Tuning the 3D microenvironment of reprogrammed tubule cells enhances biomimetic modeling of polycystic kidney disease

.


Introduction
The incidence of kidney disease is increasing globally, and there is an urgent demand for new diagnostic and therapeutic options. Kidney diseases can be modeled in animals, but for many applications, such as biochemical analysis, screens or drug candidate identification, in vitro models are more suitable to uncover pathophysiological mechanisms and identify new therapeutic targets. Primary renal derived cells are difficult to keep differentiated in culture and stable cell lines tend to lose their kidney-specific properties after immortalization [1][2][3][4]. An alternative cellular model are induced renal tubular epithelial cells (iRECs), obtained by directly reprogramming mouse and human fibroblasts using 4 transcription factors (Hnf1b, Pax8, Hnf4a, Emx2) [5]. iRECs can be cultured for months without changes to their differentiation state, and they retain many functional properties of renal tubular cells [5,6]. iRECs have been mainly employed in 2D assays, but their behavior in 3D microenvironments and their potential to model renal diseases have not yet been explored [5].
In contrast to conventionally plated culture techniques, 3D culture allows cells to freely grow and migrate without being constricted to a single plane, and significantly changes cellular morphology, cell-cell and cell-matrix interactions [7][8][9]. These effects are mediated by transcriptional changes, and thus provide stimuli that enhance physiological properties of cells. For instance, 3D cultured hepatocytes re-acquired the ability to store glycogen, produced bile salts, and showed enhanced expression of liver-specific proteins in contrast to conventional 2D cultures [10]. Similarly, by tuning the 3D microenvironment, the transcriptional signature of stem cell-derived neurons was shifted towards specific cerebral subregions at different developmental stages [11]. How changing culture conditions from 2D to 3D affects renal tubular cells, and reprogrammed iRECs in particular, has not been systematically analyzed.
A large variety of embedding techniques and biomaterials is available to generate suitable 3D microenvironments and improved culture conditions for different tissue types [12][13][14]. The kidney is considered a soft organ with a shear modulus of about 4.5 kPa, measured by magnetic resonance elastography [15], thus natural biomaterials are most suitable to support renal tubular cells [16]. Hydrogels represent modifiable networks of hydrophilic polymer chains in a water rich environment [17], and closely mimic the native extracellular matrix (ECM). The renal ECM is rich in collagen I and collagen IV [18]. Laminin and smaller molecules like entactin and other sulfated proteoglycans bridge the collagen fibrils and form a 3D network. Materials like Matrigel, mainly composed of laminin and collagen IV, are feasible candidates to reconstruct a kidney-like ECM [19]. Collagen I, fibrin and alginate are biomaterials widely employed to generate biomimetic 3D cell culture models [20,21]. For example, advanced micro-patterned microenvironments have been bioprinted using alginate [20]. Bioprinting can deliver highly reproducible results, but it requires the biomaterials to be finely adapted to each dispensing technique [22]. Despite recent advances, it remains a great challenge to fabricate functional renal tissues and tubular constructs [21,23]. We recently established a Drop-on-Demand (DoD) bioprinting process to generate defined patterns of iRECs in hydrogels and utilized their self-organizational properties to achieve tubular structures with a mean diameter of 105 ± 22 μm [24].
The physiological properties of iREC tubules have not been explored, nor has bioprinting been applied to model tubular kidney diseases.
Autosomal-dominant polycystic kidney disease (ADPKD) is the most prevalent monogenetic kidney disorder, affecting over 12 million patients worldwide [25] and leading to progressive kidney failure and end-stage renal disease [26]. The only approved treatment (Tolvaptan) delays disease progression by targeting cAMP mediated re-absorption in the collecting ducts, but it elicits thirst and increased diuresis, as well as liver toxicity, as prevalent side effects [27]. ADPKD is mainly caused by mutations of either PKD1 or PKD2. While inflammation and fibrosis contribute to subsequent renal failure, the primary insult that triggers cyst formation is a somatic "second hit" mutation of one of these genes in tubular epithelial cells [28,29]. Most mouse models of ADPKD only partially replicate the clinical phenotype of patients, as truly orthologous heterozygous knockout mice have a mild renal phenotype [30]. Cellular models for ADPKD exist, but are limited by diminished differentiation, or lack of isogenic controls [31]. Embedding cells in a 3D matrix of either collagen I [32] or Matrigel [33] elicits micro-cyst formation in spheroids. Interestingly, cAMP stimulation is required to trigger cyst formation in vitro in most cases. However, cysts also form in wild-type cells, though to a lesser degree than in PKD1/PKD2 deficient cells. These models are well suited but largely restricted to studying the role of cAMP dependent cyst progression driven by fluid secretion in the collecting duct, as no cysts are observed in the absence of cAMP agonists [32]. The positioning of seeded cells is hard to control and constructs generally do not assume a tubular architecture. iPSC-derived renal organoids have been generated that harbor mutations in PKD1 and PKD2 and form cystic expansion spontaneously [34,35]. However, due to the heterogeneous nature of organoids, effects of loss of PKD1 or PKD2 specifically on tubular cells are hard to distinguish from secondary factors that aggravate cyst progression and tubular injury, such as microenvironmental cues or fibrosis [34][35][36][37].
Therefore, iRECs represent an attractive cell culture and disease model because they maintain a state of differentiation when cultured and passaged extensively. They represent a tubular cell model that can be experimentally evaluated in isolation. Direct effects on the epithelium can be evaluated excluding the confounding effects of fibrosis and inflammation. Here, we investigated the impact of 3D microenvironments on iRECs by analyzing combinations of kidney ECM-like biomaterials and dispensing techniques. We assessed the physiological properties of cellular structures promoted by each microenvironment, including cell viability, morphology, and changes in gene expression. We detected gene signatures specific for each microenvironment and dispensing technique. Furthermore, using Cre recombinase mediated Pkd1 knockout, we generated an in vitro ADPKD model, which uncovered an overabundance of Aldh1a1 expression. Bioprinted tubule-like structures of Pkd1 − /− cells showed disrupted tubular integrity and outgrowth of radial side branches. Thus, microenvironmental conditions strongly influence the proliferative status, morphology, and gene expression profile of iRECs that may inform future experimental design strategies. In addition, Pkd1 − /− iRECs uncover a potential role for Aldh1a1 in ADPKD, and bioprinted tubular arrays are suitable for modeling cystic kidney disease in vitro.
All animals were raised in a SPF facility with a 12 h day/night cycle and free access to water and chow. Experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the German law for the welfare of animals (Regierungspräsidium Freiburg). Primers used for genotyping are listed in Supplementary Table 1.

