Stem/proliferative and differentiated cells within primary murine colonic epithelium display distinct intracellular free Ca²⁺ signal codes

Abstract

The second messenger, intracellular free calcium (Ca²⁺), acts to transduce mitogenic and differentiation signals incoming to the colonic epithelium. In this report, a self-renewing monolayer of primary murine colonic epithelial cells is formed over a soft, transparent hydrogel matrix for the scalable analysis of intracellular Ca²⁺ transients. Cultures that were enriched for stem/proliferative cells exhibited repetitive, high frequency (ca. 25 peaks h⁻¹) and short pulse width (ca. 25 s) Ca²⁺ transients. Upon cell differentiation the transient frequency declined by 50% and pulse width widened by 200%. Metabolites and growth factors that are known to modulate stem cell proliferation and differentiation through Wnt and Notch signaling pathways, including CHIR-99021, DAPT, bone morphogenetic proteins (BMPs), and butyrate, also modulated Ca²⁺ oscillation patterns in a consistent manner. Increasing the stiffness of the supportive matrix from 200 Pa to 3 GPa shifted Ca²⁺ transient patterns toward those resembling differentiated cells. The ability to monitor Ca²⁺ oscillations with the spatial and temporal resolution offered by this platform, combined with its amenability to high-content screens, provides a powerful tool for investigating real-time communication within a wide range of primary tissues in addition to the colonic epithelium.

Graphical Abstract

An automated imaging pipeline platform was built to assess the intracellular Ca²⁺ transient patterns of the self-renewing primary murine colon epithelial cells. Various biochemical and mechanical conditions were used to modulate proliferation and differentiation of the primary colon epithelial cells. The proliferative and differentiated cells exhibited distinct Ca²⁺ transient patterns with the proliferative cells exhibiting short-lived, high-frequency Ca²⁺ oscillation, while differentiated cells showed wider and low-frequency peaks.

1. Introduction

Ca²⁺ is one of several intracellular signaling molecules that integrate environmental stimuli into chemical instructions that regulate cellular functions.^[1] As a ubiquitous second messenger, Ca²⁺ is involved in several key cellular processes such as proliferation/differentiation balance,^[2] maturation/development,^[3] cell cycle progression,^{[4, 5]} and cell death,^{[6, 7]} with the magnitude and downstream effects of these Ca²⁺ signals being cell type-specific.^[3] On a tissue-level scale, the roles of Ca²⁺ in excitable cells are well-studied and include smooth and striated muscle contraction from muscle cells^{[8, 9]} or the regulated release of neurotransmitters from neurons.^[10] However, Ca²⁺signaling in non-excitable cells, such as those that reside in the intestinal epithelium, is much less appreciated. A recent study utilizing the Drosophila model demonstrated that high Ca²⁺ levels may integrate mitogenic signals and induce proliferation of intestinal epithelial cells, with distinct Ca²⁺ oscillation patterns observed in proliferative, resting, and Ca²⁺ signaling-deficient cell phenotypes.^[11] These findings suggest a parallel role for Ca²⁺ signaling in the mammalian gastrointestinal tract, which has remained largely unstudied due to the technological barriers present in existing platforms.

The analysis of living organisms or intact intestinal tissue offers the best representation of mammalian physiology, though is hindered by access to host material, strict requirements for maintaining tissue viability, and highly convoluted background signals that must be physically mitigated (e.g., through tissue clearing solutions) or accounted for by advanced computational techniques. Human cell lines derived from the colon adenocarcinoma and gene-engineered to express the extracellular Ca²⁺ sensing receptor are far more accessible, and have been utilized for the investigation of intracellular Ca²⁺ oscillations.^[12] In this study, extracellular Ca²⁺, L-phenylalanine, and L-tryptophan elicited distinct Ca²⁺ oscillation patterns through signaling pathways specific to the SW-480, HT-29, and NCM-460 tumor cell lines. The elevation of extracellular Ca²⁺ promoted sinusoidal Ca²⁺ oscillation in >80% of stimulated cells and inhibited proliferation, while the aromatic amino acids induced short lasting Ca²⁺ spikes. While these findings suggest unique correlations between Ca²⁺ signaling patterns and intestinal cell phenotypes, the aberrant and irregular mutations observed in tissue-cultured tumor cell lines combined with the artificial culture environment (e.g., polystyrene substrate) may not be representative of healthy in vivo tissue. Improving on the physiological relevance of such models, spherical organoids derived from primary murine intestinal tissue and cultured within Matrigel patties have been used to study Ca²⁺ influx.^[13] Organoids are a well-established, primary cell-derived model that recapitulates several important features of the in vivo intestinal epithelium, including the maintenance of both stem and differentiated cell lineages, and are more predictive of in vivo physiology than immortalized cell lines.^[14] However, the enclosed three-dimensional architecture significantly limits assay throughput and the ability to observe intracellular Ca²⁺ transients. Innovations in biomaterial design are necessary and critical for the investigation of colonic epithelial Ca²⁺ oscillation patterns, for which the ideal platform would combine the physiological relevance of primary organoid culture with the assay throughput present in tumor-cell culture techniques. Moreover, the ability to associate distinct Ca²⁺ oscillation patterns with specific cell phenotypes would be highly informative and has not been reported using any primary intestinal cell model.

