2.1.

Mouse Model and Necropsy

The Zfp148-CreERT2: Loxp-STOP-Loxp tdTomato mouse line was previously reported.^17,28 In brief, a transgenic mouse line was generated using a plasmid containing a bacterial artificial chromosome expressing 250 kb of the Zfp148 promoter inserted upstream of the CreERT2 cassette. The Zfp148CreERT2 line was then bred to the Loxp-STOP-Loxp ROSA-tdTomato mouse line purchased from Jackson Labs to generate the Zfp148CreERT2; loxpSTOPloxp-tdTomato hybrid line. A “loxp-STOP-loxp” site was introduced downstream of the Zfp148 promoter sequence and upstream of the tdTomato promoter to drive the Cre-mediated expression of tdTomato protein in ZBP-89 expressing cells. The Cre recombinase was activated by administering 2 mg of tamoxifen (T5648, Millipore Sigma, Burlington, Massachusetts, United States) in 0.1 mL sterile corn oil (C8267, Millipore Sigma) by gastric gavage once. The age of each mouse at the time of tamoxifen injection was variable between roughly 3 and 7 months of age. However, the gut epithelium is expected to turn over approximately every 3 to 5 days,²⁹ providing rationale for basing analysis on time since induction rather than initial starting age of each mouse. Littermate controls and tdTomato-expressing Zfp148-CreERT2 mice were euthanized by CO2 inhalation, and organs were explanted, flushed, and the luminal organs were dissected and arranged such that the luminal side was facing the illumination source. All tissues were maintained on ice in phosphate buffered saline (PBS). Images were collected to measure the expression of tdTomato-expressing ZBP-89+ cells in the GI tract, stomach, and liver at the time points, as shown in Table 1. The animal study was reviewed and approved by IACUC Protocol 18–440 PHS Animal Welfare Assurance Number: D16-00159 (A-3248–01) the University of Arizona.

2.2.

Imaging Systems

Figure 1(a) shows a schematic diagram of the wide-field fluorescence imaging systems used in the reported studies. Both studies used a similar system architecture for detection and illumination. The peak excitation of the tdTomato fluorophore is 554 nm, and peak emission is 581 nm. In general, the excitation and emission spectra for tdTomato are broad, allowing for a degree of flexibility on both the illumination and detection end.³⁰ For both studies, illumination was provided by a 300W xenon arc lamp source (LS-OF30, Sutter instruments, Novato, California, United States). Illumination light was filtered with a $561\mathrm{nm}\pm 20\mathrm{nm}$ bandpass filter (12-152, Edmund Optics, Barrington, New Jersey, United States) and delivered using a fiber bundle (LLG, Sutter Instruments) with a collimation lens. For study 1, images were collected with a 35 mm fixed focal length lens (59-872, Edmund Optics) focusing onto a 0.5-inch CCD Monochrome Camera (EO-1312M, Edmund Optics). The resolution for this system was $\sim 20\text{microns}$, and the integration time was 60 ms. A 594 nm longpass filter (BLP01-594R-25, Semrock) was used to collect emission light. The imaging system used in study 2 was reported and characterized in detail previously.²⁵ The illumination system is identical to study 1, but in this case, the imaging camera is a high performance back-illumination CMOS sensor (QHY600M, QHY, Beijing, China), coupled with a photographic lens (NIKKOR 55 mm f/2.8, Nikon, Tokyo, Japan) to produce high performance imaging. The resolution for this system was $\sim 7\text{microns}$, and the integration time was 20 ms. The vignetted appearance of images from study 2 is due to the limited numerical aperture of the illumination light guide being unable to fill the entire field of view of the sensor. To ensure high quality data acquisition, we used a brightfield image as a reference during data acquisition to orient all tissues such that they were within the illuminated region before beginning fluorescence imaging acquisitions. In future iterations of this system, this issue could be corrected by integrating additional illumination optics to expand the numerical aperture.

Fig. 1

(a) System diagram of the wide-field fluorescence imaging systems used in the two studies and (b) diagram of the data analysis procedures.

In both cases, the detection system was mounted on a rigid arm above the sample plane. Images were saved in 16 bit .tif format. Prior to imaging, a white diffuse reflectance target (58-609, Edmund Optics) was imaged to correct for spatial non-uniformity using unfiltered illumination. Additionally, a power measurement was collected for the filtered illumination light at the sample plane using a power meter (S120C, Thorlabs, Newton, New Jersey, United States).

2.3.

