Three-dimensional imaging of the enteric nervous system in human pediatric colon reveals new features of Hirschsprung disease

Abstract

Background and Aims:

Hirschsprung disease is defined by the absence of enteric nervous system (ENS) from distal bowel. Primary treatment is “pull-through” surgery to remove bowel that lacks ENS with re-anastomosis of “normal” bowel near the anal verge. Problems after pull-through are common and some may be due to retained hypoganglionic bowel (i.e., low ENS density). Testing this hypothesis has been difficult because counting enteric neurons in tissue sections is unreliable even for experts. Tissue clearing and 3-dimensional imaging provides better data about ENS structure than sectioning.

Methods:

Regions from 11 human colons and one ileal specimen resected during Hirschsprung disease pull-through surgery were cleared, stained with antibodies to visualize ENS, and imaged by confocal microscopy. Control distal colon from people with no known bowel problems were similarly cleared, stained, and imaged.

Results:

Quantitative analyses suggest age-dependent changes in myenteric plexus area, ENS ganglion area, percentage of myenteric plexus occupied by ganglia, neurons/mm², and neuron Feret diameter using human colon ranging from 3-days-old to 60-years-old. Neuron counting using 3D-images was highly reproducible. High ENS density in neonatal colon allowed reliable neuron counts using 500 × 500 μm² regions (36-fold smaller than in adults). Hirschsprung samples varied 8-fold in proximal margin enteric neuron density and had diverse ENS architecture in resected bowel.

Conclusion:

Tissue clearing and 3-dimensional imaging provides more reliable information about ENS structure than tissue sections. ENS structure changes during childhood. 3-dimensional ENS anatomy may provide new insight into human bowel motility disorders, including Hirschsprung disease.

Graphical Abstract

Lay Summary:

Enteric nervous system controls most bowel functions and when defective or missing, serious disease ensues. By making bowel translucent, enteric nervous system becomes visible. This new strategy could improve outcomes.

INTRODUCTION

Hirschsprung disease (HSCR) is a birth defect characterized by absence of enteric nervous system (ENS) from distal bowel. Because the ENS controls most aspects of bowel function, even short segments of “aganglionic” bowel (that lacks ENS ganglia), can cause intractable constipation, bilious vomiting, and predisposition to life-threatening sepsis. HSCR is treated via “pull-through surgery”, a technique first developed in the 1940s¹. In pull-through surgery, distal aganglionic bowel is removed and proximal ENS-containing bowel is re-connected near the anal verge. Despite this ostensibly curative surgery, HSCR-associated enterocolitis (HAEC) and functional obstructive symptoms are common before and after pull-through surgery^2,3. It is not well understood why some children have problems after pull-though surgery while others are symptom-free. A common hypothesis is that post-operative HAEC and obstructive complications occur because of incomplete removal of the transition zone, a region of decreased ENS density between “normal” and aganglionic bowel. Testing this hypothesis has been challenging because transition zone features are not easily defined using standard histology.

In the ENS, enteric neurons cluster into ganglia in two layers. Myenteric plexus ganglia between circular and longitudinal muscle control muscle contraction and relaxation. Submucosal plexus ganglia between circular muscle and bowel epithelium regulate epithelial function, blood flow, and immune cell activity in response to local stimuli. These interconnected neurons and associated glia are normally found along the entire bowel, but in children with HSCR, ENS is absent from distal bowel. One important problem for HSCR clinical pathology is reliance on thin (4–15 μm) tissue sections. In thin sections, enteric neurons cannot be reliably counted, even by experts using immunohistochemistry to highlight neurons⁴. Furthermore, tissue sectioning makes it very difficult to identify three-dimensional relationships between ENS cells clustered into small enteric ganglia of the myenteric and submucosal plexus. The low abundance (about 1:10,000 colon cells are ENS) and irregular spacing of ENS cells means many tissue sections need to be evaluated to adequately sample ENS anatomy⁵. Reconstruction of serial sections would also be needed to visualize organization of the ENS, because many ENS defects can only be appreciated when the whole plexus is visualized, at least based on murine disease models.

Here we tested the hypotheses that tissue clearing, antibody staining, and three-dimensional imaging could provide additional insight into HSCR mechanisms. We specifically hypothesized that ENS anatomy in small children might differ from ENS anatomy in adults and that ENS in children with HSCR might be very abnormal near proximal edges of resected tissue, at least some children. We also hypothesized that our three-dimensional imaging method could facilitate more accurate quantitative analysis of human ENS anatomy than traditional sectioning. To test these hypotheses, we took advantage of a method we recently developed to analyze human ENS anatomy without tissue sectioning⁵. We examined colon resected from children with HSCR as part of standard care and compared ENS anatomy of HSCR colon to pediatric organ donor and adult colons. We discovered that human colon ENS anatomy changes between infancy and adulthood even in people without known ENS disease. For HSCR colon, we found remarkably varied ENS features in the transition zone that would be difficult or impossible to appreciate in tissue sections.