Hydrogel matrices composed of Type 1 collagen possess the requisite functional groups for the attachment of primary mammalian intestinal epithelial cells and the establishment of focal adhesions, as well as an appropriate stiffness that facilitates their self-renewal and serial expansion.^[15] Moreover, these cells form superimposed monolayers over the clear hydrogel surface as opposed to embedded three-dimensional structures, thus streamlining image acquisition and enabling high-content imaging screens to be conducted. Here, we adapt this monolayer culture system to study Ca²⁺ transients in living intestinal epithelial cells derived from healthy mammalian donors, demonstrating the effects of selected growth factors, metabolites, and substrate mechanical properties on distinct Ca²⁺ signaling patterns. A robust image analysis pipeline based on previously reported automated image acquisition algorithms^{[16, 17]} and the open-source CaImAn library^[18] was developed to isolate Ca²⁺ oscillation patterns in the cultured monolayers with single-cell resolution. The various biochemical and mechanical conditions that were tested modulated proliferation and differentiation of the colonic epithelial cells, with these populations exhibiting strikingly different Ca²⁺ oscillation patterns in regard to frequency and pulse width. Finally, we extended the assay capabilities of this planarized format to a unique platform that enables spatial segregation of the stem/proliferative and differentiated cell zones,^[19] analogous to in vivo crypts, where tailored spatial patterning of Ca²⁺ transients was observed.

2. Results and Discussion

2.1. Identification of Ca²⁺ signaling in primary colon epithelium

Primary murine colonic epithelial stem cells were expanded from resected donor tissue by isolating the colonic crypts and culturing the cells as a monolayer within a growth factor-enriched medium^[20] over a soft collagen hydrogel scaffold.^{[15, 21]} The soft hydrogel provides a surface of optimal stiffness for growth of the stem/proliferative intestinal epithelial cells.^[15] Compared to the widely utilized organoid culture technique, the planarized nature of this culture format facilitates rapid image acquisition while preserving many properties of intestinal stem cells that are observed in vivo, including tissue self-renewal (stem cells) with normal cell doubling times, a normal genetic karyotype, and the ability to differentiate into all major absorptive and secretory cell lineages.^{[15, 22]} Ca²⁺ imaging was performed on these expanding cell colonies while immersed in an expansion medium (EM) enriched with key contributors to intestinal stem cell maintenance and proliferation, including Wnt-3a, R-spondin 3, noggin, and epidermal growth factor. As such, these cultures were expected to be highly proliferative immediately after culture initiation. The fluorescent Ca²⁺ indicator, Cal-520, was selected for labelling as its dissociation constant (K_d = 320 nM) is suitable for free, intracellular Ca²⁺ measurements and it displays minimal subcellular compartmentalization (Figure 1a, Figure S1a, Supporting Information). In order to leverage image resolution with acquisition throughput, an automated focus algorithm was adopted that enabled rapid image tiling and accurate cell quantitation along the curved meniscus of the luminal surface of the hydrogel.^[17] The above constituents were employed within a humidified and temperature-controlled incubation chamber, allowing the cells to be assayed at varying time points over several days of culture.

Figure 1.

Measurement of Ca²⁺ signaling in primary colonic epithelial cells. (a, b) Representative images (a) and sample signals (b) of the Cal-520 loaded cells with background fluorescence subtracted. A total of 360 images were captured at 2 s timepoints. (c) A simplified flow diagram of the image analysis pipeline utilizing the CaImAn and CNMF-E packages. The time lapse images of the cells loaded with the Ca²⁺ indicator Cal-520 were collected. The temporal image stack was spatially filtered across the entire stack using a Gaussian kernel to remove background noise. Then a local correlation and peak-to-noise ratio images were constructed from the temporal image stack using the CaImAn package. The areas suitable for signal extraction were identified (seed points) and used to initialize the matrix factorization. The matrix is then iteratively fit to the CNMF model until the components reach a stabilized residual sum of squares (RSS), normally 3 to 5 cycles. The final outputs are the isolated fluorescent profiles and the associated cell areas.

Time series image stacks of these cultures were initially subjected to manual analysis and spontaneous transients in free, intracellular Ca²⁺ were readily observed within the boundaries of individual cells (Fig. 1a, b). These occurred in the form of distinct Ca²⁺ spikes, in which the frequency (5–70 peaks h⁻¹) and pulse width (10–100 s) were highly variable. Interestingly, as the culture duration increased the percentage of cells displaying repetitive Ca²⁺ signals significantly decreased, with 12 ± 7% of cells displaying spontaneous spikes on culture day 1, and 2 ± 1% of cells displaying spontaneous spikes on culture day 5 (Fig. S3a, Supporting Information). Beyond the presence and absence of Ca²⁺ signals, Ca²⁺ spike patterns were readily assayed over time. By day 3 of culture, the mean peak width increased, and by day 4 of the culture, both the mean peak frequency and the mean peak width were significantly different from those at day 1 culture. (Fig. S3b,c, Supporting Information). Our previous results indicate that the number of differentiated cells relative to the number of stem/proliferative cells increase over time as the colonies grow to confluency and deplete growth factors from the medium.^[15] Taken together with this data, we posit that differentiated cells from the murine colon engage in distinct Ca²⁺ signaling patterns relative to those of stem/proliferative cells, analogous to the fly gut.^[23] However, conventional image analysis methods presented severe data bottlenecks, limiting assay scalability and hypothesis testing.

To improve the throughput of image processing, Ca²⁺ signal isolation, and data interpretation, an automated image analysis pipeline based on the open-source CaImAn/CNMF-E library was developed and implemented in MATLAB (Figure. 1c).^{[24, 25]} This software library was originally designed for the analysis of Ca²⁺ transients in neurons within the high noise background of the in vivo brain, which is similar in nature to the identification of Ca²⁺ spikes within our tightly packed intestinal epithelial colonies on a thick scaffold.^[24] In brief, a spatial filter is initially applied that enhances fluorescent signals within cell expected diameters and mitigates background. After training and optimization of loading values (e.g., minimum local correlation and peak-to-noise ratios) from manually annotated image datasets, the pipeline can assign seed pixels to automatically generate cellular regions of interest (ROIs) that are correlated to their temporal fluorescence. The ROIs, temporal fluorescent signals, and background components are then optimized through iterative fitting within a constrained non-negative matrix (typically 3–5 cycles). To assess and optimize the performance of the pipeline, the manually annotated and segmented cells were compared to the pipeline output by calculating an F-score. An F-score is a commonly used metric to assess the accuracy and precision of a binary classifier ranging from 0 to 1, with 1 representing perfect identification of target samples. (Supporting Information, Fig. S2).^[26] After training and optimization, the pipeline yielded a maximum F-score of 0.74 ± 0.14 (n = 7 datasets), similar to the pipeline’s original application on neurons with an F-score of 0.754 ± 0.003(Figure S2b).^[18] Altogether, these factors render the pipeline suitable for tracking the Ca²⁺ oscillation patterns present within individual cells of our colonic epithelial platform.