Cloning and virus production
The coding sequences of mouse transcription factors Pax8, Emx2, Hnf1b and Hnf4a were cloned into the pWPXLd lentiviral vector (Addgene, #12258). Cre recombinase was cloned from cre-NLS-p3-26 (kind gift from the European Xenopus Resource Centre) and Flp recombinase (optimized Flippase) from pDIRE (Addgene, #26745). The puromycin resistance was cloned from pLIX_403 (Addgene, #41395). Cre and Flp were combined with puromycin resistance and cloned into the pWPXLd vector for lentiviral expression (Cre-Puro and Flp-Puro). The Pkd1-containing pLXSN retroviral transfer vector was a kind gift by Tilman Busch and Michael Koettgen, University Clinic Freiburg. Here, the coding sequence of Pkd1, flanked by n-terminal Flag and cterminal V5 tag, was cloned into the pLXSN retroviral transfer vector. For production of lentivirus, three plasmids (pWPXLd; psPax2, Addgene, #12260; pMD2.G, Addgene, #12259) were transfected into HEK293T cells (ATCC®, HEK 293 T/17, CRL-11268). Virus concentration and harvest were performed as previously described [40] using polyethylene 6000 precipitation and ultracentrifugation at 6800 rpm. Analogously, for γ-retrovirus production the transfer vector pLXSN, the packaging vector pMD-MLV Gag/Pol and the envelope-containing vector pMDG VSV-G were co-transfected into HEK293T cells, and the same protocol of virus concentration was applied.

Cell culture
Induced renal tubular epithelial cells (iRECs) were developed by Kaminski et al. by direct reprogramming of mouse embryonic fibroblasts (MEFs) by transduction of the four transcription factors Emx2, Hnf1b, Hnf4a, and Pax8 [5]. Cells were cultured in MEF medium (MEFM), containing standard Dulbecco's modified Eagle's medium (Gibco, DMEM, 41966029), supplemented with 10% v/v fetal bovine serum (Sigma-Aldrich, FBS, F9665), 2 U l − 1 penicillin/streptomycin (Sigma-Aldrich, P4333), and 2 mM L-glutamine (Gibco, 25030024). Where stated, iREC samples were cultured in Renal Epithelial Growth Medium (Lonza, REGM, CC-3190). MEFs were isolated from E13.5 embryos. Limbs and tail were cut, minced with a scalpel and trypsinized in 0.25% Trypsin-EDTA (Gibco) for 30 min. Cell pellets were resuspended in MEF medium, plated on 0.1% gelatin-coated cell culture dishes, and grown until confluency. Pkd1 fl/fl iRECs were reprogrammed analogously to iRECs. In short, Pkd1 fl/fl MEFs were transduced with four lentiviruses, each containing one of the four transcription factors Pax8, Hnf1b, Hnf4a and Emx2. The concentrated lentiviruses were diluted 1:100 to 1:200 in MEFM containing 10 μg ml − 1 polybrene (Santa Cruz Biotechnology, sc-134220), and were transduced for 12 h for six consecutive days. Three weeks after lentiviral transduction, cells were singled in a FACSAria™ Fusion cell sorter (Becton Dickinson) and grown clonally in 96 well plates. After morphological and qPCR-based characterization each selected iREC clone was transduced with either a Flippase-Puromycin (Flp)-containing lentivirus (control cells) or a Cre-Puromycin (Cre)containing lentivirus (Pkd1 knockout cells) for three consecutive days. Cells were cryoconserved in freezing medium, containing 50% MEFM, 40% FBS, and 10% DMSO, and stored in liquid nitrogen. During cultivation, all cells were incubated in a humid atmosphere at 37 • C containing 5% CO 2 . For RNA sequencing experiments, 3 Pkd1 clone pairs (Flp and Cre transduced conditions) were seeded at an initial density of 1⋅ 10 6 cells in 10 cm dishes in MEFM. After confluency, cells were harvested for RNA extraction. For measurement of proliferation rates, 5⋅ 10 4 cells of Pkd1 clones were seeded per 12-well hole on day 0. Starting at day 1, cell number was counted in triplicates for 7 consecutive days. . All samples were incubated at 37 • C for 20 min prior to addition of culture media (MEFM or REGM), with aprotinin for the fibrin condition. Alginate and RGD-alginate solutions were obtained dissolving 1 mg or 2 mg alginate powder in 100 ml dH 2 O, for 1% and 2% w/v solutions respectively. To form alginate hydrogels, the cross-linking agent calcium chloride was dissolved in dH 2 O and alginate solutions were cross-linked with 10% w/v calcium chloride for biomaterial characterizations, or 2%, 5% or 10% w/v calcium chloride for bioprinting experiments. For cell encapsulation, iREC pellets were resuspended in 2% w/v alginate or RGD-alginate solutions and utilized directly for bioprinting, separately loaded in the inkjet bioprinter from the calcium chloride solution.

Hydrogel preparation and embedding
For 3D cell culture of Pkd1 iRECs in Matrigel, cells were passed through a 50 μm cell strainer to avoid cell aggregates, counted to yield a final cell concentration of 100 cells μl − 1 , and resuspended in Matrigel at 4 • C (final concentration of 6 mg ml − 1 ). Cell suspensions in Matrigel were incubated at 37 • C in 8 well ibidi® dishes (ibidi®) for 20 min to allow thermally induced gelification. Afterwards, hydrogels were incubated in REGM for up to 10 days before staining. Flp-and Cretransduced Pkd1 clones were embedded in parallel, n = 6 clones. For hanging drop culture, Pkd1 iRECs were used at a concentration of 25 cells μl − 1 in REGM containing 2.4 mg ml − 1 methylcellulose. 64 drops (8 × 8) with a volume of 20 μl were pipetted onto the inner side of the lid of a 10 cm cell culture dish. Lids were carefully turned and hanging drops were cultured in a humidified atmosphere for 6 days.

Biomaterial characterization
All biomaterials were characterized by applying various tests on cross-linked samples (volume of 150 μl). Hydrogels were prepared (Matrigel®, fibrin, collagen I and alginate) and exposed to cell culture medium (DMEM). All conditions were weighed with a high precision scale, after removing the excess of solvent with adsorbent laboratory paper. The initial weight of each sample was registered as weight 0 . Then, samples were weighed at referred time points until complete degradation (weight t ). Degradation was quantified by weight loss over time, Δweight, using formula (1): Alginate hydrogels were further characterized with a dynamic mechanical analyzer (DMA Q800 T A. Instruments). Cylindrical samples (diameter = 5 mm, height = 2 mm, n = 3) were assessed using an unconfined compression test in physiological conditions (i.e., wet environment, 37 • C), applying a force ramp of 0.2 N min − 1 until rupture. From the obtained curves stress values at 50% deformation (σ 50% ), the maximum reached strain (ε max ), and the related maximum stress (σ max ) were compared for the two sample types.
In addition, viscosity measurements of 1% w/v alginate, 2% w/v alginate and 2% w/v RGD-alginate solutions were conducted prior to crosslinking, in absence of calcium chloride solution, with a rotational rheometer (MCR 52 Anton Paar GmbH, Austria). Measurements were carried out at room temperature (T room = 24 • C), with a cone plate geometry (gap = 96 μm, α cone− plate = 0.997 • , φ cone = 49.966 mm). The viscosity was assessed in relation to the shear rate (1-19000 s − 1 ). The working shear rate was estimated according to the printing parameters [41]. With a nozzle diameter of 500 μm and an outlet velocity of 2 m s − 1 , the estimated shear rate was 8000 s − 1 . This value was used to infer the viscosity of alginate solutions.