METHODS

Human tissue acquisition:

Colon was acquired with Institutional Review Board approval from Children’s Hospital of Philadelphia (CHOP; IRB 13–010357) and Perelman School of Medicine at University of Pennsylvania (IRB 804376). Human HSCR colons were obtained either (a) with informed consent and full access to medical records or (b) de-identified with limited data (age, sex, type of surgery). HSCR specimens were all obtained during clinically indicated pull-through surgery, using tissue not required for clinical pathology. We did not ask surgeons to resect tissue beyond what was clinically indicated. We focused on HSCR colon from primary pull-through procedures where aganglionosis was restricted to colon, but examined ileum from one child with total colonic aganglionosis. Pediatric organ donor colon was obtained from the Gift of Life Donor Program (IRB exempt). Pediatric specimens were compared to previously acquired data from adult colon⁵.

Tissue processing:

HSCR specimens were transferred from the operating room to CHOP pathology at room temperature. Bowel lumen was opened longitudinally along the anti-mesenteric border. Three segments were excised for clinical pathology including proximal margin, distal margin, and a longitudinal strip all the way along the bowel (Fig. 1 A). Remaining tissue not required for clinical analysis was divided into “proximal neo-margin” (width approximately 2/3 of bowel circumference) and a longitudinal strip (width approximately 1/3 of bowel circumference). Within 1 hour of tissue initial arrival in clinical Pathology, these separate sections were placed into sterile ice-cold 1x phosphate-buffered saline (PBS) or in University of Wisconsin (UW) Belzer solution (NC0410019; Fisher Scientific, Waltham, MA). We received coded specimens in solution on ice. For organ donors, colons were placed immediately into UW Belzer solution on ice and transported to our laboratory. Organ donor colon lumen was opened longitudinally along the mesentery and cleared of stool. Descending colons from organ donors were evaluated to compare to HSCR pull-through segments since 80% of people with HSCR have aganglionosis restricted to recto-sigmoid colon. For most imaging, taenia were avoided when they could be identified because thick taenia are more challenging to clear and image.

Figure 1. Colon analysis strategy.

(A) Schematic of HSCR pull-through tissue obtained from clinical pathology. (B) Neomargin and longitudinal strip of bowel from HSCR colon resection stretched and pinned on a Sylgard^®-coated plate. (C) Fixed human colon with approximately 0.5 cm × 0.5 cm sections removed for staining and imaging. (D) Human colon resected from a child with HSCR after tissue clearing and staining with HuC/D (magenta) and nNOS (green) antibodies. (E) Tissue from (D) was imaged using confocal microscopy demonstrating the complete absence of ganglion cells in the entire imaged region.

Human colon immunostaining and clearing:

Our detailed clearing and staining protocol was published, including step-by-step methods with details about reagents and sample handling^5,6. Briefly, while specimens were in solution (UW Belzer or PBS), visceral fat was removed and tissue was stretched and pinned with serosa facing upward onto Sylgard 184 Silicone Elastomer (Dow, Midland, MI), using insect pins (Fig. 1 B). Tissue was fixed overnight (4% paraformaldehyde, 4°C). Then pins were removed. Tissue not immediately processed was stored in 50% glycerol, 50% PBS with 0.05% sodium azide at 4°C. Scissors were used to cut segments of fixed colon (Fig. 1 C). Stained regions were usually ~5 mm × 5 mm. Segments were initially washed in PBS (3 × 5 minutes, room temperature), permeabilized with 100% methanol (1 hour on ice), treated with Dent’s bleach (2 hours, room temperature), and placed in blocking solution (three days on a shaker, 37°C). This was followed by incubation in primary antibody (Supplemental Table 1) (14 days, 37°C) on a shaker. Tissue was then washed in PBS (2 × 2–3 hours, and 1x overnight) followed by incubation in secondary antibody (Supplemental Table 1) (3 days. 37°C) on a shaker, dehydrated in serial methanol dilutions, and cleared with benzyl alcohol-benzyl benzoate (BABB) (1:2) until tissue was translucent (Fig. 1 D). Tissue was mounted on glass slides in BABB for imaging (Fig. 1 E). For the 3-day-old organ donor we imaged full bowel circumference. For the 5-year-old organ donor, we imaged half the colon circumference. To image these large areas, we devised a new mounting method that flattens tissue more completely (Supplemental Fig. 1) and used ethyl cinnamate as mounting medium instead of BABB, which preserved tissue for imaging for months.