Data Analysis

Data were analyzed identically for both studies, diagrammatically, as shown in Fig. 1(b). First, using reflectance images as reference, regions of interest were manually drawn for each organ using ImageJ to create a binary mask corresponding to the organ area of interest. These binary masks were saved as separate image files.³¹ Masks were created for stomach, liver, and the GI tract, including duodenum, jejunum, ileum, and colon.

All acquired images were dark subtracted, normalized by the flat field image to correct for illumination non-uniformity, and scaled by the light source power. The binary masks were then applied separately for each organ and the mean value of the region of interest was calculated to assess fluorescence intensity. Statistical comparisons were conducted between wild-type and transgenic mice for each time point using a two-tailed t-test.

To assess potentially overlapping and mixed effects of different biological variables,––namely the sex of the subject and whether the reporter was homozygously expressed,—we used LMEs modeling to evaluate which variables influence expression of the fluorescence reporter.^32,33 We first aimed to estimate the effect of time on the reporter expression to confirm the results found using the t-test. To do so, we can model the measured fluorescence intensity using standard LME notation in the following equation:

For the second study, we confirmed our results for organ-level expression using LME for the expression, as shown in Eq. (3). Note that the subject distribution did not contain sufficient numbers of homozygous and heterozygous subjects to include this effect. Here, all samples were grouped according to time point with a random intercept for each group

All LMEs models created and optimized using Python 3 (Python Software Foundation, Beaverton, Oregon, United States) with the statsmodel package. All models were optimized using restricted maximum likelihood and were found to converge. In building these models, wild-type mice and stomach tissues were removed from the analysis, as we are investigating the effects within the positive experimental group.

2.4.

Tissue Fixation and Fluorescence Microscopy

Immediately following wide-field fluorescence imaging using our systems, all tissues were fixed in 4% paraformaldehyde for 30 min prior to dehydrating overnight at 4° in a 30% sucrose-PBS solution. Tissues were embedded in optical cutting temperature (OCT) compound (Sakura Finetek USA, Torrance, California, United States) and frozen at $-80°\mathrm{C}$. OCT-embedded tissues were sectioned at $-20°\mathrm{C}$ to a 10-micron thickness using a TN50 semi-automatic cryostat (Tanner Scientific, Sarasota, Florida, United States). Frozen sections were adhered onto glass slides, then mounted with a #1.5 glass coverslip using Prolong Gold Anti-fade mounting medium containing 4’,6-diamidino-2-phenylindole (Life Technologies, Rockville, Maryland, United States). Slides were imaged using the same acquisition (exposure time) settings for the tdTomato channel at 200× magnification using an Olympus BX53F epifluorescence microscope (Center Valley, Pennsylvania, United States). This microscopic fluorescence imaging serves as a gold standard reference for the wide-field fluorescence imaging completed in studies 1 and 2.

3.1.

3 Month High Sample Size Study

Our results confirm that ZBP-89 is expressed ubiquitously in the GI tract in our transgenic line. Representative raw fluorescence images from 3-month mice in study 1 are shown in Fig. 2, which illustrate significant expression of the fluorescent reporter is visible using wide-field imaging. The signal appears to manifest as bright spots with a diffuse background. As expected, wild-type mice have almost no detectable signal.

Fig. 2

Example images of whole-organ fluorescence imaging taken during study 1 for various murine organs taken at 3 months for wild type [WT; (a)–(c)] and transgenic [Zfp148; (d)–(f)]. Qualitatively, there is a strong signal in transgenic mice compared to WT mice throughout the gastrointestinal tract. Autofluorescence is observed in the stomach for both transgenic and WT littermate control mice. Brightness of all images shown was digitally increased by 40% and contrast was increased by 20% for better visualization.

Quantification of the images, as shown in Fig. 3, illustrates a statistically significant increase of fluorescent signal across the GI tract between transgenic and wild-type mice. There were no significant differences observed in the stomach tissues, and differences in signal with other tissues become more significant at longer time points, which indicates the label becomes increasingly expressed as time goes on. These results also suggest a gradient of decreasing expression from the proximal to distal intestine, which suggests a higher abundance of ZBP-89+ stem cells and differentiated progeny cells in the proximal intestine. We also note there is no significant signal at the 24 h time point as anticipated as mRNA transcription, translation, and protein folding may take longer than 24 h and up to 72 h.³⁴ Thus, there is likely no fluorescent protein being produced by 24 h. In transgenic mice, Fig. 4(a) shows the collected signal relative to wild type as a function of time, and Fig. 4(b) shows the acquired fluorescent signal as a function of tissue in the GI tract, further supporting the observations of spatial and temporal gradients.