2.2. Impact of signaling pathways that determine cell fates on intracellular Ca²⁺ oscillation

The observed evolution of the Ca²⁺ oscillation pattern over time suggests that differentiated cells may exhibit a distinct Ca²⁺ spiking pattern from proliferative, undifferentiated cells. To verify this, we placed the murine colon epithelial cells for 48 h in either EM that contains the growth promoting factors (Wnt-3a, R-spondin-1, noggin) and thus maintains undifferentiated and highly proliferative states, or differentiation medium (DM) that lacks the growth promoting factors thus inducing differentiation. Secreted near the stem cell niche in vivo by the underlying stroma, Wnts and R-spondin activates Wnt signaling pathway while noggin inhibits BMP signaling, which are essential in maintaining stem cell population and proliferation.^[27–30] In the in vivo colon, proliferative cells migrate away from the stem cell niche located at the base of an invagination or a crypt, and differentiate into either absorptive (i.e., colonocytes) or secretory (e.g., goblet, enteroendocrine, and tuft) cell lineages as the concentrations of these growth factors diminish along the crypt axis. Unlike the small intestine, the large intestinal epithelium does not possess Paneth cells (growth-factor-secreting cells) so that in vitro the only source of growth factors is that exogenously added.^[27–30] Thus these biochemical cues in vivo were exploited to modulate the proliferation and differentiation in vitro. Proliferation and differentiation states of the cells were confirmed by examining EdU incorporation (that occurs in the S phase of proliferative cells) and performing immunofluorescence for a differentiated cell marker, Krt20, respectively. In the EM, a large fraction of the cells was EdU+/KRT20− indicating that the cells were largely in an undifferentiated proliferative state, while DM-exposed cells were predominantly EdU−/KRT20+ identifying these cells as nondividing, differentiated cells (Figure 2a,b, S4). Next, cells cultured in either EM or DM were loaded with Cal-520 and examined for spontaneous Ca²⁺ spikes. The murine colon cells in EM possessed a significantly higher frequency of Ca²⁺ spiking (mean frequency of 25 peaks h⁻¹) compared to that of cells in DM (14 peaks h⁻¹) (Figure 2c). In contrast, the mean pulse width of cells in EM was significantly narrower than that of cells in DM (24 s vs. 59 s respectively, Figure 2d). The Ca²⁺ spike pattern of the cells in EM, enriched with Wnt-3a, is consistent with the sharp Ca²⁺ peak in response to Wnt-3a in PC3 cells, a prostate cancer cell line.^[31]

Figure 2.

Ca²⁺ properties signaling in stem/proliferative and differentiated cells. (a, b) Quantification of EdU incorporation (a) and a differentiation marker Krt20 (b) for proliferative capacity and differentiation status, respectively. The EdU and Krt20 positive area above the fluorescence intensity threshold was normalized to the nuclear area labeled by Hoechst 33342. n = 3 separate culture wells for each condition and the bar and the error bars represent the mean and a single standard deviation. Unpaired, two-tailed t-tests were performed for comparison between EM and DM conditions. P = 0.0203 for (a), 0.0014 for (b). (c,d) Box-and-whisker plot showing the properties of the Ca²⁺ signals over 1 h (c) and mean pulse widths of Ca²⁺ signals (d) in EM or DM. Boxes are centered on the median value and extend to the 25th and 75th percentiles. Whiskers extend to the minimum and maximum values. n = 13 cells for EM and 20 for DM. Unpaired, two-tailed t-tests with Welch’s correction was used to compare the mean Ca²⁺ peak frequency and mean Ca²⁺ peak width data. P = 0.0405 for (c), 0.0014 for (d). (e,f) Measured Cal-520 fluorescence of cells cultured in (e) EM and (f) DM normalized to the maximum Cal-520 fluorescence during the measurement. For all statistical comparisons, p values are represented as follows: * for p ≤ 0.05, ** for p < 0.01, and *** for p < 0.001.

Next, we investigated how growth-factor signaling pathways and their agonists modulated Ca²⁺ signaling in the colonic epithelial cells. CHIR-99021 is a potent Wnt signaling activator that acts via inhibition of glycogen synthesis kinase 3β (GSK3β).^[32–34] Addition of CHIR-99021 to EM did not alter Ca²⁺ signaling spiking in the cells (Figure 3a,b), most likely due to the already high concentrations of a Wnt/β-catenin signaling ligand Wnt-3a in the EM. For this reason, the amount of Wnt-3a (as well as R-spondin and noggin) in the medium was titrated down ten-fold until EdU incorporation was reduced to 20% of cells, normalized to nuclear area (Figure 3c). Cells incubated with CHIR-99021 demonstrated a more than two-fold increased frequency in Ca²⁺ spikes and a 50% decreased pulse width compared to cells in the reduced-growth factor medium without CHIR-99021 (Figure 3a,b). This shift in Ca²⁺ oscillation aligns with prior results in EM and DM and suggests that alterations in canonical Wnt signaling are sufficient to modify Ca²⁺ signaling attributes.

Figure 3.