Image acquisition and analysis:

Most cleared tissue was imaged on a Zeiss LSM 710 confocal microscope (10x/0.3 and 20x/0.8 Plan-Apochromat objectives, Zen software version 2.3 14.0.14.201; Zeiss, Oberkochen, Germany). Z-axis intervals were 4 μm (10x objective) or 1 μm (20x objective). For each segment imaged, confocal Z-stacks were obtained with 10x objective using tile scan (5% overlap) and stitch features to visualize large regions of full-thickness tissue. Multi-channel images were acquired sequentially using laser-scanning operated under multitrack. Excitation/long-pass emission filters were Alexa Fluor 488, Alexa Fluor 594, and Alexa Fluor 647. Full and half-circumference imaging from organ donor colon was obtained on a Zeiss LSM 980 confocal microscope (Plan-Apochromat 10x/0.45 WD=2.0 M27, Zeiss Zen Blue 3.5 software, variable laser power) at 2 μm intervals (5% overlap). Full-thickness Z-stacks were used for myenteric plexus area and submucosal plexus assessments. Four additional images per segment (708 × 708 μm²) were obtained with a 20x objective, one in each quadrant. These Z-stacks were utilized for neuron density and subtype counting. Image analysis was performed using ImageJ (Java1.8), including features of Z-project, scale bar, polygon selection, rectangle selection, specify, straight line tool, measurements, cell counter, and ROI manager.

Quantitative analysis of myenteric plexus density:

Two-dimensional flattened Z-stacks obtained with a 10X objective were manually outlined. The myenteric plexus was defined as myenteric ganglia and thick neuron fiber bundles between ganglia. Areas selected for analysis were the largest contiguous rectangular areas in which plexus was visible in the plane of flattened Z-stacks. Ganglia were defined as clusters of >1 neuron soma separated by <1 soma diameter. Individual ganglia were separately outlined. We calculated percent of bowel wall containing plexus, percent of bowel wall containing ganglia, and percent of the plexus containing ganglia.

Quantitative analysis neuron subtypes and size:

Myenteric plexus neurons (HuC/D+) were counted manually in 3-dimensional 20X Z-stacks. For each bowel region we counted all neurons in randomly selected 500 μm × 500 μm regions (one region in each quadrant of stained tissue). Data are presented in neurons per mm². Neurons expressing neuronal nitric oxide synthase (nNOS, NOS1) were counted as nitrergic. Neuron size was measured as the longest Feret diameter.

Statistics:

Analyses employed Prism 7 (GraphPad Software, San Diego, CA) and RStudio Version 1.4.1717 (ggplot2, ggpubr, tidyverse, broom, and dplyr packages). D'Agostino-Pearson tests were used to assess normality. Unpaired t-tests were used to compare means of normally-distributed data. Paired t-tests were used to compare means of repeated counts of the same images for inter-observer variability. Mann-Whitney tests were used to compare means of non-normally distributed data. Data are presented as mean +/− standard error of the mean (SEM). Best fit lines are plotted as linear regression.

RESULTS

We obtained tissue from 16 children with HSCR undergoing primary (i.e., first) pull-through surgery, and from three human organ donors (Supplemental Table 2). The organ donors were 3-days, 5-years, and 18-years-old and had no known bowel disease. Of the 16 HSCR pull-through subjects, two had total colonic aganglionosis (HSCR12, HSCR16), one has Down syndrome (trisomy 21) (HSCR13), one has multiple endocrine neoplasia 2A (MEN2A) (HSCR11) (RET (c.1858T>G, p.Cys620Gly in exon 10)) (HSCR11), and one has cystic fibrosis (HSCR8). Two were not evaluated because additional colon was resected during primary surgery, but not available to our laboratory (HSCR14, HSCR15). Colon from the child with Down syndrome had no neurons in the proximal resection neo-margin, so quantitative analyses were not pursued. Terminal ileum from HSCR12 was imaged and compared to control ileum from one organ donor. For the 12 HSCR pull-through segments imaged and analyzed, there were 4 different surgeons. These data highlight the complexity of HSCR and HSCR surgery. To interpret data from HSCR resections, we evaluated ENS anatomy in distal colon from pediatric organ donors presumed to have normal ENS anatomy (Control1, Control2, Control3) and in adult descending colons (Adult1 to Adult4) previously analyzed by our group⁵.

