The effect of hypoxia on the proteomic signature of pig adipose-derived stromal/stem cells (pASCs)

Human adipose-derived stem cells (ASCs) have potential to improve wound healing; however, their equivalents from domestic animals have received less attention as an alternative cell-based therapy for animals or even humans. Hypoxia is essential for maintaining stem cell functionality in tissue-specific niches. However, a cellular response to low oxygen levels has not been demonstrated in pig ASCs. Hence, the goal of our study was to characterize ASCs isolated from the subcutaneous fat of domestic pigs (pASCs) and examine the effect of hypoxia on their proteome and functional characteristics that might reproduce pASCs wound healing ability. Analysis of immunophenotypic and functional markers demonstrated that pASCs exhibited characteristics of mesenchymal stem cells. Proteomic analysis revealed 70 differentially abundant proteins between pASCs cultured under hypoxia (1% O2) or normoxia (21% O2). Among them, 42 proteins were enriched in the cells exposed to low oxygen, whereas 28 proteins showed decrease expression following hypoxia. Differentially expressed proteins were predominantly involved in cell metabolism, regulation of focal and intracellular communication, and attributed to wound healing. Functional examination of hypoxic pASCs demonstrated acquisition of contractile abilities in vitro. Overall, our results demonstrate that hypoxia pre-conditioning impacts the pASC proteome signature and contractile function in vitro and hence, they might be considered for further cell-based therapy study on wound healing.

via up-regulation of vascular endothelial growth factor (VEGF) 18 . Frazier et al. 19 showed that pre-conditioning of human ASCs with low oxygen reduced secretion of extracellular matrix (ECM) proteins and type 2 cytokines (interleukin [IL]-6, IL-13, monocyte chemoattractant protein [MCP]-1, CD40 ligand) involved in the initiation of fibrosis. These results suggest that enhancing these ASCs characteristics with hypoxia may increase their regenerative potential and thus augment healing following cell delivery to the injured tissues.
The hypothesis that hypoxic pre-conditioning of ASCs provides functional enhancement 15,18,20 and makes the cells more attractive for clinical translational science has encouraged an interest in characterizing the hypoxiaresponsive characteristics of ASCs on the proteomic level 21 . This -omics technique might provide a tool for predicting ASCs safety, feasibility, and efficacy . Therefore, clarification of hypoxia-altered proteins together with follow-up functional studies are needed prior to clinical ASCs delivery to damaged tissues, particularly when stem cells from non-human donors are considered for testing.
Adipose tissue stem cells from livestock animals have gained increased attention in recent years due to their promising role in developing regenerative medicine protocols that can be utilized in veterinary medicine 22 and humans. Of particular interest are large animals such as pigs that share multiple similarities with human and are necessary to conduct pre-clinical studies. To date, adult stem cells have been identified in a variety of pig tissues, including bone marrow 23 , endometrium 24 , ovary 25 , and abdominal fat 26 . However, only a limited number of studies have investigated porcine stem cells in the context of cutaneous wound healing 27 . Moreover, there are no data related to effect of hypoxia on pig ASCs proteome and the cell function.
The present study was undertaken to characterize pASCs from the subcutaneous fat of prepubertal domestic gilts. We evaluated the effect of hypoxia (1% O 2 ) on pASCs proteome using two-dimensional difference gel electrophoresis (2D-DIGE) to quantify proteins and matrix-assisted laser desorption/ionization (MALDI) mass spectrometry to identify proteins with different abundances in hypoxia-treated pASCs. Finally, in vitro functional studies were performed to determine the impact of hypoxia on pASC properties associated with skin wound healing.
Proteomic analysis of pASCs exposed to hypoxia (1% O 2 ) or normoxia (21% O 2 ). The proteomic data showed that 93 protein spots, corresponding to 70 proteins, differed significantly (p < 0.05) between pASCs cultured for 24 h under hypoxia (1% O 2 ) or normoxia (21% O 2 ). Sixty two spots representing 42 proteins were enriched in cells exposed to low oxygen, whereas 31 spots corresponding to 28 proteins showed decreases upon hypoxic exposure (see Supplementary Table S2). Identified proteins were 1.03-1.89-fold regulated by oxygen tension. The largest changes were observed in response to hypoxia for N-myc downstream regulated 1 (NDRG1) protein (fold change -1.89) and translocon-associated protein subunit delta isoform X2 (SSR4; fold change -1.78, see Supplementary Table S2). The group of 27 proteins corresponding to 31 spots that demonstrated fold-enrichment ≤ 1.10 were excluded from further analysis (see Supplementary Table S2 and Supplementary Figure S1).