Immunofluorescence staining and viability assays
Cell viability was determined with a live/dead assay, using ReadyProbes™ Cell Viability Imaging Kit, Blue/Red (Thermo Fisher Scientific, REF R37610) to stain cell nuclei with DAPI and dead cell nuclei with propidium iodide, according to manufacturer instructions. In short, samples were washed in 1x PBS and exposed to the staining solution for 25 min (65 μl of staining dye per 1 ml of DMEM). To measure intensities for dead cells, positive control samples were either frozen at − 20 • C overnight in medium without DMSO (alginate and RGD-alginate) or exposed to 4% PFA before staining. The impact of the bioprinting process on cell viability was evaluated via a live/dead assay using inkjet bioprinted and manually dispensed iREC suspensions at a concentration of 1.68⋅ 10 5 cells ml − 1 , directly into 8-well ibidi dishes, pre-filled with 250 μl MEFM. Cell viability was measured at day 1, 3 and 7 after seeding.

Imaging
Imaging for live/dead assay and morphology was carried out with either an Apotome2 Zeiss, Colibri 2, Power Supply 232, Zeiss Axio Observer Z1, with a 20× air objective, or a confocal system NLO-I-LSM Zeiss LSM 880, with a 25× air objective.
Acquisitions of proliferation behaviors via pHH3 staining and ZO-1 localization were performed on the confocal SP8 Falcon 25× glycerol objective (10 Z-stacks per each triplicate and control), and three 80 μm Z-stacks per biological replicate, respectively. The validation of transcriptomic analyses was imaged with the Olympus Spinning disk with a 20× air or 30× oil objective, acquiring three 80 μm Z-stacks for each triplicate and control. Single frames of WT mouse kidney cryosections were imaged at a Leica DMI 6000 with a 40× oil objective. Images of Matrigel® embedded and DoD bioprinted Pkd1 iRECs were made with a confocal laser scanning microscope (Zeiss, LSM 510 DUO with inverted microscope Axiovert 200) or a Zeiss Axio Observer.Z1 with ApoTome.2 ZEN and ZEN 2010 software. Bright-field images of hanging drop culture were taken with a Zeiss Axio Imager.M2 upright microscope. For second harmonic generation imaging, cells were pipetted manually onto a first collagen I layer instead of bioprinting to obtain less localized cell patterns before a second covering layer of collagen I was applied, and imaged at the multiphoton Leica SP8 MP DIVE FALCON, as previously described [42].

Image analysis
A custom-made code was developed with Fiji software (version 1.52p) to analyze images of live/dead assay experiments. Each acquired set of images consisted of image stacks in z direction and 3 fluorescence channels. First, the original multichannel images were split into separate stacks per fluorescence channel. A duplicate of the blue fluorescence channel was processed to detect the ROIs (regions of interest): nuclei of both living and dead cells stained with DAPI. All ROI positions were collected and overlapped to the unprocessed DAPI images at the corresponding z level to verify the correct nuclei detection. As further quality control of the automated segmentation, ROIs were overlapped to corresponding GFP images to check if each region was actually surrounded by cell surfaces, which express green fluorescent protein intrinsically in iRECs. Last, the detected ROIs were overlapped to the corresponding red fluorescence images to measure the median red intensity of each ROI and to assess cell viability with respect to positive control images. If the median red intensity value was lower than the average median red intensity of positive control data, that single cell was counted as a living cell, and vice versa. The quantitative analysis of iREC morphology in 3D was performed manually with imageJ, reporting the number of structures per z-stack, the number of cells composing each structure, the area at the largest aggregate diameter, and the respective circularity (range 0-1, perfect circle = 1).
The analysis of proliferating iRECs embedded in Matrigel, fibrin and collagen I was performed by processing acquired z-stacks with Imaris software (Oxford Instruments Group) for nuclei segmentation (Hoechst channel). Each nucleus that showed a staining in the pHH3 channel was manually counted as a proliferating cell. Images of Pkd1 iRECs were analyzed with Fiji ImageJ and images of iREC spheroids were further processed with Imaris software (Oxford Instruments Group). The orientation of actin filaments was performed with the Fiji plugin Ori-entationJ Analysis, as previously described [43].

Bioprinting
The bioprinting was performed using a Drop-on-Demand (DoD) technology based on a commercial piezo-driven dispenser (Biofluidix GmbH, PipeJet®, Germany). The dispenser used disposable sterile capillaries, which could be directly connected to pipette-tips as bioink reservoirs. This allowed voxelated deposition of single bioink droplets with adjustable volumes between 1 and 70 nl, using a nozzle diameter of 200 or 500 μm. Detailed descriptions of the bioprinting process can be found below [24,44].

Micropatterned alginate single-and multilayer substrates
First, single-layer alginate substrates were optimized by printing calcium chloride droplets varying D x (distance between consecutive drops) along the same line and D y (distance between parallel lines) for parallel lines, and by changing droplet dimension depending on the chosen nozzle diameter Ø nozzle . Multilayer iRECs-laden alginate samples were bioprinted with a single 10 mm line geometry. Distance between consecutive droplets met 50% nozzle size, so that droplets overlapped about 50% (D x = Ø nozzle /2 = 250 μm). The final sample was obtained by overlapping 5 hydrogel layers, composed of iRECs-laden alginate bioprinted with a 0.5 mm nozzle diameter. On top of each layer a 5% w/v CaCl 2 solution was bioprinted with a 0.2 mm nozzle diameter. For improved stability in height and hydrogel crosslinking, a waiting time of 60 s elapsed between each consecutive layer.