Imaging ENS anatomy in human colon without tissue sectioning

To visualize interconnections between enteric ganglia of myenteric and submucosal plexus, we pursued 3-dimensional confocal imaging in cleared human colon without sectioning. Our surgeons performed their usual pull-through surgery. Our clinical pathologists took tissue needed for clinical care and provided us with remaining bowel that would normally be discarded (Fig. 1 A). For organ donors we analyzed descending colon, the most common region for proximal resection margin in HSCR pull-through surgery. Tissue was pinned flat and fixed (Fig. 1 B). Then 5 mm × 5 mm pieces were cut with a scissors (Fig. 1C), stained with antibodies, and cleared (Fig. 1D) to make colon translucent before confocal imaging (Fig. 1E). Antibodies to HuC/D and neuronal nitric oxide synthase (nNOS, NOS1) reliably stained human colon with our method. HuC/D antibody stains ELAV family proteins in the cell body marking all enteric neurons. nNOS antibody stains nerve cell bodies and neuronal fibers for inhibitory motor neurons and descending interneurons in myenteric plexus⁷. The 416 confocal Z-stacks we generated are available as supplemental data (Supplemental Table 3). Selected Z-stacks were converted to videos to show images analyzed (Supplemental Videos 1–31b, 55 total, Supplemental Table 3). All videos and images are available via the SPARC Data Portal, https://doi.org/10.26275/whgt-j5vy. Staining tissue in this manner allowed for visualization of both overall ENS structure and two neuron classes: nitrergic neurons (nNOS-positive) and non-nitrergic neurons (nNOS-negative).

Characterization of pediatric myenteric plexus in organ donor colon

Since most people with HSCR have pull-through surgery as infants or toddlers, interpreting HSCR ENS anatomy requires data about normal human colon ENS at various ages. Based on murine data, we hypothesized that quantitative features of ENS anatomy we described in adult colon might differ from pediatric ENS^5,8. To test this hypothesis, we first compared HuC/D and nNOS antibody-stained myenteric plexus from 3-day-old organ donor colon (Fig. 2 A) to adult colon myenteric plexus (Fig. 2 D). The 3-day-old colon ENS appeared far denser and comprised a larger percentage of bowel wall area than adult colon ENS. To pursue this observation, we determined the percent of bowel wall occupied by myenteric plexus (Fig. 2 E, F), percent of bowel wall occupied by myenteric ganglia (Fig. 2 G, H), and proportion of myenteric plexus occupied by ganglia (Fig. 2 I, J) within the myenteric plexus plane (Supplemental Table 4). Similar analyses were pursued using 5-year-old and 18-year-old organ donor distal colon (Fig. 2B, C) and using distal colon from 28-, 36-, 37-, and 60-year-old adults with no history of bowel dysfunction. These data show the proportion of bowel wall occupied by myenteric plexus and by myenteric ganglia progressively declined from birth to 28 years of age, as did proportions of myenteric plexus occupied by ganglia (Fig. 2 E–J, Supplemental Table 4). We initially tried fitting 3-day-old to 60-year-old colon data to a single linear regression (Fig. 2 E, G, I) but discovered data fit better to separate pediatric and adult linear regressions (Fig. 2 F, H, J). Collectively these data confirm the need for age-specific quantitative data about ENS anatomy, and suggest that myenteric plexus and ganglion cell density declines as children grow and colon enlarges.

Figure 2. Normal colon myenteric plexus anatomy varies with age.

(A-D) Flattened Z-stacks of myenteric plexus stained with antibodies to HuC/D (magenta) and nNOS (green). (A) 3-day-old, (B) 5-year-old, and (C) 18-year-old descending colon from organ donors with no known bowel dysfunction. (D) 60-year-old descending colon adjacent to resected diverticulitis. Tissue sections looked normal by hematoxylin and eosin staining. Scale bar = 500 μm. (E-J) Quantitative analyses of myenteric plexus using maximum intensity Z-projections. (E, F) Percentage of flattened Z-stack containing myenteric plexus (ganglia plus thick nerve fiber bundles). (G, H) Percentage of flattened Z-stack containing myenteric ganglia (clusters of nerve cell bodies). (I, J) Percentage of myenteric plexus within the Z-stack occupied by myenteric ganglia. (E, G, I) Linear regression shows a single best fit line based on all data. (F, H, J) Same data as (E, G, I) showing separate linear regressions for 3-day-old to 28-year-old and for 28-year-old to 60-year-old data sets.

Minimal inter-observer variability for neuron counting in Z-stack images

Neuron counting is notoriously difficult for human ENS, leading to discrepancies even between expertly trained pathologists^4,9. We hypothesized that neurons in 3-dimensional confocal images could be more reliably counted than neurons in tissue sections currently used by pathologists (Fig. 3). To test this hypothesis a new investigator counted neurons in 15 adult colon confocal 20X Z-stacks our group previously analyzed⁵. These images were stained for HuC/D, nNOS, and choline acetyltransferase (ChAT), permitting us to define 4 neuron subtype classes. Counts were performed blinded so the new investigator was unaware of prior results. When data from 15 regions were combined, there was no statistically significant differences between any neuron subtype counts for the two observers (Fig. 3 A). For individual regions, neuron counts agreed closely for all HuC/D+ neurons (R-squared 0.94), total nNOS+ neurons (R-squared 0.84), and total ChAT+ neurons (R-squared 0.74, Fig. 3 B), but varied more for scoring “Neither nNOS nor ChAT” (R-squared 0.61), probably because ChAT staining is not ideal as we previously reported (Supplemental Table 5)⁵. These observations suggest that 3-dimensional imaging facilitates reliable neuron counting when staining is effective.