To characterize functionality and define the types of proteins differentially expressed between pASCs treated with hypoxia or normoxia, their genes were classified into four main GO categories including "Biological processes," "Molecular function," "Cellular component," and "Protein class" using the PANTHER Classification System (Fig. 2, distribution of hypoxia up-and down-regulated proteins between individual GO categories are presented in Supplementary Tables S3 and S4). Generally, with respect to "Biological processes, " proteins involved in cellular (37%) and metabolic (24.1%) processes comprised the largest group of proteins affected by hypoxia ( Fig. 2A). Furthermore, according to the classification category of "Molecular function, " 64% of differentially expressed proteins identified by PANTHER Classification System demonstrated catalytic function, whereas 32.0% were found to be engaged in binding (Fig. 2B). In the "Cellular component" category, proteins associated with cell (45.9%) and organelle (16.2%) constituted the largest group of oxygen-sensitive molecules (Fig. 2C). Considering the "Protein class" cluster, many hypoxia-responsive proteins were represented by metabolite interconversion enzyme (55.6%) and cytoskeletal protein (14.8%) (Fig. 2D).
STRING database query (Fig. 3) and subsequent KEGG pathway analysis allowed us to specify processes enriched by differentially abundant proteins. Of the 27 proteins up-regulated upon hypoxia, 26 were identified in the STRING database (Fig. 3A). The proteins that interacted with each other ( www.nature.com/scientificreports/  Table 1, see Supplementary Table S3). Importantly, the majority of hypoxia up-regulated proteins (ALDOC, TPI1, GAPDH, PGK1, PGAM1, ENO1, ENO2, LDHA) were linked with anaerobic metabolism (glycolysis). The second large group of molecules that increased with hypoxia comprised proteins involved in focal adhesion and intracellular communication (MAPK1, ZYX, VCL). With regard to proteins associated with wound healing, we found that prolyl 4-hydroxylase subunit alpha-1 precursor (P4HA1) was increased with hypoxia ( Fig. 3A). In addition, heat shock protein 70.2 (HSP70.2) exhibited increased abundance in low oxygen environment (Fig. 3A).
In the group of hypoxia-down-regulated proteins, there were 14 that were included in the protein network. The most connected proteins (average local clustering coefficient, 0.4; PPI enrichment p value < 5.e−10) participate in glycolysis/gluconeogenesis, tryptophan, pyruvate and carbon metabolism, and biosynthesis of amino acids, as well as degradation of valine, leucine, and isoleucine (Fig. 3B, Table 2, see Supplementary Table S4).
Moreover, down-regulation of phosphoenolpyruvate carboxykinase [GTP] (PCK2) and dihydrolipoyl dehydrogenase (DLD), both of which play roles in the citrate cycle, confirmed that aerobic metabolism is suppressed in hypoxia-treated cells. With regard to proteins critical/fundamental for cutaneous healing, collagen type I chain alpha 2 (COL1A2) was down-regulated following hypoxia (Fig. 3B. Table 2). Among the hypoxia-downregulated proteins, we found heat shock proteins including HSPCB (also known as HSP90AB1), HSPD1, and HSPA9 (Fig. 3B). mRNA level verification of selected hypoxia-responsive proteins confirmed their up-or downregulation under low oxygen (Fig. 3C).

Effect of hypoxia on pASCs functional features.
We performed a series of in vitro studies to define the role of hypoxia on pASCs functional features, particularly those involved in wound healing. Collagen gel contraction assays were performed to assess pASC contractile abilities that are crucial for wound closure in vivo. Hypoxia stimulated pASC-mediated contraction of collagen gels over time relative to cells exposed to normoxia (p < 0.001 for days 1-9, Fig. 4A; see Supplementary Table S5a). Low oxygen gradually promoted pASC contractile abilities between days 0 and 9 with gel size reductions of 61. 22 Table S5b.). Migratory properties of pASCs were examined with scratch assays. Low oxygen exposure had no effect on motility compared to atmospheric oxygen level ( Fig. 4B; see Supplementary Table S6a). Based on the  Table S6b). Because pASCs were pre-treated with mitomycin C, the "wound closure" capacity was exclusively due to cell migration. However, considering that wound healing is also driven by cell proliferation we examined pASC propagation abilities using BrdU incorporation and cell counting. Neither approach showed any effect of hypoxia on pASC proliferation at 24, 48, or72 h compared to normoxic cells (   www.nature.com/scientificreports/ Furthermore, given that collagen deposition is an important hallmark of cutaneous healing which is consistent with our proteomic data (see Fig. 3B. and Table 2), we next examined the impact of hypoxia on pASCs ability to produce collagen type I alpha 2 (Fig. 4E). The results of these analyses showed that there was no change in collagen secretion to the culture media after 24 h of hypoxia ( Fig. 4E; see Supplementary Table S9a and S9b).