Bioprinted tubule in collagen/fibrin sandwich
To mimic the complex structure of the kidney extracellular matrix, we produced bioprinted samples combining two biomaterials (collagen I and fibrin). Cell growth was supported by bioprinting a high density iRECs-laden fibrinogen bioink in a line pattern on top of a collagen I layer as substrate. To stabilize the construct, a second collagen I layer was pipetted on top.
In detail, collagen I (Corning, collagen I, rat tail, REF 354236) was pipetted on a glass coverslip to form the first hydrogel layer. Collagen I was used at a final concentration of 3 mg ml − 1 and was mixed with thrombin (Calbiochem, thrombin, human plasma, final concentration 10 U ml − 1 , #605190), required for the crosslinking of the upcoming fibrinogen layer. The pH of the solution was brought to alkalinity by adding sodium hydroxide (NaOH, 1 M), double distilled water, and PBS (Sigma-Aldrich, PBS, 10X) to allow gelification of collagen I. Crosslinking was thermally induced by incubation at 37 • C for 30 min. Cells were suspended in fibrinogen (Calbiochem, fibrinogen, human plasma, final concentration 5 mg ml − 1 , #341576) as a carrier at a concentration of 1.5⋅ 10 7 cells ml − 1 (adjustable concentration range of 1⋅ 10 6 to 2.5⋅ 10 7 cells ml − 1 , leading to droplets containing 10 to 250 cells, respectively). The cell suspension was deposited as droplets (volume of 10 nl) by the piezo-driven DoD dispenser (nozzle diameter 200 μm). Fibrinogen was cross-linked by thrombin, contained in the first hydrogel layer, to form on-site cell patterns and a fixed fibrin scaffold. Single droplets were combined by overlap printing to form a continuous line with a droplet-to-droplet distance of 200 μm. The freshly bioprinted cell pattern was then covered with a second collagen I layer (3 mg ml − 1 ) for encapsulation. Bioprinted cells were incubated in REGM for up to 10 days. To avoid enzymatic digestion of fibrin by cell secreted proteases, aprotinin was added to cell culture medium (Sigma, aprotinin, bovine lung, 20 μg ml − 1 , A1153). During incubation in REGM cell aggregates self-assembled and developed into a single tubular structure. Cell compatibility with the printing process and iREC viability were assessed by applying ReadyProbes™ Cell Viability Imaging Kit, Blue/Red (Thermo Fisher Scientific) to stain cell nuclei with DAPI and dead cell nuclei with propidium iodide.

mRNA extraction and quantitative real-time PCR
RNA was extracted in triplicates from iRECs embedded into four different microenvironments (Matrigel®, fibrin, collagen I and bioprinted) after 7 days of incubation, having 2D seeded iRECs as a control. In particular, each replicate corresponded to passages p35, p48 and p53, covering a time of six weeks in culture among biological replicates. For each condition, including the 2D control, 8 identical samples, produced at day 0 with 6000 cells each, were pooled for RNA extraction. For ADPKD experiments, Pkd1 iREC lysates and whole kidney lysates of iPkd1 and control mice were used. Total RNA was extracted using the QIAzol Lysis Reagent (Qiagen, no. 79306), and isolated with the RNeasy Plus Universal Mini Kit (Qiagen, no. 73404), according to the manufacturer's instructions. For 3D hydrogels, initial steps suggested by Khetan and Burdick were applied [45].

Transcriptomic analysis
For RNA sequencing, total RNA was extracted as described above. Library preparation and RNA sequencing were performed by Eurofinsgenomics (GATC Biotech AG, Konstanz) for Pkd1 iREC samples and by Novogene (HK) for iREC samples. Respectively, single-end 50 bp mRNA reads and 150 bp paired-end reads were generated and sequenced on an Illumina HiSeq platform. Fastq files were uploaded to the Galaxy platform and further processed with the usegalaxy.eu server [47]. After quality check using FastQC (version 0.67 and 0.72) and read trimming with TrimGalore! (Version 0.4.3.0), reads were aligned to the genome build GRCm38.89 (mm10) with RNA STAR (version 2.5.2b-2 and 2.6.0b-1) [48]. For differential gene expression analysis, gene counts were computed with featureCounts (version 1.6.0.3 and 1.6.3) [49] and statistically evaluated with DESeq2 (version 1.22.1 and 2.11.40.1) [50]. The DESeq2 analysis for the 3D cultured iRECs was run considering each microenvironment and individual replicate as a factor. Further bioinformatic analysis and visualization were conducted with Rstudio (version March 1, 1093, R version 4.0.2). Kidney enrichment was determined for genes with an expression in the mouse kidney larger than the median of all other mouse tissues profiled in Shen Y. et al. [51]. The gene ontology (GO) enrichment analysis of Pkd1 data was run with clusterProfiler (version 3.10.1) with a p-value cutoff of 0.01, corrected with the Benjamini-Hochberg method, and a q-value cutoff of 0.05 [52]. Further analyses were performed with the g:Profiler web version [53] (version: e101_eg48_p14_baf17f0, Benjamini-Hochberg FDR = 0.05) and the REVIGO platform [54]. Volcano plots and heatmaps were made with Bioconductor packages [55] and R packages ggplot2 (version 3.2.1), heatmap.plus (version 1.3), EnhancedVolcano (version 1.10.0), pheatmap (version 1.0.12) and scatterplot3d (version 0. . Previously published RNA sequencing data were reanalyzed using the same processing algorithm on the usegalaxy.eu server as described above.

Statistical analysis
Data were statistically analyzed and visualized using GraphPad Prism8 (GraphPad, version 8.4.2) software. To test for normal distribution, Shapiro-Wilk and Kolmogorov-Smirnov tests were used. Biological and technical replicates were reported along with each experiment (N), together with the statistical test. When one-way or twoway Anova proved significant differences, relevant individual comparisons were tested (Sigma, GraphPad). Significance levels of statistical tests as indicated in the figures: n.s. -non-significant, *p < 0.05, ** <0.01, *** <0.001, **** <0.0001.

Selection of hydrogels and assessment of biocompatibility with iRECs
First, we assessed structural features of three biomaterials widely used for 3D cell culture and bioprinting, namely Matrigel, fibrin and collagen I [14,56]. Our selection was based on the approach to use tissue-like biopolymers, reflecting the physicochemical microenvironment in natural tissues [56]. Matrigel and fibrin matrices have excellent biocompatibility, and can be enzymatically degraded and modified by cells, providing a high degree of freedom for cellular self-assembly [57]. This also accounts for collagen I, which is the most abundant interstitial protein, with well-defined compositions and tunable properties by setting the protein concentration before crosslinking [58]. The stability of these hydrogels was assessed in the absence of cells over 10 days (Supplementary Figs. 1a, b, c). Matrigel showed initial swelling between day 1-4, while no weight loss was detected in comparison to its initial weight ( Supplementary Fig. 1a). Weight loss and a subsequent degradation were more pronounced in fibrin, followed by collagen I, and ranged from 22.4 to 35.2% of initial weight at day 10. When hydrogels were incubated with iRECs ( Fig. 1a and b), initial swelling was only observed in Matrigel, and weight loss of all hydrogels was accelerated ( Supplementary Figs. 1a, b, c). The addition of cells led to the complete degradation of fibrin after 24-48 h. As previously shown [59], the serine protease inhibitor aprotinin was necessary to prevent premature hydrogel dissolution (Supplementary Fig. 1b).
We next evaluated the biocompatibility of Matrigel, fibrin and collagen I by measuring iREC viability with a live/dead assay (Supplementary Figs. 1d, e, f). In Matrigel and fibrin, a high iREC viability was maintained over 11 days, while it decreased significantly in collagen I ( Supplementary Fig. 1f). Interestingly, cell survival also depended on the culture medium. On day 6, renal epithelial growth medium (REGM) showed a beneficial impact on cell survival compared to standard culture medium (MEFM) in Matrigel and collagen I. In fibrin, however, viability was significantly reduced when iRECs were incubated in REGM ( Supplementary Fig. 1f).