Figure 3. 3D confocal Z-stacks provide reliable estimates of neuron density and percentage of nNOS expressing neurons in human colon.

(A) Comparison of myenteric plexus cell counts by two independent observers using the exact same colon 3D confocal Z-stacks. Neuron subtypes were identified by staining with HuC/D, nNOS, and ChAT antibodies. Circles show neuron counts from each of 15 images (3 adult subjects, 5 images each, dimensions 708 × 708 μm × 40–123 μm). All subtype counts by each observer were normally distributed. All inter-observer differences were statistically equivalent by paired t-test. Lines indicate mean neuron counts. (B) Plot shows neuron counts for 15 individual images evaluated by two observers. R-squared values varied among neuronal subtypes and are reported in the text. (C) Flattened 126 μm thick Z-stack of myenteric plexus from colon of 3-day-old human organ donor divided into 49 squares (100 × 100 μm each). (D) Each circle shows counts obtained from a region of the size indicated on the X-axis from the Z-stack depicted in image (C). Lines indicate mean values. (E) Quantitative analysis of myenteric plexus neuron density using Z-stacks obtained with a 20x objective. All neurons were counted in 4–5 regions (500 μm × 500 μm each) per individual. Linear regression shows best fit lines for 3-day-old to 28-year-old and for 28-year-old to 60-year-old colon. (F) Percentage of myenteric neurons that are nitrergic (HuC/D+ nNOS+) at various ages. Error bars are +/− SEM.

Neuron count reliability as a function of area using Z-stack images

Because enteric neurons cluster into ganglia, neuron counts depend on which region is counted and how large a region is analyzed. To assess the minimum bowel area required for reliable neuron counts in a 3-day-old infant, a representative 700 µm × 700 µm (20X) image was divided into 49 squares (each square = 100 µm × 100 µm) (Fig. 3 C). Neurons were manually counted in each square revealing convergence of neuron counts/mm² when areas analyzed measured 5 squares × 5 squares (Fig. 3 D). To ensure reliability in our data, we therefore counted neurons in 500 μm × 500 μm areas in pediatric colons for subsequent analyses.

Neuron density, percent nitrergic neurons, and neuron size differ in infants and adults

Analyzing regions between taenia, we found that myenteric plexus neuron density in control colons declined during childhood (Fig. 3 E, Supplemental Table 6), while the percentage of myenteric neurons that were nNOS+ increased (Fig. 3 F, Supplemental Table 6). Myenteric neurons in the 3-day-old control colon were smaller than in older colons (Supplemental Fig. 2A) based on the longest neuron Feret diameter (3-day-old = 13.0 +/− 0.46 µm, 5-year-old = 20.7 +/− 0.92 µm, 18-year-old = 22.8 +/− 0.99 µm, p < 0.0001 versus 3-day-old for both older ages). There was no statistically significant difference in neuron Feret diameter between 5-year-old and 18-year-old colons (p = 0.115). Additionally, there was no difference in neuron size between nitrergic and non-nitrergic neurons in control pediatric colons assessed (Supplemental Fig. 2B).

Enteric neuron density is similar in regions with and without taenia

To evaluate how ENS structure varied around the bowel circumference, we imaged the entire bowel width in the 3- day-old organ donor after HuC/D and nNOS antibody staining (Fig. 4, Supplemental Fig. 3). Enteric neuron density varied little across the bowel circumference including in regions with and without taenia (neurons/mm²: taenia 1889.0 +/− 22.2, non-taenia 1962.0 +/− 78.9, P = 0.34). In 5-year-old organ donor we could only visualize half the bowel circumference before files became too large for imaging software to open. ENS density also appeared similar around the 5-year-old colon circumference and neuron counts were similar in regions with and without taenia (neurons/mm²: taenia 806.0 +/− 29.41, non-taenia 770 +/− 35.34, P = 0.69, Mann-Whitney U test). Because thick muscle bands in taenia make imaging more challenging, all other images were from regions between taenia.

Figure 4. Myenteric neuron density is similar in regions with taenia and between taenia.