Fig. 3

Quantification of wide-field fluorescent images in study 1, showing an increase over time in fluorescent signal in transgenic mice compared to wild-type littermate controls for all GI tissues except stomach. Error bars denote standard deviation. Sample sizes are listed in Table 1. Significance is denoted as * ($p<0.05$) and ** ($p<0.01$) using a two-tailed t-test.

Fig. 4

(a) Fluorescent signal relative to wild type as a function of time shows consistent increase from 1 day to 12 weeks. (b) At all time points, a gradient of higher to lower tdTomato expression across the proximal to distal GI tract was observed. Sample sizes are listed in Table 1. Error bars denote standard deviation. Duo – duodenum and Jej – jejunum.

To quantitatively assess the potential dependence of expression on these parameters, LME models were used. Table 2 summarizes the results from the LME model including effects from time (in weeks), sex, and homozygosity of tdTomato sequences on both alleles. The results show a highly significant ($p<0.001$) dependence on time and tdTomato homozygosity. The positive coefficient for time post-induction indicates that the expression will increase with time, as was qualitatively observed in Fig. 4(a), and the positive coefficient for homozygosity indicates that homozygous subjects will have an increase in expression compared to heterozygous mice with a single tdTomato allele. There was no significant dependence on sex as a biological variable for ZBP-89-tdTomato expression. The intercept parameter would imply that there is a nonzero offset at the first time point, which would be influenced by tissue autofluorescence levels and any observed reporter expression.

Table 2

LMEs model results for model fitted to Eq. (1). This models the effect of time, sex, and homozygosity on the fluorescence reporter expression. Results are shown as LME coefficient, standard error, Z-score, and statistical significance denoted as *** for p<0.001. Sex coefficient refers to male relative to female. Homozygosity refers to mice with genotyping positive for floxed tdTomato sequences on both alleles relative to heterozygous mice in which one allele is the wild type for tdTomato expression (negative). Upper and lower confidence intervals (C.I.) are shown for 95% confidence.

Similarly, Table 3 summarized results for the LME model built to assess effects of organ location and cross-effects of sex and homozygosity. Compared to the first LME model, we observe a similar result for the overall intercept and the homozygosity parameter ($p<0.001$), which supports the previous finding. We also see that there is no significant difference due to sex for this model, again confirming the previous result. From this LME model, we see that we obtain highly significant effects due to organ location. Coefficients are shown relative to signal obtained in the colon, and we see that the duodenum has the highest increase, followed by jejunum and ileum, (all $p<0.001$) consistent with the qualitative observations made in Figs. 3 and 4(b). The coefficient for the liver is negative, which indicates a lower overall signal, and confirms the apparent relationship observed in Fig. 3.

Table 3

LMEs model results for model fitted to Eq. (2). This models the effect of organ, sex, and homozygosity on the fluorescence reporter expression. Results are shown as LME coefficient, standard error, Z-score, and statistical significance denoted as * for p<0.05 and *** for p<0.001. Sex coefficient refers to male relative to female. Homozygosity refers to homozygous relative to heterozygous. All organ coefficients are shown relative to colon. Upper and lower confidence intervals (C.I.) are shown for 95% confidence.

3.2.

6 Month High-Resolution Study

Figure 5(a) shows quantitative results from the high-resolution imaging study for fluorescent signal collected from transgenic and wild-type mice at 6 months post-induction. Note that as the earlier time points for study 2 were limited to two transgenic samples each and no wild-type control, these time points were not powered for statistical significance and instead were primarily used for qualitative analysis. The overall relationship of expression is consistent with the first study, strongly supporting the reproducibility of the results. As before, we see significant increases in tdTomato fluorescence for transgenic mice in every organ except the stomach. Figure 5(b) shows the signal as a function of organ location at 6 months, once again following the decreasing expression gradient from proximal to distal intestine that was originally observed.

Fig. 5

(a) Fluorescent signal relative to wild type at the 6 month time point, showing a similar trend of expression across organ locations as in study 1. $n=5$ for each Zfp and WT. (b) Fluorescence intensity in transgenic samples as a function of organ location, showing decrease from the proximal to distal GI tract. Error bars denote standard deviation. Duo – duodenum and Jej – jejunum.

Table 4 summarizes results obtained from the LME model used to quantitatively evaluate signal as a function of organ location, as well as assess cross-effects of subject sex. Note that compared to study 1 (Tables 2 and 3), the coefficient scale will vary due to differences in detection system, such as different dynamic ranges and sensitivity. As with study 1 (shown in Table 4), we see no significant difference due to sex, and the results are also consistent in that the proximal intestine (e.g., duodenum) illustrates the highest signal increase compared to the distal intestine (e.g., colon), with a gradient in between. Note that compared to study 1, the coefficient scale will vary due to differences in detection system due to different dynamic ranges and sensitivity.