iRECs self-assemble into complex structures depending on the embedding conditions
All hydrogel microenvironments promoted self-assembly of embedded cells into multicellular structures (Fig. 1c). However, each hydrogel had a different impact on the number, size and morphology of iREC structures. Over 11 days, the number of self-assembled structures was highest in fibrin. In Matrigel and collagen I, the incubation of cells in REGM led to higher numbers of cell structures compared to MEFM (Supplementary Fig. 1g). In Matrigel, self-assembly of iRECs resulted in small aggregates with high sphericity. In fibrin, mostly spheroids and sporadic tubule-like aggregates were detected. Matrigel spheroids grew bigger in size in REGM (Fig. 1d, Supplementary Fig. 1h). In fibrin, the size of spheroids was only slightly decreased by REGM, and it led to a significant increase in circularity (Fig. 1e). Cell structures in collagen I were mainly characterized by elongated and branching tubular shapes when cultured in REGM for 7 days (Fig. 1e and f).
To further characterize the differential impact of hydrogels on embedded cells, we investigated the apico-basal polarization of iREC structures by ZO-1 staining. In Matrigel, it could be detected towards the apical side of spheroids that contained a lumen. In collagen I, the apicobasal polarization was only partially conserved, as ZO-1 accumulated in the core of tubular aggregates, which in most cases did not form lumen. In fibrin, however, ZO-1 staining was seen at the inner and outer boundaries of spheroids, hinting towards a disturbed apico-basal polarization (Fig. 2a).

Transcriptional changes specific to matrigel, fibrin and collagen I
As iRECs are a directly reprogrammed kidney cell line originating from primary MEFs, we verified that cells maintained their differentiation status when incubated in 3D hydrogels. The expression levels of epithelial and kidney specific genes like Cdh1 (adherens junctions), Cdh16 (kidney specific protein), Lrp2 (megalin, absorptive epithelial tissues) and Slc17a1 (renal specific sodium-phosphate cotransporter) were measured in iRECs (Fig. 2b) at early (p24/p36) and late passages (p44/p53) separated by a culture period of 3 months. After 7 days in 2D or 3D culture, the gene expression of differentiation markers was maintained in both early and late passaged iRECs compared to primary MEFs. Thus, iRECs did not de-differentiate substantially over time or when cultured in 3D microenvironments. Comparing iRECs embedded in 3D microenvironments with 2D plated cells, we detected a significant upregulation of Cdh1 in Matrigel and a downregulation of Slc17a1 in fibrin (Fig. 2b). Therefore, the respective microenvironment had a small  but detectable influence on the expression of differentiation markers.
Next, we analyzed how the 3D microenvironment affected gene regulation and expression in iRECs across the transcriptome. We performed RNA sequencing of iRECs cultured in Matrigel, fibrin and collagen I for 7 days, and compared gene expression levels of 3D embedded cells to conventionally 2D plated iRECs (Fig. 2c). A total of 2356 genes were differentially regulated (DEGs) in all 3D versus 2D conditions (Matrigel/2D, fibrin/2D, collagen/2D). We identified 322 DEGs, which were concordantly up-or downregulated in all 3D microenvironments. Interestingly, 3D embedding mainly resulted in upregulation of genes, whereas the portion of downregulated genes was smaller in all three conditions (Fig. 2c, d, e).
A GO term analysis of the 200 commonly upregulated DEGs detected enrichment in terms like DNA replication, chromosome segregation, nuclear division and DNA repair, hinting towards enhanced proliferation rates of 3D embedded cells (Fig. 2f). Immunostaining against Phospho histone H3 Ser10 (pHH3) confirmed increased proliferation rates between 2D and 3D embedded cells at day 7 after hydrogel embedding in fibrin or collagen I ( Fig. 2g and h, Supplementary Figs. 2a  and b).

Bioprinting of alginate hydrogels
We evaluated how enhanced control of iREC dispensing in the bioprinting process could affect cell viability and characteristics. Because of its favorable biophysical characteristics, alginate was selected as a bioink to determine compatibility of iRECs to the printing process [60]. Alginate crosslinks in the presence of CaCl 2 ions and can be coupled to an RGD sequence, a tripeptide which enhances cell-matrix interaction and promotes cell proliferation and migration [61]. We took advantage of an established DoD bioprinting technique to generate microstructured alginate hydrogels [24].
First, we assessed the degradation rate of alginate hydrogels and their compatibility with the bioprinting process. Low alginate concentrations (1% w/v) maintained at least 50% of the initial hydrogel weight after 7 days in culture. As expected, higher concentrations (2% w/v) led to slower degradation (Δ weight = 50% at t = 7 days versus t = 14 days) ( Supplementary Fig. 3a). Additionally, higher concentrations showed improved mechanical stability, exhibiting a significantly higher compression stress value at maximum deformation compared to 1% w/v samples ( Supplementary Figs. 3b and c). As DoD bioprinting requires specific hydrogel viscosity (working range: 0.5-500 mPa*s), we measured the viscosity of 1% w/v alginate, 2% w/v alginate, and 2% w/ v RGD-modified alginate. All three solutions met the bioprinter requirements (Supplementary Fig. 3d). Exhibiting sustained mechanical stability, we used 2% w/v alginate and RGD-alginate solutions for further experiments.
The biocompatibility of the DoD bioprinting technology with iRECs was analyzed by applying a live/dead assay. Interestingly, bioprinted cells displayed an increased viability (95.3 ± 7.4%) compared to manually dispensed cells (87.2 ± 12.6%) one day after bioprinting. Yet, both methods resulted in similar iREC viability after 3 and 7 days, respectively (Fig. 3a). Cell proliferation was detected at each timepoint, with no significant differences between bioprinted and manually pipetted conditions (Fig. 3b).
We compared the effect of three calcium chloride (CaCl 2 ) concentrations (2%, 5%, 10% w/v) on the printing process. When CaCl 2 was bioprinted on top of the alginate layer, the formation of a repetitive micropattern was observed, which could be modulated by altering the drop overlap distance (Supplementary Fig. 3e) and the crosslinker concentration (Fig. 3c). Multi-layered iRECs-containing alginate and RGD-alginate samples were generated, which showed a homogeneous distribution of cells (Supplementary Fig. 3f). Importantly, cells started to self-assemble and form 3D structures at day 7 in RGD-alginate samples, whereas they remained mostly singled in pristine alginate (Supplementary Figs. 3g and h). RGD-alginate improved the viability of cells after longer periods of incubation (day 21) (Supplementary Fig. 3i).
In conclusion, the automatic DoD dispensing process had no detrimental effect on printed iRECs. However, since iRECs remained mostly as single cells up to day 7 in both alginate and RGD-alginate, this biomaterial had to be excluded for further comparative analysis.