Discussion
There has been tremendous progress in pre-clinical studies of human ASCs in the context of their use in regenerative medicine. Although considerable advancement in identification and characterization of animal ASCs has been observed, these cells have not received the attention paid to their human equivalents, so animal-derived ASCs are a relatively new area to explore. The present studies provide a protocol of efficient ASCs isolation from the subcutaneous fat of domestic pigs (pASCs) and their culture-expanded characterization.
For optimal pASCs recovery, we adapted a method originally established to isolate human ASCs 4,19 . Because no criteria have been established to identify porcine stem cells, we facilitated those defined by The International Federation for Adipose Therapeutics and Science (IFATS) and the International Society for Cellular Therapy (ISCT) to identify human ASCs 28 . These criteria were also used in other studies for characterization of pig ASCs derived from visceral 29 and subcutaneous interscapular sites and buccal fat pads 30 . Our results show that pASCs isolated from the subcutaneous fat of prepubertal domestic gilts fulfill the criteria for MSCs with regard to morphology, clonogenicity, differentiation potential and immunophenotype. Flow cytometry analysis confirmed the expression of primary MSC surface markers (CD29, CD44, CD90, CD105) that is with agreement with the studies on human 31 and pig ASCs 32 . Likewise, the presence of small fraction of CD45 positive cells has been demonstrated in numerous independent investigations [33][34][35] .
The major goal of the present study was to examine the effect of hypoxia on the pASCs protein signature with particular attention to proteins associated with wound healing. Because damaged tissues are often characterized by oxygen deprivation, much work has been done to identify strategies that improve ASCs survival in hypoxic environments 15,36 . In this respect, hypoxia pre-conditioning plays a protective role in human ASCs 16 and concomitantly enhances their reparative/regenerative properties 15 .
In our study, we confirmed the observation of Riis et al. 21 that hypoxic preconditioning of human ASCs caused subtle differences in protein expression. Changes in protein abundance ranged between 1.89 and 1.03 in our study. The largest changes were observed in response to hypoxia for N-myc downstream regulated 1 (NDRG1) protein (fold change -1.89, see Supplementary Table S1). Similarly to our study, Riis et al. showed that differences in protein levels between human ASCs cultured in hypoxic and normoxic conditions were less than two-fold for most identified proteins 21 . Interestingly, mass spectrometry analysis of the cells revealed that a relatively small fraction of the proteome (9.6%) is significantly affected by hypoxia 21 , which is comparable to the 9.16% of affected proteins in our study.
A variety of proteins were differentially expressed between hypoxia-and normoxia-exposed cells. However, only a small proportion of these were strictly related to the process of cutaneous wound healing. In particular, low oxygen increased the expression of P4HA1, which catalyzes the formation of 4-hydroxyproline that is essential for the correct folding of triple helical collagen molecules 37 . This is consistent with research on human ASCs that showed increased P4HA1 and P4HA2 levels following 24 h of hypoxia 21 . The authors also demonstrated that hypoxia stimulated the expression of other proteins involved in matrix regulation including COL1A1, COL3A1, www.nature.com/scientificreports/ www.nature.com/scientificreports/ and COL7A1 21 . Surprisingly, in our study we found that although hypoxia increased the levels of P4HA1, it concomitantly led to decreased expression. These findings are in agreement with a report by Frazier et al. 19 that evidenced lower levels of COL1A1, COL1A2, and fibronectin released to CM upon hypoxia exposure of human ASCs. Since overabundance of these proteins were found in multiple models of tissue fibrosis including skin 38 , the hypoxia-induced decrease might suggest that hypoxia pre-conditioning prevents ASCs from excessive ECM deposition. Consistent with the literature, exposure to low oxygen causes cells to switch from oxidative phosphorylation to glycolysis, which becomes the primary method for ATP production 21,39,40 . As expected, we found that the largest group of proteins affected by hypoxia were enzymes in the glycolytic pathway including ALDOC, TPI1, GAPDH, PGK1, PGAM1, ENO1, ENO2, and LDHA. These metabolic alterations provide the energy necessary for cell adaptive responses and thus support cell survival 16 and function 41,42 .