Full thickness descending colon from organ donors stained with HuC/D (magenta) and nNOS (green) antibodies imaged at 10X (2 μm intervals between confocal slices). (A-G) 3-day-old and (H-N) 5-year-old colons were selected from larger full-circumference or half circumference (respectively) Z-stacks in Supplemental Figure 3C-F. (A, B, D-G, H, I, K-N) Flattened Z-stacks from the myenteric plexus region. (A, B, H, I) Dotted lines separate regions with or without taenia. (C, J) Gross tissue indicating regions imaged. (B) HuC/D of region in (A). (I) HuC/D or region in (H). (D-G) Enlarged images of regions in boxes at the left (D, E) or right (F, G) in panels (A, B). (K, L, M, N) Enlarged images of regions in boxes at the left (K, L) or right (M, N) in panels (H, I). (A, H) Scale bar = 1000 μm. (D, E, F, G, K, L, M, N) Scale bar = 200 μm.

Myenteric plexus in HSCR proximal resection margin varied markedly among children

One primary goal of evaluating ENS anatomy in control distal colon (Figs. 2–3, Supplemental Figures 2, 4, Supplemental Videos 29–31 and 52–55, Supplemental Table 3 confocal Z-stacks 231–284 and 413–416, Supplemental Tables 3–6) was to put in context ENS anatomy in children with known or suspected ENS disease, including HSCR. For HSCR, surgeons remove distal bowel that lacks ENS and attempt to remove the hypoganglionic “transition zone” that is presumed dysfunctional. The decision about how much bowel to remove is made in the operating room based on frozen sections (15 μm) from small biopsies, but these sections provide limited views of ENS anatomy (Fig. 5 A, Fig. 6 Y, Z, Fig. 7 U, V, Supplemental Figure 5I). To gain deeper insight into HSCR anatomy, colon 0.5 to 2.4 cm from proximal pull-through resection margins was cleared, stained with antibodies (HuC/D, nNOS), and imaged by confocal microscopy. Full thickness regions approximately 5 mm × 5 mm were imaged. Confocal images demonstrated striking variability in HSCR proximal margin ENS anatomy (Fig. 5 B–K). Myenteric plexus area in proximal margins varied from 16.0% to 59.7% of total colon area, while ganglia occupied 14.7% to 70.4% of myenteric plexus (Supplemental Table 4). The mean density of myenteric plexus neurons varied over 8-fold (range = 234 to 1972 neurons per mm²), with no apparent correlation to age at the time of surgery (Fig. 5 B–L, Supplemental Table 6). The percentage of nitrergic neurons at proximal margins of children with HSCR also varied considerably (range = 28–63%) but these percentages were similar to values in organ donor colon (Supplemental Figure 4, Supplemental Table 6).

Figure 5. Colon myenteric plexus in Hirschsprung disease (HSCR) varies considerably at the proximal surgical margin.

(A) Hematoxylin and eosin stained clinical pathology section of colon proximal margin from a HSCR pull-through surgery (HSCR1, 13 mm from proximal margin). Scale bar = 250 µm. ENS is difficult to see in these sections even when myenteric plexus region (boxed) is enlarged ~ 3.5-fold (inset, Scale bar = 50 μm). Arrow points to a myenteric neuron. (B-K) Representative 20X images from colon myenteric plexus of HSCR 1 to HSCR10 were stained with antibodies to HuC/D (magenta) and nNOS (green). The Key in Figure 5L indicates how each HSCR number matches Figures 5B–K. Images show flattened 3-dimensional Z-stacks at the proximal neomargin of HSCR pull-through segments. Scale bar = 250 µm. (L) Neurons per mm2 in Z-stacks containing myenteric plexus at varied distances from the proximal resection margin. Dots indicate mean neuron counts in four 500 × 500 µm² regions per colon at each location. Error bars are +/− SEM. Horizontal dashed lines indicate mean neuron counts for the 3-day-old (Control1) and 5-year-old (Control2) colon ENS. (L) Key indicates colors for the lines in (L) and correlation between HSCR numbers (HSCR1 to HSCR10) and Figure 5 letters (B-K). “HSCR1 (B)” in the Key indicates that the image in Figure 5B was from HSCR1. The same logic applies to all HSCR numbers.

Figure 6. HSCR transition zone anatomy for HSCR1 and HSCR3.

(A-X) Representative flattened Z-stacks show myenteric and submucosal plexus at various locations in resected colon stained with antibodies to HuC/D (magenta) and nNOS (green). The second and fourth columns show magnified views of boxed areas in first and third columns respectively. Schematics to the left of images indicate distance in cm from proximal resection margin and relative location (% = distance from proximal margin/length of resected bowel × 100). (Y, Z) Hematoxylin and eosin stained colon sections (4 μm) from the same region as the 3-dimensional confocal images on the same horizontal row. Scale bars = 500 µm.

Figure 7. HSCR transition zone anatomy for HSCR5 and HSCR8.