DoD bioprinting promotes formation of tubule-like iREC structures
Next, iRECs were bioprinted in a fibrin and collagen I matrix, arranged as a cell-containing layer of fibrin embedded between two layers of collagen I [24]. This enabled cells to self-assemble and to form tubule-like structures. As schematically shown in Fig. 3d, an iREC-fibrinogen suspension was bioprinted on top of a collagen I layer. The cell-containing bioink (fibrinogen) was dispensed as overlapping droplets by DoD based bioprinting with a droplet-to-droplet distance of 200 μm. As fibrinogen was enzymatically crosslinked to fibrin by thrombin, a serine protease contained in the first layer of collagen I, cells were positioned in a linear arrangement. Afterwards, a second layer of collagen I was printed on cell-laden fibrin, and the hydrogels were incubated in REGM for up to 10 days [24].
In contrast to alginate embedded cells, the bioprinted constructs in fibrin/collagen I formed a compact cell layer. This process of guided cellular self-assembly was already observed one day after printing (Fig. 3e). Subsequently, linear structures shrank in diameter and formed a three-dimensional tubule with a diameter of about 100 μm (Fig. 3e and   f).

3D microenvironments enhance nephron segment specificity
The transcriptome of bioprinted cells was analyzed after selfassembly to tubule-like structures at day 7 and compared to the profiles obtained from iRECs manually embedded in Matrigel, fibrin, and collagen I. 2483 genes were significantly differentially regulated with respect to 2D cultured iRECs (Supplementary Table 3). In the principal component analysis (PCA), bioprinted cells clustered separately from 2D cells. More importantly, they could be clearly distinguished from the other 3D microenvironments (Supplementary Fig. 4a). 113 DEGs were detected which were concordantly upregulated in all 4 conditions of 3D culture in comparison to 2D culture (Fig. 3g). Interestingly, 31 genes were kidney enriched, including Tmem27 encoding collectrin, the amino acid transport regulator, or the sodium-independent organic anion transporter Slco4a1 (Fig. 3h). Thus, this analysis identified a core set of genes that were differentially regulated in 3D cultured iRECs, independent of the embedding substrate.
We also detected substrate specific transcriptional changes that only occurred specifically in one of the embedding or printing conditions (Fig. 3i, Supplementary Fig. 4b). 60.5% of DEGs uniquely upregulated in Matrigel were kidney enriched, 30.2% in fibrin, 43.7% in collagen, and 45.3% in the bioprinted samples. Interestingly, the gene signature in Matrigel showed the strongest proximal tubular specificity with enhanced expression of AldoB, Cubn, or Lrp2 and an accumulation of genes associated with the GO terms 'brush border' and 'apical membrane' (Supplementary Fig. 4c), also in comparison to the other microenvironments ( Supplementary Fig. 4d). The urate transporter Slc22a12 and the anion transporter Slc26a1, a nephrolithiasis risk locus, were significantly upregulated in Matrigel embedded samples.
Ace2 encoding for angiotensin converting enzyme 2, essential to the renal control of blood pressure and the functional receptor of SARS-CoV-2, was specifically upregulated in fibrin embedded iRECs. Among the transcripts with enhanced expression in collagen was Pcbd1, the Hnf1a co-factor, and the magnesium transporter Trpm6 was found uniquely upregulated in the bioprinted sample. In conclusion, the different embedding conditions elicited specific transcriptomic responses with a surprisingly high number of kidney enriched genes in each of the differentially upregulated gene sets. The uniquely enhanced expression within each condition may allow tailoring the 3D culture conditions to specific experimental requirements and thereby even allow controlling certain tissue-specific functionalities.

Validation of differentially expressed genes
To correlate differential expression of transcripts with protein presence, a number of proteins encoded by genes commonly upregulated in 3D vs. 2D culture conditions was validated by immunofluorescence. Confirming the transcriptomic analysis, Nqo1, Arg2, and Olfm1 showed high and distinct protein expression in iRECs (Fig. 4a). In parallel, the renal tubule specific expression of respective genes was confirmed in adult mouse kidneys (Fig. 4b). Upregulation of Slc25a48 and Gsta4 was confirmed by RT-qPCR (Fig. 4c).
Similarly, we analyzed proteins encoded by genes with enhanced expression in only one condition. AldoB was upregulated in Matrigel and the expression unique to this condition was confirmed by immunofluorescence (Fig. 4d). The endogenous localization was restricted to proximal tubules, in line with a shift of iRECs towards a proximal transcriptomic profile in Matrigel (Fig. 4e). Spp1 was detected in fibrin embedded iRECs and nuclei of intermediate tubule segments. Podxl, which was upregulated in collagen, showed detectable immunostaining also in fibrin embedded samples and 2D cultured iRECs, thus partially confirming the RNA-Seq results. Immunostaining also confirmed the protein expression of Nqo1, Arg2 and Olfm1 in bioprinted samples. Trpm6, specifically enriched in the printed condition, was detectable at cell membranes (Fig. 4f).

Disease related genes are differentially regulated according to microenvironment
Among DEGs upregulated in at least one condition, genes were detected which are associated with or causative for renal diseases. These included tubulopathy related genes such as Wnk4 (hyperkalemia) enhanced in Matrigel, Car2 (pyelonephritis susceptibility) detected in fibrin, Pcbd1 (hypomagnesemia and renal magnesium wasting) expressed in collagen or Fxyd2 (Renal hypomagnesemia-2) found in collagen and the bioprinted samples (Fig. 4g). We conclude that in vitro models of renal diseases may be enhanced by employing culture conditions that favor the expression of respective genes.

Transcriptome analysis identifies a dysregulated gene signature in Pkd1 knockout iRECs
ADPKD is a highly prevalent genetic disease with a primary (caption on next page) manifestation in renal tubular cells. We generated induced renal tubular epithelial cells with a loss-of-function allele for Pkd1 and analyzed the cellular transcriptome and morphology in 3D culture (Fig. 5a). MEFs from mice homozygous for floxed Pkd1 alleles were reprogrammed to Pkd1 fl/fl iRECs, clonally expanded, and their tubule-specific differentiation state was confirmed by RT-qPCR ( Supplementary Fig. 5a). As each clone showed an individual composition of marker gene expression, we selected clones that maintained expression of a broad range of tubule specific markers. In total, 6 Pkd1 fl/fl clones were chosen. Each of these clones was transduced with Flippase (Flp) as control, and, in parallel, with Cre recombinase (Cre) to generate a Pkd1 knockout, respectively. Hence, we could directly compare knockout cells to their corresponding wild-type clones. Cre recombination and knockout of Pkd1 were confirmed by genotyping. The reduction of the Pkd1 gene product PC1 was tested by Western blot analysis ( Supplementary Figs. 5b and c).
Transcriptional changes were analyzed by RNA sequencing in 3 pairs of Pkd1 clones. 303 DEGs were detected between control (Flp) and the Cre expressing cells. 121 genes were downregulated, and 182 genes were upregulated in Pkd1 knockout clones. Overall, 21 DEGs had a fold change of >2 (Fig. 5b and c, Supplementary Table 4). As expected in the Cre condition, the expression level of Pkd1 was reduced and no reads aligned to the genomic region of exons 2-4 of the Pkd1 gene, confirming successful Cre recombination (Supplementary Fig. 5d). The functional enrichment analysis showed an accumulation of GO terms associated with epithelial development, e.g., cell-cell adhesion, transmembrane and ion transport, tube development, apoptotic signaling pathway, actin cytoskeleton organization, and urogenital system development (Supplementary Fig. 6a).