Impact of growth and differentiation signals on the Ca²⁺ signaling profile in intestinal epithelium (a,d) Box-and-whisker plot showing distributions of total Ca²⁺spikes over 1 h (a) and mean pulse widths of Ca²⁺ spikes (d) with 50% or 5% L-WRN conditioned medium and addition of CHIR-99021. Boxes are centered at the median and extend to the 25th and 75th percentiles and the whiskers extended to the minimum and maximum values. N = 12 cells for 5% L-WRN and 19 for 5% L-WRN with CHIR-99021. P values are as follows (a) 0.9479 for 50% L-WRN vs. 50% L-WRN + CHIR, 0.3514 for 50% L-WRN vs. 5% L-WRN, 0.0042 for 5% L-WRN vs. 5% L-WRN + CHIR, (d) 0.8396 for 50% L-WRN vs. 50% L-WRN + CHIR, 0.0108 for 50% L-WRN vs. 5% L-WRN, 0.0297 for 5% L-WRN vs. 5% L-WRN + CHIR. (b,e) Box-and-whisker plot showing distributions of total Ca²⁺ spikes over 1 h (b) and mean pulse widths (e) of Ca²⁺ signals in response to activation of differentiation signaling pathways. Boxes are centered at the median and extend to the 25th and 75th percentiles, while the whiskers extend to the minimum and maximum values. N = 29 cells for the EM control, 14 for butyrate, 20 for BMP-2, 12 for BMP-9, and 13 for DAPT. P values (b) 0.0003 for EM vs. Butyrate, < 0.0001 for EM vs. BMP-2, 0.0007 for EM vs. BMP-9, 0.9986 for EM vs. CHIR, 0.0261 for EM vs. DAPT, 0.6159 for EM vs. 5% L-WRN, 0.1029 for EM vs. 5% L-WRN +CHIR.,(e) 0.0596 for EM vs. Butyrate, 0.0049 for EM vs. BMP-2, 0.4444 for EM vs. BMP-9, 0.9814 for EM vs. CHIR, 0.0319 for EM vs. DAPT, 0.0237 for EM vs. 5% L-WRN, 0.0409 for EM vs. 5% L-WRN +CHIR. (c, f) Quantification of EdU incorporation (c) and a differentiation marker Krt-20 (f) for assessing proliferation and differentiation, respectively. The EdU and Krt20 area above an empirically determined threshold was normalized to the nuclear area (Hoechst-3342) and bars and error bars represent means and standard deviation (n = 3 separate culture wells for each condition). P values (c) <0.0001 for EM vs DM, 0.0001 for EM vs. Butyrate, 0.9927 for EM vs. CHIR, 0.0077 for EM vs. DAPT, <0.0001 for EM vs. BMP-9, <0.0001 for EM vs. BMP-2, <0.0001 for EM vs. 5% L-WRN. (f) 0.0003 for EM vs DM, 0.0002 for EM vs. Butyrate, 0.9998 for EM vs. CHIR, 0.0072 for EM vs. DAPT, 0.0036 for EM vs. BMP-9, 0.0396 for EM vs. BMP-2. One way ANOVA tests were used with Welch’s correction for (a,b,d,e) and without for (c,f). For statistical comparisons, p values are represented as follows: * for p ≤ 0.05, ** for p < 0.01, and *** for p < 0.001

Notch signaling works in unison with Wnt signaling to maintain proliferation of the colon epithelial cells. The progenitor cells favor secretory cell fates when Notch signaling is reduced or inhibited. To assess independently whether Notch signaling might impact Ca²⁺ signaling patterns, a Notch signaling inhibitor DAPT, which blocks γ-secretase, which is essential in Notch signaling transduction, was added to cells cultured in EM for 48 h. Notch inhibition decreased cell proliferation and promoted differentiation, which was demonstrated by a significant decrease in the EdU incorporation and increase in Krt20 expression (Figure 3c,f). Inhibition of Notch signaling alone induced a 50% decrease in Ca²⁺ peak frequency, a similar Ca²⁺ signaling pattern to that of Wnt/noggin/R-spondin withdrawal in DM with a significant decrease in spike number over time and wider Ca²⁺ spikes (Figure 3d,e). This is consistent with the decreased Ca²⁺ oscillation by knockdown of Notch in Drosophila model, despite the opposite effect of decreased Notch signaling on proliferation (increased proliferation by Notch knockdown in Drosophila).^[11]

Differentiation of the colon epithelial cells is facilitated by multiple signaling factors at the luminal end of the colon crypt. Among these signals are bone-morphogenetic proteins (BMPs) that act to oppose Wnt-mediated self-renewal and proliferation. Bmp-2 and Bmp-4 are most prevalent in the intestine, produced by mesenchymal cells, and through Smad-mediated transcription repression, suppress proliferation independent of Wnt signaling.^{[35, 36]} To investigate how Bmps might impact Ca²⁺ signals, intestinal epithelial cells were cultured on the collagen scaffold in EM supplemented with BMP-2, or BMP-9. Differentiation of the cells in these cultures was confirmed by the significantly decreased EdU incorporation and increased Krt20 expression in the presence of the two Bmps relative to that of EM cultures (Figure 3c,f). Compared to the EM condition, Bmp-2 significantly decreased the mean peak count by 75% and increased the mean peak width 1.9 times (Figure 3d,e). Bmp-9 also significantly decreased the mean peak frequency by 65% but the change in peak width was not statistically different from that of the EM control (Figure 3d,e). A decreased Ca²⁺ pulse frequency with no change of peak width suggests that the mean intracellular free Ca²⁺ concentration has decreased, which is consistent with results from the Drosophila model.^[11] Bmp-2 and Bmp-9 belong to different subfamilies of Bmps and have different binding affinities to Bmp receptors.^[37] In the human colon tissue, BMP-2 and its receptor bone morphogenic protein receptor I (BMPR-1) are abundantly expressed while BMP-9 and its receptor, Anaplastic lymphoma kinase (ALK), are much less abundant^{[38, 39]} suggesting that BMP-2 and BMP-9 may have different potencies in inducing signaling changes in the colon epithelial cells. Nevertheless, addition of Bmps to EM-cultured cells induced differentiation of the murine colon epithelial cells and adoption of the Ca²⁺ signaling pattern of differentiated intestinal epithelial cells.^{[35, 36]}

