogenic role in (healthy) tendons. 44 This implies a positive feedback loop of low tissue vascularization promoting Tnmd expression, which in turn prevents angiogenesis. On the other hand, Tnc might be a pro-angiogenic factor, which was increased in the vascularized (serum-rich) model and is similarly found in blood vessel-infiltrated tendon tissue. 3 The serum-induced phenotypic drift during in vitro cell expansion appears to be reversible by serum deprivation. This is a simple and effective procedure to obtain a large number of tenocyte-like cells for in vitro studies, compared to mechanical, topological, biological, or biochemical stimuli that are currently applied to induce tenogenic differentiation of stem cells or maintaining phenotype in primary TDC expansion. 15,26 Previously, cell alignment has been shown to increase tenogenic markers (eg, Tnmd: 80-to 200-fold) during stem cell differentiation, 22,23,45 although this effect was not detected consistently. 24 However, the increase in tenogenic markers upon cell alignment in primary TDCs rarely exceeds 5-to 10-fold, 21,25,45 similar to our results.
The options and boundaries of serum deprivation as experimental phenotypic drift-limiting method need to be further determined. Firstly, prominent tendon markers Col1a1 and Tnmd did not completely recover to native levels in this research, but expression of these genes can potentially be stimulated biochemically with VAN | 1569 ascorbic acid or additional bioactive stimuli. 46 Secondly, all in vitro readouts in this research were performed at p4. However, phenotypic drift may also be reversible at higher passages, which would be ethically favorable due to a maximized cell yield per animal, reducing the number of animals sacrificed for in vitro tendon research. Thirdly, with qPCR and RNA sequencing, the most important readouts in this research focused on gene expression levels. These are indicatorsbut no direct measures-for protein production, cell behavior, and phenotype. 47 However, "alterations in (…) gene expression" are explicitly mentioned in the definition of phenotypic drift, 15 justifying the choice for these readouts. It is important to consider that gene expression analysis by transcriptomics and qPCR requires a certain level of mRNA production to exceed the detection limits. Serum starvation decreases the production of mRNA 27 and therefore requires high numbers of cells in order to guarantee feasibility of the readout. The clinical implications need to be explored as well. Despite the fact that rodent animal models of (tendon) tissue repair can deviate from human system behaviors, 48 for instance in immune system involvement during healing, 49 an impressive phenotypic plasticity is observed in both murine and human TDCs. 50 This arguably gives confidence for the clinical translatability of the presented findings in the search for an established tendinopathy treatment. Improving properties of tendinopathic cells and tissues by controlling vascularization -analogous to serum deprivation in this research promoting a healthier tenocyte-like phenotype-might be a potential path to follow in the quest for a clinical tendinopathy treatment aiming at (partial) tissue recovery.
In summary, this study quantified the phenotypic drift away from native gene expression in different tendon culture models revealing an extremely dynamic phenotype of tendon cells. It provided a valuable set of tendon maker genes that contributed to the further characterization of the underresearched tendon tissue. We conclude that serum supplementation exacerbates phenotypic drift in tendon tissue explant and in vitro cell culture systems, and overrides cues from ECM integrity, cell orientation, and morphology. Serum deprivation limits phenotypic drift in cells of explanted tendon tissues, reverses it after in vitro expansion, and therefore represents a method to potentially increase physiological relevance of in vitro studies.