Table 4

LMEs model results for model fitted to Eq. (3). This models the effect of organ and sex on the fluorescence reporter expression for mice at 6 months post-induction in study 2. Results are shown as LME coefficient, standard error, Z-score, and statistical significance denoted as * for p<0.05 and *** for p<0.001. Sex coefficient refers to male relative to female. All organ coefficients are shown relative to colon. Upper and lower confidence intervals (C.I.) are shown for 95% confidence.

Taking both studies together, the use of LME models proves to be a powerful tool for the analysis and interpretation of this data. To make impactful conclusions from complex biological models, it is important to assess whether, and how, different variables influence the expression of markers. This approach could be applied generally for other models and labeled markers to evaluate the combined effects of multiple variables.

Figure 6 shows wide-field images collected for the duodenum and jejunum in the high-resolution imaging study to illustrate the sensitivity of tdTomato detection. Figure 6(a) shows that wild-type animals have negligible signal and that transgenic mice at one day post injection [Fig. 6(b)] are similar in expression levels. At 1 week, fluorescence begins to appear as diffusely distributed across the organ [Fig. 6(c)]. As time progresses, we observed punctate expression of fluorescent signal that grows in intensity over time [3 weeks, 12 weeks, and 6 months, as shown in Figs. 6(d)–6(f), respectively]. Inspecting the resolving capabilities of the system in the early signal at 1 week [Fig. 6(g)] appears to have some minor texture to the pattern but few concentrated strong areas of signal. This may suggest that the markers are expressed broadly as cells differentiate across the organ. As time progresses, for example, 3 weeks [Fig. 6(h)], higher resolution inspection of the punctate pattern reveals that these bright sources of signal have long thin hair-like projections, consistent with the villus tissue structure of the luminal GI tract. This finding suggests that the cells expressing ZBP-89 differentiate and/or migrate up the length of the villus with time, but these cells may originate in stem-cell like populations in the crypt zones deeper in the tissues, giving rise to the early diffuse signal.

Fig. 6

Wide-field fluorescence images collected in study 2 for the duodenum and jejunum. (a) Wild-type mice up to 6 months and (b) transgenic mice 1 day post-induction show no noticeable signal. (c) Transgenic mice at 1 week post-induction show a diffuse distribution of signal, and the signal begins to concentrate into localized regions at (d) 3 weeks, (e) 12 weeks, and (f) 6 months. (g) Higher resolution inspection of the diffuse signal at 1 week [red box in panel (c)] shows texture patterns to the signal, whereas at (h) 3 weeks, the signal is localized to small thin regions that are thought to be the villus. Jej – jejunum and duo – duodenum. Panel (a) is the only wild type image and (b)–(h) transgenic mice at different time points.

To confirm these findings, Fig. 7 shows representative fluorescence microscopy images of cryopreserved tissue sections from our mouse model at each time point. The expression pattern of tdTomato is consistent with our wide-field results. We observe that tdTomato expression occurs more diffusely immediately following Cre recombinase induction, suggesting that ZBP-89 is expressed early on in the crypts of the epithelial mucosa. Over time, the tdTomato signal localizes to long hair-like projections, consistent with the distinct villi microstructure of the luminal mucosa. We also observe increased tdTomato expression in the proximal intestine, suggesting that ZBP-89 is more highly expressed in this organ compared to the stomach and distal GI tract.

Fig. 7

Fluorescence microscopy images of frozen tissue sections from transgenic mice previously imaged using wide-field fluorescence imaging. These images were taken with an Olympus BX53F epifluorescence microscope with a $200\times$ objective. TdTomato expression is consistent between wide-field and microscopic images with a gradient of higher to lower expression from proximal to distal intestine and increased signal intensity at longer time points following Cre recombinase induction with tamoxifen.

Figure 8 is a graphic detailing the aforementioned observed spatiotemporal gradients in transgenic mice. In summary, fluorescence expression increases with time and decreases from the proximal to distal intestine. To the left, representative images from one sample at each time point have been arranged in order of increasing temporal and spatial properties. To the right, an illustration depicts these trends in a simplified form. The color of the dots indicates increase in signal over time from 1 day to 3 months, as indicated by Fig. 4(a), and the density of dots indicates fluorescence abundance across all time points relative to the colon, as calculated in Table 3.

Fig. 8

A diagram showing trends in fluorescence in transgenic mice over location and time. The color of the dots indicates increase in signal over time from 1 day to 3 months, and the density of dots indicates fluorescence abundance across all time points relative to the colon. Images were collected during study 2. Duo – duodenum and Jej – jejunum.