The intestine is also host to multiple microbiota that give rise to a vast array of metabolites, many of which regulate intestinal epithelial cell physiology.^[40] One of these metabolites, butyrate, has shown to be a potent inhibitor of intestinal progenitor-cell proliferation driven by the Fox3 transcription factor.^[41] The murine colon epithelial cells cultured with butyrate supplemented in the EM for 48 h demonstrated increased differentiation compared to cells cultured in EM without butyrate, which is indicated by decreased EdU incorporation and increased Krt20 expression (Figure 3c,f). This is consistent with previous findings that butyrate induces differentiation into absorptive colonocytes in vitro.^[42] As expected, butyrate significantly decreased the mean peak frequency of Ca²⁺ spikes by 67% (Figure 3d). The mean pulse width (Figure 3e) was not significantly different between the butyrate treated group and the control (p > 0.05).

In summary, we demonstrate that signaling factors promoting proliferation or differentiation alter Ca²⁺ spike signaling attributes in murine colon epithelial cells. Increased Ca²⁺ peak frequency was promoted by Wnt agonist, decreased by Notch inhibition, and decreased by BMP-pathway activation, consistent with the Ca²⁺ spike patterns of the cells in the EM and DM. Also, butyrate, a bacterial derived differentiation promoting metabolite, decreased Ca²⁺ peak frequency similar to the other differentiation promoting signaling factors. These results are consistent with the previous finding that inhibited proliferation decreased intracellular Ca²⁺ oscillations in Drosophila model.^[11]

2.4. Impact of extracellular matrix stiffness on Ca²⁺ signaling

Mechanical cues within the local environment of the colon crypt, such as matrix stiffness, guide intestinal stem cell fate determination through mechanosensitive, YAP/TAZ-mediated pathways.^[43–46] Previous results suggest that intestinal stem cells undergo optimal growth over substrates that possess a stiffness (elastic modulus, E) of 100 – 1000 Pa, with cell differentiation increasing on stiffer surfaces.^{[15, 47–49]} To investigate the impact of matrix stiffness on Ca²⁺ signaling, primary murine colonic epithelial cells were cultured over an unmodified collagen hydrogel (E ~ 200 Pa),^[50] a highly cross-linked collagen gel (E ~ 10 kPa),^[51] and polystyrene (E ~ 3 GPa).^[52] A thin, surface coating of collagen was applied to each of these surfaces to allow cell attachment,^[22] and all cells were cultured for 48 h under expansion medium (EM), providing sufficient growth factors for normal proliferation.^[15] Cells grown over soft matrices were predicted to display Ca²⁺ signaling behaviors that are representative of stem/proliferative cells, while the cells cultured over the stiffer matrices might shift their Ca²⁺ signaling patterns to reflect those of differentiated lineages.

The Ca²⁺ peak frequency for cells grown over the two stiffer matrices (i.e., cross-linked collagen and polystyrene) was significantly lower than those grown over unmodified soft collagen (Figure 4a), suggesting that stiffness of the matrix also regulates cell outcome potentially by alteration of Ca²⁺ signaling. Interestingly, no significant differences were observed between the Ca²⁺ pulse widths for any of these cultures. The cells were assessed for their differentiation status by assaying EdU uptake and the presence of Krt20, with the cells grown over unmodified collagen demonstrating significantly more EdU incorporation than those grown over the two stiffer surfaces (Figure 4c,d). Surprisingly, however, only the cells grown over polystyrene exhibited higher Krt20 expression, suggesting that the cells grown over cross-linked collagen might be in a state intermediate to that of the softest and stiffest surfaces. Taking these results together with those of the growth factor studies, the cells exhibited increased differentiated cellular characteristics as they were cultured over stiffer surfaces with respect to EdU/Krt20 presence and Ca²⁺ peak frequency. Given that the pulse widths did not significantly change over any surface, it is possible that full differentiation may still require the removal of stem cell supporting growth factors from the culture medium. These results align with previous studies of matrix stiffness on stem/proliferative vs. differentiated cell allocations,^{[48, 53]} though here, correlates these phenotypes with distinct calcium signaling patterns. However, we recognize that a multitude of signaling pathways are utilized to sense the compression of neighboring cells and the stiffness of the surrounding matrix, including Raw/MAPK, PI3K/Akt, RhoA/ROCK, Wnt/β-catenin and TGF-β pathways,^[43–46] which warrant further investigations for their individual contributions to intestinal Ca²⁺ signaling patterns.

Figure 4.

Impact of matrix stiffness on Ca ²⁺ signaling in colonic epithelial cells. (a,b) Box-and-whisker plot showing Ca²⁺ spike frequency (a) and mean pulse width (b)after culture for 48 h on the different matrices. The box is centered at the median value and extends to the 25th and 75th percentiles. Whiskers extend to the minimum and the maximum values. N = 13 cells for neutralized collagen, 31 for cross-linked collagen, and 19 for polystyrene. One way ANOVA with Welch’s correction was used for statistical analysis. P values (a) 0.003 for neutralized collagen vs. crosslinked collagen, 0.003 for neutralized collagen vs. polystyrene, >0.99 for crosslinked collagen vs. polystyrene, (b) 0.14 for neutralized collagen vs. crosslinked collagen, 0.20 for neutralized collagen vs. polystyrene, >0.99 for crosslinked collagen vs. polystyrene. (c,d) EdU+ area (c) and Krt20+ area (d) normalized to the nuclear area (Hoechst 33342). Bars and error bars represent standard deviation (n = 3 separate culture wells for each condition). One way ANOVA was performed for statistical analysis. P values (c) 0.002 for neutralized collagen vs. crosslinked collagen, 0.006 for neutralized collagen vs. polystyrene, 0.79 for crosslinked collagen vs. polystyrene, (d) 0.06 for neutralized collagen vs. crosslinked collagen, <0.001 for neutralized collagen vs. polystyrene, <0.001 for crosslinked collagen vs. polystyrene. For statistical comparisons, p values are represented as follows: * for p ≤ 0.05, ** for p < 0.01, and *** for p < 0.001