Hypoxia is a natural shock signal to cells, so we put our focus on stress-related proteins. Oxygen deficiency increased the presence of heat shock 70 kDa protein 1B (HSP70.2) in pASCs. HSP70 is a crucial negative regulator of the mitochondrial pathway of apoptosis that is able to block cell death at both pre-and post-mitochondrial levels 43 . Recent studies demonstrated that HSP70 upregulation in induced pluripotent stem cells-derived cardiomyocytes (iPS-CMs) exposed to hypoxia promoted cell survival by suppressing apoptotic pathways 44 . Clinically, this observation implies that increased HSP70 might improve the survival of transplanted stem cell and enhance their therapeutic potential. It was previously shown that high HSP70 levels correlate with proper wound healing, likely by enhancing endothelial cell function, whereas its weak or lack of expression results in chronic wound development 45 . In addition, delivery of RAW264 macrophages pretreated with HSP70 to full-thickness cutaneous wounds in mice accelerated wound closure by stimulating macrophage-mediated phagocytosis of wound debris 46 . Together, these data suggest that cells subjected to hypoxia release specific pro-repair and/or pro-regenerative agents and might be adopted as a "carrier cell system" for delivering molecules that enhance damaged tissue recovery.
The initial expectation was that low oxygen tension may increase the expression of all identified HSPs, as in general they play an important protective role in the cellular response to hypoxia. However, proteomic data revealed that the presence of some of HSPs representatives, such as stress-70 protein, mitochondrial (HSPA9), heat shock 90kD protein 1 beta (HSP90AB1), and mitochondrial 60 kDa heat shock protein (HSPD1), were decreased following hypoxia. Among them, HSP90 exhibited well-documented abilities to maintain stem cell fate 46 . Indeed, the results showed that HSP90 is essential for maintaining pluripotency of mouse embryonic stem cells (ESCs) through regulating the pluripotency factors Oct4 and Nanog. Dramatic decreases in the expression levels of both HSP90α and HSP90β were observed during ESCs differentiation into embryoid bodies. Furthermore, HSP90 inhibition led to lower Oct4 mRNA levels and increased markers for mesoderm lineage 47 . Along this line, the decrease in some HSPs following hypoxia exposure might indicate a reduced stress defense capacity of pASCs and/or an ongoing differentiation process. Although these concepts remain to be elucidated, the increase in pASCs contractility led us to presume that hypoxia improves acquisition of the smooth muscle cell phenotype (See Supplemental Fig. S2). This viewpoint is consistent with the results obtained from human ASCs, showing that low oxygen conditions influenced their differentiation into smooth muscle cells 17 .
With regard to wound healing, the effect exerted by HSP90 remains under debate as the protein has been shown to improve skin cell motility and accelerate wound closure 48 , whereas others reported that overexpression of HSP90 led to excessive collagen synthesis and keloid formation 49 . Altogether, the consequences of decreased HSP90AB1, HSPD1, and HSPA9 in response to hypoxia require further investigations.
The cellular response to hypoxia involved cell components associated with focal adhesion and cell to cell communication. Besides regulating cell morphology and polarity, these structural systems also control cell dynamics essential for multiple physiological processes such as migration, adhesion, ECM remodeling, and signaling [50][51][52] . The present study revealed that hypoxia increased pASC expression of zyxin (ZYX), vinculin (VCL), vimentin (VIM), and caldesmon 1 (CALD1). These results are in agreement with data obtained from human ASCs demonstrating increased integrin alpha 5 (ITGA5) and intercellular adhesion molecule 1 (ICAM1) in response to hypoxia 21 . The increase in proteins responsible for intracellular communication and cells' connections with the extracellular environment suggest that hypoxia-exposed pASCs might exert their effect in injured tissue by replenishing damaged cells. However, this concept merits further studies of pASCs ability to differentiate into the phenotype of injured tissue (e.g., skin) and further in vivo evaluation.