(A-X) Representative flattened Z-stacks show myenteric and submucosal plexus at various locations in resected colon stained with antibodies to HuC/D (magenta) and nNOS (green). The second and fourth columns show magnified views of boxed areas in first and third columns respectively. Schematics to the left of images indicate distance in cm from proximal resection margin and relative location (% = distance from proximal margin/length of resected bowel × 100). (U, V) Hematoxylin and eosin stained colon sections (4 μm) from the same region as the 3-dimensional confocal images on the same horizontal row. Scale bars = 500 µm.

Complex features of HSCR transition zone ENS anatomy

To characterize the transition zones of HSCR colon (HSCR1 to HSCR10), we stained tissue from proximal resection margins plus 1–2 additional distal regions with antibodies to HuC/D and nNOS (Figs. 6, 7, Supplemental Figure 5–8). Assessed transition zones had thin nerve fiber bundles between ganglia in the myenteric and submucosal plexus, in addition to the gradual disappearance of neurons distally. In some children, the transition zone appeared to terminate at the same distance from the proximal margin in both myenteric and submucosal plexus (Fig. 6 E–H, Fig. 7 I–L, Supplemental Fig. 6 A–H). In other children, the transition zone extended over a longer distance in the myenteric plexus compared to the submucosal plexus (Supplemental Fig. 8 E–L), and in some cases the opposite was true (Supplemental Fig. 8 U–X). In one child, a cluster of HuC/D+ cells organized atypically was present in muscularis propria within a region of aganglionosis (Supplemental Fig. 6 Q–T). Analysis of neuron density at various distances from proximal margins demonstrated different degrees of hypoganglionosis throughout the transition zone. The transition zone-aganglionosis border ranged from 3.2 cm to >13 cm distal to the proximal resection margin (Fig. 5 L). These features of the transition zone would be very challenging to appreciate using any standard clinical pathology technique. The difficulty assessing ENS anatomy is further emphasized by our analysis of HSCR11, a 5.3-month-old with MEN2A (Supplemental Figure 5). Although we found 1451 neurons/mm² at the proximal resection margin of colon we received (similar to 3-day-old organ donor control), intra-operative assessment of ENS anatomy led to a decision to remove an additional 23.5 cm of more proximal colon (not available to us). Finally, we imaged distal ileum from a child with total colonic aganglionosis (HSCR12). Compared to ileum from the 5-year-old organ donor (Control2), ENS appeared sparse in the HSCR ileum (Supplemental Figure 9).

DISCUSSION:

Anatomic defects underlying bowel motility disorders remain incompletely understood in part because human ENS is difficult to visualize through the opaque colon wall. While most prior ENS studies use tissue sectioning, this approach provides very limited views of ENS anatomy and can be misleading because enteric neurons and glia cluster into ganglia. This clustering means that in normal human colon, some sections have many enteric neurons and other adjacent sections have no neurons⁵. Furthermore, even for experts, it is challenging to unambiguously identify enteric neurons in tissue sections as elegantly demonstrated by Swaminathan and Kapur⁴. These problems probably explain the 150-fold range for “normal” human colon enteric neuron density reported in the literature⁴. To overcome these problems, a few prior studies employ tissue microdissection, which provides much better views of ENS anatomy^10,11. However, microdissection requires specialized skills, disrupts normal structures, does not visualize submucosal neurons well, and is labor intensive. In contrast, we visualize ENS in human colon without sectioning using a recently developed and simple method for tissue clearing, immunohistochemistry, and confocal imaging. Our analyses demonstrate substantial changes in ENS anatomy during childhood, indicating the need for age-specific normal ranges. Our studies also show remarkable diversity of ENS structures in children with HSCR, an observation that might underlie variable outcomes after HSCR pull-through surgery.

Colon ENS anatomy changes as children grow

To determine if ENS anatomy at the proximal resection margin predicts outcomes after HSCR surgery, we first need to define normal ENS anatomy. Our data show that “normal” varies depending on the age of the individual, confirming and extending prior results in rodents and humans^12–15. We focused on distal colon as this region is most commonly resected from children with HSCR, recognizing that ENS anatomy varies between bowel regions¹⁶. Consistent with prior studies¹⁵, our images show neonatal human colon myenteric neurons in closely spaced ganglia. Between 3 days and 28 years of age there was a progressive decline in the percentage of colon containing myenteric plexus, the percentage of myenteric plexus containing ganglia, and the density of myenteric neurons. Individual myenteric neurons were also smaller in young children than in adults. These data provide a framework for understanding human distal colon ENS anatomy and make it obvious how inadequate sampling might provide misleading information. Because neonatal myenteric ganglia are closely spaced, our analyses suggest that reliable enteric neuron counts can be obtained from a 500 × 500 μm² (250,000 μm²) region in neonates. In contrast, our prior adult data suggested 3 × 3 mm² regions were needed for reliable neuron counts (a 36-fold larger area). Consistent with our data Swaminathan and Kapur showed using tissue sections showed that reliable data from 8-week-old human rectum required neuron counts in at least 5 full circumference regions⁴. Since Swaminathan and Kapur evaluated 4 μm sections separated by at least 24 μm, we estimate that 5 full circumference regions covered 500,000 μm² (5 sections × 4 μm thick × 25,000 μm [circumference estimate] = 500,000 μm²) or 3,500,000 μm² (5 sections × [4 μm thick + 24 μm spacing] × 25,000 μm [circumference estimate] = 3,500,000 μm²) if including spacing between sections. Neither of these analyses consider the possibility that neuron density might vary around the bowel circumference as suggested by Swaminathan and Kapur. However, our full circumference or half-circumference images from 3-day-old and 5-year-old organ donor colon show ENS ganglia distributed all the way around the bowel wall with similar enteric neuron density in regions with and without taenia.