4.5. Ca²⁺ signaling in flat crypt model

Using the planar hydrogel model, it was demonstrated that culture environments supporting stem/proliferative cells are associated with higher Ca²⁺ spike frequencies and narrower pulse widths relative to those of the differentiated cells. However, all of these culture systems possess a mixture of cells i.e. proliferative plus differentiated cells and assignment of the Ca²⁺ signaling attributes to the differentiation state of the cell is at best indirect. For these reasons, we used a previously described “flat crypt” culture system^[54] to form a tighter link between the cell differentiation state and Ca²⁺ signaling properties (Figure 5, Figure S5). In the flat crypt system, through-holes (50 μm diameter) are present beneath a monolayer of primary colonic epithelial cells and are connected fluidically to a basal reservoir supplied with EM. The luminal reservoir possesses DM. Since the large intestinal epithelium does not possess growth-factor-secreting cells, the only source of growth factors is that supplied via the through-holes from the basal reservoir.^[28–30] Therefore only those cells growing directly above the patterned array of through-holes have access to stem-cell supporting growth factors. We observed significantly greater EdU incorporation and diminished Krt20 expression in the cells at the through-holes relative to the cells spaced between the through-holes, confirming the expected segregation of stem/proliferative cells from differentiated cells (Figure 5b,c). The cells were thereafter grouped based on their distance from the nearest through-hole center, either within a 50 μm radius from the center or between 50 and 100 μm from the center, and their calcium signals, EdU incorporation, and Krt20 presence were compared. As expected, the cells above the through-hole showed significantly greater Ca²⁺ peaks per time and narrower peak widths compared to those that were spaced further away (Figure 5d–g). These results confirm that Ca²⁺ signal patterns supporting stem/progenitor cells utilize a Ca²⁺ code of higher frequency yet narrower Ca²⁺ pulse width relative to that in differentiated cells.

Figure 5.

Ca²⁺ signaling pattern in the stem/proliferative and differentiated cell zones of flat crypts. (a) A schematic side-view diagram (lower right 2 panels) and microscopy images of flat crypts. Shown in the upper left panel is a brightfield image of the 4 crypt regions with the center of each flat crypt positioned at the through-hole. The upper right 3 panels show the same 4 flat crypts stained for Hoeschst 33342, EdU incorporation or Krt20. Scale bar = 100 μm. (b, c) Spatial distribution of cells that are EdU+ (b) and Krt20+ (c). Measurements were taken in concentric circles emanating from the center of a microhole at 5 μm intervals to175 μm, the midpoint to the nearest microhole. Values reflect the average, and the error bars the standard deviation. N = 12 microholes evenly distributed across 3 devices. (d,e) Box-and-whisker plot showing Ca²⁺ spike frequency (d) and mean pulse width (d) vs their distance from the nearest microhole center. Boxes are centered at median and extend to the 25th and 75th percentiles and whiskers extend to the minimum and maximum values. N = 10 cells for 0–50 μm from hole center and 16 cells for 50 – 100 μm. T-test with Welch’s correction was used for statistical comparisons with p values represented as follows: * for p ≤ 0.05, ** for p < 0.01, and *** for p < 0.001. Exact p values for (c) 0.0027 and (d) 0.0002. (f,g) Three example signals from each category normalized to their maximum raw fluorescence. The signals were collected at 10.6, 21.5, and 13.9 μm from microhole center (e) and at 57.9, 61.5, and 89.2 μm (f) from microhole center.

3. Conclusions

We have demonstrated a powerful and automated method for the real-time monitoring of Ca²⁺ signaling with single-cell resolution in densely packed monolayers of primary murine colonic epithelial cells formed over planar and transparent scaffolds. Advantages of this method are its applicability to any Ca²⁺ indicator i.e. genetically encoded or small molecule and single or multiple wavelength measurement format.^[55] The method should be fully compatible with organ-on-chip systems with complex backgrounds and multiple imaging interfaces i.e. scaffolding hydrogel supports, microdevice walls, and surrounding liquids.^{[56, 57]} The open-source machine learning library, CaImAn/CMNF-E, was readily trained to segment individual colonic epithelial cells, identifying distinct calcium transients that could be correlated with stem/proliferative vs. differentiation status. Cells in stem/proliferative states exhibited Ca²⁺ oscillations with narrow peaks and high frequencies, while differentiated cells displayed wider peaks at significantly lower frequencies. Mechanistically, these assignments were supported by the addition of targeted growth factors or inhibitors to the growth medium, including the Wnt/β-catenin signaling agonist, CHIR 99021, the Notch inhibitor, DAPT, and BMPs that forced differentiation. As the stiffness of the extracellular matrix is also known to influence intestinal signaling and differentiation, this property was modulated through covalent crosslinking of collagen scaffolds and utilization of ECM-coated polystyrene. Compared to neutralized collagen, each of these stiffer matrices decreased the Ca²⁺ peak frequency, aligning with an observed decrease in EdU incorporation and an increase in Krt20 expression. Finally, these patterns were spatially modulated using a microfabricated flat crypt model that patterns and segregates actively dividing cells from terminally differentiated cells. Beyond demonstrating a clear association between Ca²⁺ transient patterns and cell phenotype, this technology will find widespread applications given its amenability to high-throughput screens, as well as the ubiquity of Ca²⁺ signaling throughout fundamental biology, pharmaceutical development, and disease physiology.