Comparison to other transcriptomes identifies a PC1 dependent induction of Aldh1a1
To further evaluate the findings made in reprogrammed renal tubular cells, the datasets were compared to publicly available RNA sequencing studies of two Pkd1 knockout mouse models. We reanalyzed transcriptomic data of Pkd1 fl/fl ;HoxB7-cre mice [62] and a tamoxifen inducible kidney-specific Pkd1 deletion mouse model (iKsp-Pkd1 del ) [63] by applying the same processing algorithms as used for iREC RNA sequencing ( Fig. 5d and e, Supplementary Table 4). Interestingly, among 15 concordantly regulated genes found in all three datasets, the highest fold change could be found for Aldh1a1 (Pkd1 iRECs fold change (FC) = 2.07, Pkd1 fl/fl ;HoxB7 FC = 6.45, iKsp-Pkd1 del FC = 2.29, p adj < 0.001).
We selected a list of overlapping genes found in the Pkd1 knockout datasets and, for further validation, analyzed them by RT-qPCR. In parallel to Pkd1 iRECs, we made use of RNA from an inducible Pkd1 mouse model. Adult onset polycystic kidney disease was induced in Pkd1 fl/fl ;Pax8 rtTA ;TetO-cre mice (iPkd1 mice) by doxycycline treatment between P28 to P42 [39], and RNA was extracted at 12 weeks of age. The upregulation of Aldh1a1 expression after Pkd1 knockout could be confirmed in all iREC clones and the iPkd1 mouse model, where a significant upregulation of Aldh1a1 was detected in comparison to wild-type littermates. Moreover, Itga9, Slc16a10, Fbxo40, Heg1, and Pla1a were also concordantly up-or downregulated in the iPkd1 model (Fig. 5f). For other genes, the Pkd1 iREC RNA sequencing results could be confirmed by RT-qPCR, but results in iPkd1 mice were either non-significant or showed non-concordant regulation ( Supplementary   Fig. 7a).
Next, we analyzed if differential regulation of selected genes found in the RNA sequencing datasets was directly dependent on Pkd1 expression. Therefore, the Pkd1 coding sequence was transduced retrovirally into Pkd1 knockout iREC clones to rescue PC1 expression. Using an Anti-PC1 antibody, re-expression of PC1 could be confirmed in three Pkd1 knockout clones (Supplementary Fig. 7b, upper line). As the Pkd1 coding sequence in the retroviral vector was marked with an amino-terminal Flag tag and a carboxy-terminal V5 tag, immunoprecipitation was applied to check the expression of full-length PC1 protein (Supplementary Fig. 7b, lower line).
The expression level of several genes found in the Pkd1 knockout datasets was determined after Pkd1 rescue. Again, the higher mRNA level of Aldh1a1 could be confirmed in three Pkd1 knockout clones in relation to the corresponding wild-type clones, and Aldh1a1 expression returned to wild-type levels in the rescue condition in each clone pair ( Supplementary Fig. 7c). The reconstitution of the wild-type expression level was also observed for Slc25a37, Serpine2, Heg1, Exoc5, and Itga9. For other genes, the rescue yielded a non-significant result, or it showed an enhancement of deregulation (Fig. 5g, Supplementary Fig. 7d).

Bioprinting uncovers a tubular morphology defect not discernible in 2D culture
First, we checked the influence of Pkd1 knockout on cell proliferation in 2D cell culture. All Pkd1 iREC clones were grown in standard culture medium (DMEM) supplemented with 10% FBS. The knockout of Pkd1 did not alter cell proliferation rates compared to corresponding wildtype clones during a one-week course ( Supplementary Fig. 8a). Next, we focused on morphological features of Pkd1 iRECs in 3D cell culture. 6 Pkd1 iREC clone pairs were embedded in Matrigel and cultivated in Renal Epithelial Growth Medium (REGM) for up to 10 days. Cells were singled and seeded in low concentrations (100 cells μl − 1 ) to allow cell proliferation and formation of spheroids. Spheroid size and volume were determined, but no significant differences were detected between control (Flp) and Pkd1 knockout (Cre) conditions ( Fig. 6a and b and Supplementary Fig. 8b). Pkd1 iREC clones were also seeded in hanging drop culture. Cells aggregated and evolved into spheroidal structures. Again, no significant differences were detected concerning area, perimeter, and circularity of hanging drop spheroids (Fig. 6c, Supplementary Fig. 8c).
To further explore the behavior of Pkd1 knockout iRECs in a more tubule-like conformation, we employed the bioprinted tubule model depicted in Fig. 3d to obtain bioprinted samples of cell-containing fibrin matrix sandwiched between collagen I layers, analogously to the tubules analyzed earlier in this work [24]. Again, bioprinted cells started to form tubular aggregates after a few days in culture. In control (Flp) conditions, bioprinted cell aggregates developed into well-defined and sharply bounded tubule-like structures over time (Fig. 6d and e). Interestingly, two days after the printing process, Pkd1 knockout (Cre) tubule-like structures started to develop lateral protrusions, which diverted from the main axis, mainly perpendicularly, and grew in length. This phenomenon was significantly more prominent in the Cre conditions. We performed bioprinting for all 6 Pkd1 iREC clone pairs (Flp and Cre) and analyzed the number of side tubules per millimeter of main tubule, the cross-sectional dimension of the main tubule and the length of side tubules. In Cre conditions, the mean number of side    tubules per millimeter of main tubule exceeded the number of side tubules in Flp conditions significantly. No relevant differences could be detected concerning the mean diameter of the main tubule and the mean length of side tubules ( Fig. 6f and g). In addition, we used second harmonic generation imaging to investigate the orientation of collagen I fibers in the microenvironment of side tubules. Pkd1 wild-type clones showed an accumulation of collagen fibers in the vicinity of unbranched tubule boundaries, while in Pkd1 − /− samples, fewer collagen I fibers were detected, and some fibers were aligned along the direction of tubular branching (Supplementary Fig. 9a).
We also detected alignment of the actin cytoskeleton following the direction of side tubule branching, as shown in a vector analysis of phalloidin staining (Supplementary Fig. 9b). Moreover, pHH3 staining indicated enhanced proliferative cell activity at the basis of and within side tubules (Supplementary Fig. 9c).