While oxygen deprivation did not affect cell migration, proliferation, or collagen type 1 alpha 2 secretion, hypoxia considerably enhanced pASCs contractile abilities. These findings are consistent with those reported by Wang et al. that demonstrated increased contractility of human ASCs cultured at 5% oxygen 17 . Other investigators have reported that stimulation of processed lipoaspirate cells with the cholinergic agonist carbachol or 60 mM KCl into smooth muscle-like cells differentiated from human ASCs induced their contractile function in standard ambient culture conditions (20% O 2 ) 53,54 . The lack of extrinsic stimuli in culture medium might partially explain the static behavior of pASCs exposed to normoxia. This topic requires further investigation. There is no definitive answer for the mechanisms that regulate pASCs functions in vitro, and further assessments are needed.
To the best of our knowledge, this is the first report demonstrating the effect on hypoxia on the proteome and functional characteristics of pASCs. In veterinary medicine, pASCs xenotransplantation can be an alternative to costly and inconvenient autotransplantation. Currently, ASC-related therapies are largely used to treat musculoskeletal tissue injury in horses 55 and dogs 56 . This trend advocates for the need of animal ASC availability as an off-the-shelf product that will allow vets to use them directly at the point of care. With regard to humans, since wound healing studies have been limited by a lack of relevant in vivo experimental models studies, autotransplanting pig stem cells into damaged tissues can provide a clinically relevant model to evaluate the outcome of allogenic cell grafts. Furthermore, for a number or reasons including size, anatomical, and physiological similarities with humans, the pig is a preferred donor species. The results of preclinical studies of transplantation of pig www.nature.com/scientificreports/ cells (e.g., islets, neuronal cells, or hepatocytes) into non-human primates have encouraged their further use in clinical trials 57 . Given that significant progress has been made in many areas of xenotransplantation, it is likely that pASCs will be considered for cell-based therapies in humans. On the other hand, the wide availability of human ASCs and their well-documented therapeutic properties support the consensus that future application of ASCs in human medicine will be conducted mostly in an autologous form.

Conclusion
In searching for alternative sources of stem cells to facilitate wound healing in animals and/or humans, we isolated and characterized of ASCs from the subcutaneous fat of domestic pigs and showed that they fulfilled the criteria of MSCs. The data demonstrated that hypoxia (1% O 2 ) altered the expression of proteins associated with cell metabolism, ECM components, and intracellular communication. Changes induced by oxygen deficiency might suggest pASCs potential to improve wound healing by preventing excessive collagen deposition and enhancing cell-to-cell communication and ECM interaction. By combining a proteomics approach with functional pASC assessment following hypoxia, we performed the initial step toward qualifying pASCs as attractive candidates in cell-based applications for wound healing therapeutics. However, as present report characterizes pASCs exclusively in the in vitro settings further in vivo studies on animal models of wound healing are required to verify pASCs contribution in this process. Moreover, given that human ASCs exert their effects on wound healing in a paracrine manner, further study of the pASC secretome is necessary to fully understand their specificity. This will require additional studies exploring the mechanism underlying pASC action in injured skin and further pre-clinical evaluation in both rodents and large animal models.

Isolation of pASCs. Subcutaneous fat samples collected from back and leg ham regions of each donor
were obtained post-mortem from domestic gilts (Sus scrofa; n = 10) in a slaughterhouse. pASCs were isolated according to the method describe by Frazier et al. 19 , with modifications. Briefly, the adipose tissue was washed with phosphate-buffered saline (PBS), 70% ethanol, and 10% betadine; minced with scissors; and enzymatically digested with 0.1% collagenase type I (Sigma-Aldrich Co.) in PBS with 100 IU/mL penicillin and 100 μg/mL streptomycin (Sigma-Aldrich Co.) for 3 h at 37 °C. The floating adipocytes were separated from the stromal vascular fraction (SVF) by centrifugation (250×g) for 5 min at 28 °C. The supernatant containing mature adipocytes was discarded, and the remaining pellet identified as the SVF was suspended in Dulbecco's minimum essential medium ( Differentiation. pASCs (n = 4) were seeded in duplicate at the concentration of 5.0 × 10 4 into 24-well plates (p = 1) and cultured until they reached 60-70% confluency. Adipogenic differentiation was induced according to the procedure described previously by Yu et al. 58  www.nature.com/scientificreports/ Growth kinetics. pASCs (n = 9) were seeded in duplicate in 24-well plates at densities of 5.0 × 10 4 , 1.0 × 10 5 , and 2.5 × 10 5 cells/well in ADSC-BM supplemented with 10% FBS, L-glutamine, and gentamicin/amphotericin. At 80-90% confluence, cells were passaged by digestion with 0.05% trypsin, counted, and reseeded. Serial passaging was continued until the harvested cell numbers dropped below the initially plated number. Cell-doubling times (DT) were calculated by using the following formula: DT = CT/ln(Nf/Ni)/ln (2)

Gene ontology (GO) annotation and bioinformatics analyses.