4. Experimental Section

Materials and Reagents:

Cal-520 in acetoxymethyl (AM) ester form (Abcam, Cambridge, MA) was utilized as the primary Ca²⁺ indicator, which was reconstituted in dimethyl sulfoxide (DMSO) at a concentration of 1 mM and stored at −20 °C. DMSO, probenecid, and sodium butyrate were obtained from Millipore-Sigma (St. Louis, MO). Rat tail type I collagen, polystyrene tissue culture plates (6-well, 12-well, and 24-well), 12-well Transwell inserts, and HEPES were purchased from Corning (Corning, NY). DAPT was obtained from Xcess Biosciences (Chicago, IL). Recombinant forms of bone morphogenetic protein 2 (BMP-2) and bone morphogenetic protein 9 (BMP-9) were obtained from R&D Systems (Minneapolis, MN). CHIR-99021 was obtained from Tocris Bioscience (Minneapolis, MN). All other materials were purchased from ThermoFisher Scientific (Waltham, MA). Deionized (DI) water (≥17.8 MΩ·cm), purified through a Barnstead NANOpure Diamond (Thermo Scientific) filtration system, was used for all reagent preparations.

Primary Murine Colonic Epithelial Cell Culture:

Murine colonic tissue specimens were collected from female and male mice aged 6–10 weeks at the University of North Carolina (UNC) at Chapel Hill. All experiments were performed in compliance with international laws and institutional guidelines under UNC Institutional Animal Care and Use Committee (IACUC) protocol #13–200. The epithelial cells were expanded following a previously reported monolayer culture protocol (Supporting Information).^{[21, 58]} When seeding the cells for Ca²⁺ imaging, cells from one well of a 6-well maintenance/expansion plate (9.6 cm² well⁻¹, ≥80% confluency) were split to 14 wells of a 24-well plate (9.6 cm² well⁻¹) containing 500 μL of neutralized collagen per well. For investigations of matrix stiffness, neutralized collagen, crosslinked collagen, and collagen-coated polystyrene supports were prepared in 24-well plates (Supporting Information) and the cells were subjected to the same passaging protocol. All cells were thereafter expanded for 48 under an expansion medium (EM) containing Wnt-3a, R-spondin 3, and noggin (Table S1, Supporting Information) and transitioned to their respective experimental culture medium for an additional 48 h with Ca²⁺ imaging and any fixation/downstream assay conducted on culture day 4. Experimental culture media included: EM, EM + DAPT (10 μM), EM + butyrate (1 mM), EM + BMP-2 (10 ng mL⁻¹), EM + BMP-9 (10 ng mL⁻¹), stem medium (SM, 5% (v/v) L-WRN conditioned medium, Table S1, Supporting Information), SM + CHIR-99021 (2.5 μM), and differentiation medium (DM, 0% L-WRN conditioned medium, Table S1, Supporting Information). The reduced Wnt-3a concentration in SM enabled the testing of exogenous Wnt signaling agonists on cellular proliferation and Ca²⁺ signaling.

Ca²⁺ Imaging Microscopy Setup:

An inverted TE300 Nikon microscope (Nikon Instruments Inc., Melville, NY) equipped with a CoolSNAP Myo CCD camera (Photometrics, Tucson, AZ) was used for live-cell imaging of Ca²⁺ transients. The microscope was equipped with a PS3H122R motorized focus drive and H138A motorized XY translational stage (Prior Scientific Inc., Rockland, MA), which were controlled using a ProScan III Controller (Prior Scientific Inc., Rockland, MA). The two objectives utilized during experimentation were a Plan Fluor 4x (N.A. 0.60, Nikon) and an ELWD Plan Fluor 40x (N.A. 0.13, Nikon) objective. A filter set containing UV-2E/C and B-2E/C filter cubes (excitation/emission: 360 ± 20 nm/460 ± 25 nm and 480 ± 15 nm/535 ± 20 nm, respectively, Nikon Instruments Inc., Melville, NY) permitted imaging in the blue and green wavelengths. Excitation light was provided by a Lumen 200 arc lamp (Prior Scientific Inc., Rockland, MA). A SmartShutter (IQ12-SA) in combination with a Lamba-SC (Sutter Instrument, Novato, CA) was used to control exposure time. The focus drive, microscope stage, camera, and shutter were controlled using customized software written in MATLAB v2017b.

Fluorescent Imaging of Intracellular Ca²⁺:

The loading of primary murine colonic epithelial cells with fluorescent Ca²⁺ indicators was facilitated by addition of 1% Pluronic-F127 and 2% Probenecid to the cell culture medium. Cal-520 AM (5 μM) was added to this mixture over the luminal surface of the cells and incubated for 1 h at 37 °C and 5% CO₂. Hoechst 33342 (2 μg mL⁻¹) was added over the final 30 min of incubation. The cells were then rinsed three times with PBS and overlaid with fresh loading medium containing no Ca²⁺ indicator or Hoechst 33342. The plate was transferred to a microscope stage equipped with a customized incubation chamber at 37 °C in a water-saturated environment. A brightfield area scan was performed at 4× magnification with a 5% overlap between fields of view using a previously described automated focus algorithm.^[17] The images were stitched together and selected sites were imaged under brightfield and fluorescence modes at 40× magnification. Ca²⁺ fluorescent images were then collected for each site at 2 s intervals over 12 min to yield 360 sequential time-lapse images.

Raw image stacks were first subjected to manual inspection. If Ca²⁺ signals were present, a best fit region of interest (ROI) was drawn. The mean fluorescence for each ROI was measured over time and a signal profile was constructed. Using previous neuronal Ca²⁺ traces as a reference,^[24] the threshold for Ca²⁺ peak detection was set to three times the standard deviation of the background. A minimum spacing of 6 s (3 frames) was implemented to account for Nyquist sampling and prevent redundant peak identifications.^[59] A final percentage of cells that exhibited Ca²⁺ oscillations were transposed over the Hoechst 33342 image channel at the same 40× image site for manual correlation of localized Ca²⁺ activity with specific nuclei.