Discussion
Here, we investigated the influence of specific 3D microenvironments on the viability, gene expression profiles and morphological behavior of renal tubule-like cells (iRECs) generated by direct reprogramming embedded in fibrin, collagen I, and Matrigel. Considering that iRECs showed differential morphological and transcriptomic responses in the diverse microenvironments, we established a bioprinted tubular disease model using Pkd1 knockout cells. This demonstrated a link to the retinoic acid synthetase Aldh1a1 and uncovered disrupted tubular selfassembly mechanisms at the ECM-interface of PC1 depleted iRECs.
2D cell culture models cannot represent the complex architecture of renal tissues and thus fail to mimic disease states in vitro. Novel 3D models aim to replicate cell-matrix interactions and allow tissues to assume conformations that more closely resemble their in vivo state. In contrast to models based on iPSC derived renal organoids [64] or immortalized cell lines [21], we used directly reprogrammed cells. These combine two unique features that benefit the establishment of an in vitro tubule model: iRECs are proliferative and remain stably differentiated over many months in culture (up to 3 months) [5], even allowing for clonal expansion. Proliferation rates can be either inhibited or accelerated in a cell line and hydrogel dependent manner [65]. Remarkably, placing iRECs into 3D microenvironments further stimulated their growth into complex multicellular aggregates, independent of the specific embedding substrate.
Systematically comparing commonly used biomaterials as embedding substrates allowed us to identify specific cell responses. Both cellular morphology and transcriptional responses varied substantially depending on the different hydrogels. This suggests that iRECs display a considerable degree of plasticity influenced by their 3D microenvironment. Genes encoding for tubular transporters, such as Slc22a12 (Oat4l), Slc23a1 (Vitamin C transporter Svct1), or Slc26a1 were strongly expressed in Matrigel embedded cells in comparison to 2D cultured iRECs, confirming high substrate specific responses. Matrigel enhanced a proximal tubular signature, and induced iREC aggregates to assume roundish spheroidal shapes, as did fibrin. In contrast, collagen led to a tubular morphology. Collagen cross-links into 10-200 nm thick fibrils [66], which might encourage iRECs to grow along the main direction of fibril organization and subsequently form an elongated shape. Among the uniquely differentially regulated genes in collagen embedded iRECs was Sgk2, a regulator of tubular transport activity [67,68]. We conclude that by choosing an appropriate microenvironment, cellular properties can be tuned to optimize experimental assays, such as enhanced expression of proteins of interest or regulatory components of cellular morphology. Our transcriptomic and morphological data across materials may provide a useful resource to identify and exploit such beneficial conditions.
Bioprinting approaches make it possible to position biomaterial ink or cell-laden bioinks and thus control the spatial organization of renal cells [21,64]. Here, we chose collagen as substrate because of its tubule enhancing properties, and bioprinted a line of iRECs diluted in fibrin between two collagen layers. This proven bioprinting technique utilizes self-assembly mechanisms of iRECs and resulted in a defined geometrical structure with a sharp cell-matrix interface, likely achieved by the guidance cues of collagen fibrils [24]. Collagen fibrils are known to regulate vascular network formation when oriented in 3D endothelial cultures [69], and human bone marrow-derived mesenchymal stromal cells displayed enhanced chondrogenic differentiation potential when embedded in aligned collagen-tyramine bioinks [70]. Our transcriptomic analysis identified gene signatures uniquely enhanced in printed samples, such as the magnesium channel Trpm6. Thus, bioprinting offers a customizable tool for generating a defined tubular geometry. In contrast to casting a hollow tubule or generating an artificial lumen using sacrificial ink that is later seeded with cells [21], harnessing the self-organizational capacities of iRECs resulted in tubular formations with dimensions much closer to the physiological values of renal tubules [24].
Here, we utilized this bioprinting approach as an appealing platform to phenotypically characterize Pkd1 − /− iRECs. The tubular shape of bioprinted constructs allowed us to examine Pkd1 knockout-induced deviations from this geometry, of which the formation of side branches was the most striking manifestation. The pathogenetic chain of events after second hit-induced PC1 loss can conceptually be broken down into loss of polarity, disruption of cell-matrix interactions and fluid accumulation [71,72]. The observed alterations in tubule geometry may thus reflect earliest disease manifestations and could be caused by disturbed cell-cell or cell-matrix interactions and a partially disorientated growth behavior. This hypothesis is supported by our findings of altered collagen I fiber reorientation in Pkd1 knockout samples and thus represents an interesting entry point for further investigations. The microenvironment plays a crucial role at early stages of ADPKD cystogenesis, and kidney epithelia can remodel the microenvironment through migratory forces [35]. PC1 is proposed to act as a mechanosensor and at cell-cell and cell-matrix contacts, where it modulates multiple signaling pathways, which may be cilia independent [73]. PC1 influences the interaction between the extracellular matrix and integrins [74] and modulates cell-cell adhesions [75]. Interestingly, PKD1 − /− kidney organoids showed impaired contraction when embedded in collagen I droplets, but cystogenesis mainly occurs under conditions of low adhesion [35]. Partial orientation of Pkd1 − /− cells along the axis of side tubules and enhanced proliferative activity could hint towards a microenvironment-driven adjustment of signaling pathways involved in processes like cell cycle or cell migration of PC1 deficient cells. Thus, our bioprinted tubular model may allow us to address if the underlying molecular alterations are due to disruption in signaling pathways such as planar cell polarity (PCP) and oriented cell division (OCD) [76,77], or Hippo signaling [78], which have been implicated in cyst formation. The lateral protrusions in our disease model are a consistent and striking phenotype, which can be used as an experimental read-out for other ADPKD specific questions. Due to the scalability of bioprinting, pharmacological screening could be envisioned.
An additional finding was the upregulation of the retinoic acid synthesizing enzyme Aldh1a1 in Pkd1 − /− iRECs, consistent with published transcriptomes of multiple in vivo samples [62,63,79], which we independently confirmed. Aldh1a1 and RA signaling seem to enhance repair mechanisms after acute kidney injury [80]. The expression of Aldh1a isoenzymes switches from Aldh1a2, which is dominant in the fetal period, to Aldh1a1 postnatally [81]. Aldh1a1 − /− mutant mice do not display defects in nephrogenesis [82], but are more susceptible to oxidative damage [83] and show metabolic alterations of adipogenesis, lipid metabolism and gluconeogenesis [84,85]. The PKD1 promoter, located 200 bp upstream of the transcription start site, can be activated by retinoic acid via the Sp1 transcription factor [86], suggesting a potential regulatory control of RA. Representing a key enzyme in RA production, Aldh1a1 could be a potential disease modifier of ADPKD. Further studies are required to characterize the role of Aldh1a1 in healthy and ADPKD kidneys, and to determine if the knockout of Aldh1a1 mitigates the course of cystic kidney disease.
As animal models or renal organoids can make it difficult to differentiate confounding effects from inflammation, fibrosis or secondary tubular damage, our model allows a more focused analysis of tubule specific alterations. Our findings may thus be more relevant to the initial cystogenesis triggered by a loss of PC1 and warrant further investigation.