The GI accession numbers of identified proteins were mapped to the UniProtKB database (www.unipr ot.org), and Sus scrofa was selected as the organism. The gene ontology (GO) annotation of differentiated proteins was performed by using online bioinformatics tools of PANTHER Classification System, version 14.1 (released 2019-04; https ://panth erdb.org/). The proteins that differed in abundance between hypoxia-and normoxia-treated pASCs were classified based on the following GO categories: "biological processes, " molecular function, " "cellular component, " and "protein class. " The Kyoto Encyclopedia of Genes and Genomes (KEGG) was used for an annotation of biological pathways involved in the cellular response to hypoxia [60][61][62][63] . The analysis of potential protein-protein interactions were performed with the Search Tool for the Retrieval of Interacting Genes (STRING) online database, version 11.0 (https ://string-db.org) 64 with a medium confidence score cutoff of 0.4. Collagen gel contraction assay. Three-dimensional collagen gels were prepared according to a previously described protocol 65 . Details are provided in Supporting Information Methods.

Validation of proteomic results by real-time polymerase chain reaction (PCR
In vitro wound migration assay. Scratch assays were performed using a previously described procedure 66 .
Details are provided in Supporting Information Methods.
Proliferation assay. Proliferation of pASC (p = 1) was assessed by incorporation of 5′-bromo-2′deoxyuridine (BrdU, BD Pharmingen) and by counting using automatic cell counter (Life Technologies by Thermo Fisher Scientific). Cells (n = 4) were plated on 60-mm Petri dishes at a density of 2.0 × 10 6 cells and cultured in ADSC-BM with 10% FBS, L-glutamine, and antibiotics. At 50-60% confluence, the culture medium was replaced, and plates were continually cultured at 37 °C under normoxia (21% O 2 , 5% humidified CO 2 ) or hypoxia (1%, O 2 , 5% humidified CO 2 ) conditions for 24, 48, and 72 h. The cells were then washed with PBS, and BrdU was added at 10 μM for 16 h. Subsequently, pASCs were detached with 0.025% trypsin with EDTA and counted. Then, the cells were fixed, treated with DNase to expose BrdU epitopes, and labeled with FITCconjugated anti-BrdU antibodies according to manufacturer's instruction (FITC BrdU Flow Kit BD Pharmingen). Flow cytometry assays were performed using a BD LSRFortessa Cell Analyzer flow cytometer (Becton Dickinson) and BD FACSDiva v6.2 Software (Becton Dickinson). The total number of cells analyzed for each sample was 20,000.
Enzyme-linked immunosorbent assays for Col1a2. The concentrations of collagen type 1α2 in conditioned media (CM) collected from pASCs (n = 10) cultured under hypoxia or normoxia were determined using Col1α2 enzyme immunoassay kits (Cloud-Clone Corp.) according to the manufacturer's protocol. The standard curve for Col1a2 ranged from 3.12 to 2000 pg/mL. The intra-and inter-assay coefficients of variation (CV) were < 10% and < 12%, respectively. www.nature.com/scientificreports/ Statistical analysis. In order to conduct statistical analyses of pASCs growth kinetics (Fig. 1B), contractile abilities (Fig. 4A), migration (Fig. 4B), proliferation (Fig. 4C, D) and collagen I alpha 2 secretion (Fig. 4E) linear mixed-effects models were used, in which individuals were included as random intercept and different factors and their interactions were included as fixed effects. In case of insignificant random effects, linear models were used instead. As long as models were valid, based on them lsmeans were calculated and compared (for mixed-effects models Kenward-Roger method for degrees of freedom was used; Tukey p-value adjustment used where appropriate). All calculations were performed in R (ver. 3

Data availability
The data that support the findings of this study are available from the corresponding author JB upon reasonable request.