The open-source CaImAn library was adapted to isolate Ca²⁺ signals from the collected fluorescence data in an automated fashion.^[18] Each nucleus was outlined and the associated pixels were used as positive control pixels during optimization of the Ca²⁺ transient isolation pipeline (Figure S2b).^{[18, 24]} The kernel used in the initial Gaussian spatial filtering depends on the mean diameter of the cells in each image, which reduces background noise while enhancing any objects with a smaller diameter than the Gaussian kernel (i.e., intracellular Ca²⁺ oscillations). Seeding pixels are implemented to create regions of interest (ROIs) within each image stack for greedy initialization, which define the starting cellular spatial components in the base model for constrained non-negative matrix factorization (CNMF) according to:

$$Y=AC+B+E$$

where Y is the three-dimensional matrix of pixel values, C is the spatial footprint of target cells, A is the Ca²⁺ activity of target cells, B is the fluctuating background, and E is uncorrelated noise. The pipeline is allowed to iteratively fit the model until the final components fit the constrained non-negative matrix. For the purposes of optimization, no manual interventions were introduced. The initial value range (10–60 px or 2–9 μm) for the Gaussian kernel was based on measured diameters for murine colonic crypt cells,^[60] while the initial minimum local correlation (0.5–0.9) and PNR (3–15) ranges were based on the previous CaImAn library report.^[18] Optimal performance was observed at a Gaussian kernel of 13 px (~2 μm), minimum local correlation of 0.8, and PNR of 5, (Figure S2a) which provided a mean F-score of 0.74 ± 0.14 and a runtime of 159 ± 36 s.

Measurement of EdU Incorporation and KRT20 Expression:

Paraformaldehyde-fixed cells were assayed by fluorescence imaging of DNA (Hoechst 33342, B2261, Millipore-Sigma), 5-ethynyl-2’-deoxyuridine (EdU) (3 h exposure, A10044, ThermoFisher Scientific), and cytokeratin-20 (KRT20, 13063, Cell Signaling Technology) utilizing established labeling methods (Supporting Information).^{[58, 61]} Image acquisition was performed on an inverted Olympus Fluoview 3000 confocal fluorescence microscope (Waltham, WA) utilizing a 4× objective (0.16 NA, UPlanSApo4X) or 20× objective (0.45 NA, LUCPlanFLN20X) in conjunction with a galvanometer scanner. Fluorophore excitation was achieved using the 405 nm, 488 nm, and 640 nm laser diodes and emission was collected at 430–470 nm (Hoechst 33342), 505–545 nm (AlexaFluor 488), and 650–750 nm (Cy5) wavelengths, respectively. Z-series optical sections were collected at 4× magnification spanning the depth of the entire cell colony (z-step size: 25 μm), while manually defined regions of interests (ROIs) across a single z-plane were collected at 20× magnification. For comparison of biomarker area coverage, z-series optical stacks were flattened into two-dimensional representations through summed z-projections and analyzed by a custom MATLAB script. Each fluorescent image channel was converted from grayscale to binary using a global Triangle threshold selection algorithm.^[62] The average area occupied by the suprathreshold level of fluorescence was measured and exported to a spreadsheet. Extracted suprathreshold area values were normalized to the area that was positive for Hoechst 33342 fluorescence in each image.

Fabrication of Flat Crypt Microdevices:

Microhole arrays composed of 1002F were fabricated by photolithography and fixed to the base to cell culture chamber inserts.^{[54, 63]} The negative photoresist, 1002F formulation 10 (1002F-10), was spin coated (Model WS-650MZ-23NPP, Laurell Technologies) over a glass side in two steps, first at 550 rpm for 10 s, then at 1700 rpm for 30 s. The glass slides were then baked at 95 °C for 1 h. The photoresist films were exposed to ultraviolet light (400 mJ cm⁻²) through a patterned chrome photomask (Front Range Photomask, Lake Havasu City, AZ) with a 20×20 array of open circles that were 50 μm in diameter and 350 μm in center-to-center distance. The exposed photoresist was developed in propylene glycol methyl ether acetate (PGMEA) and baked at 95 °C for 12 h to produce the final 20×20 grid of microhole arrays (~ 5 μm). This photopatterned photoresist film on a glass slide was secured to the base of a 12-well Transwell cassettes from which the permeable membrane was removed using a biocompatible transfer adhesive (1504XL, 3M). The photoresist/slide/cassette assembly was incubated in DI water at 70 °C for 2 h to release the photoresist from the glass backing, and excess photoresist film was carefully excised from the cassette base. Collagen matrix deposition and murine cell culture were performed in the devices as described previously (Supporting Information),^[54] with Ca²⁺ imaging and any fixation/downstream assays conducted on culture day 4.

Statistical Analyses:

Statistical analyses and data plotting were performed in GraphPad Prism v9.1.1 (GraphPad Software, San Diego, CA) at a significance level (α) of 0.05. Group means were compared using unpaired, two-tailed t-tests (Figure 2a,b) or one-way ANOVA (Figure 3c,f, Figure 4c,d) followed by Tukey’s test for multiple comparisons. Welch’s correction was applied for the mean Ca²⁺ peak frequency and mean Ca²⁺ peak width data (Figure 2c,d, Figure 3a,b,d,e, Figure 4a,b, Figure 5d–e) to account for unequal variances between datasets (F-test, p < 0.05). Unless specified, data are presented as sample means with error bars depicting one standard deviation. The box plot crossbars represent sample medians, 25th percentiles, and 75th percentiles, with the whiskers depicting the minimum and maximum measurements from each dataset. For statistical comparisons, p values are represented as follows: * for p ≤ 0.05, ** for p < 0.01, and *** for p < 0.001.