Three-dimensional imaging permits highly reproducible quantitative analysis

Another key observation from Swaminathan and Kapur is that counting enteric neurons is difficult in thin tissue sections⁴. Even with nicely stained immunohistochemically marked neurons, specific criteria for what should be counted, repeated training, and expert pathologists, their inter-observer counts varied by up to 5-fold for the exact same slides. In contrast, using our 3-dimensional confocal images, HuC/D stained neuron counts varied on average by 7% (and at most by 20%) for the exact same image analyzed by two observers. This >25-fold improvement in the accuracy of neuron counts probably occurs because it is easier to unambiguously identify an enteric neuron in 3-dimensional Z-stacks than in thin sections. Our 3-dimensional neuron counts for nNOS+ enteric neurons also varied by an average of 8% between observers (and at most 18%) for any image. Both HuC/D and nNOS antibodies stain tissue beautifully, but even for ChAT antibody, where staining is more challenging to interpret, average inter-observer counts varied by only 28%. Thus, in addition to providing data about how the ENS is organized, neuron counting is dramatically better in confocal Z-stacks than in tissue sections.

New insight into ENS anatomy in Hirschsprung disease

Children with HSCR undergo pull-through surgery where distal aganglionic bowel is removed and surgeons determine the proximal resection margin based on small biopsies and frozen sections. In our studies, the ENS in colon just distal to the resection margin varied dramatically among children with HSCR. Some specimens had only a few small ganglia with very thin nerve fibers, while others had thick nerve fiber bundles connecting large dense ganglia. Quantitative analysis showed neuron density similar to pediatric organ donors in 6 of the 10 proximal margins evaluated (within ~30% of controls). The other 4 HSCR colons had 3.4- to 8-fold fewer neurons near the proximal resection margin compared to the most densely innervated colon resected. While it might be predicted that low neuron density leads to poor post-operative outcomes, HSCR1 had the fewest enteric neurons/mm² near the proximal resection margin and did well after surgery. During 31 months follow up, he was hospitalized only once (5 months after pull-through) because of vomiting and diarrhea. While these are Hirschsprung-associated enterocolitis symptoms, he had norovirus, respiratory syncytial virus and rhinovirus at the time of admission making infectious enteritis most likely. Other than this episode, HSCR1 ate well, and grew well without obstructive symptoms or enterocolitis after pull-through surgery. These observations suggest we have more to learn about anatomic features that predict good or bad outcomes after HSCR surgery.

Three-dimensional ENS imaging also revealed other features that would be difficult to appreciate in tissue sections. First, length of colon containing enteric neurons varied widely among pull-through resections (2–15 cm). This suggests some children might have done well with shorter resections if ENS could be visualized prior to or during pull-through surgery. This hypothesis is highlighted by the observation that HSCR11 had an additional 23.5 cm of proximal colon resected even though neurons were abundant in the proximal margin of the colon we received. Second, density of myenteric neurons differed markedly over short distances for some children. For example, for HSCR4, a region 4 cm from the proximal resection margin had 1150 neurons/mm² whereas a region 2.3 cm more distal had only 42 neurons/mm². Similarly, for HSCR2, neuron density was 426 neurons/mm² at 3.5 cm from the proximal resection margin and only 17 neurons/mm² in a region 1.5 cm more distal. Abrupt changes in neuron density like this might explain why HSCR13 had no neurons at the proximal neo-margin we analyzed, since clinical pathology retained the true proximal margin that contained myenteric and submucosal neurons in approximately 80% of HSCR13 sampled circumference. Interestingly, HSCR13 was admitted for enterocolitis 6 months after pull-through surgery. Third, submucosal and myenteric neuron density may differ markedly for defined transition zone regions. In 2 of our resections (HSCR2, HSCR9), we found myenteric neurons in distal colon that lacked submucosal neurons and one child had submucosal neurons in a distal region with nearly absent myenteric neurons (HSCR4). Consistent with our results, analysis by other investigators of circumferential HSCR sections showed myenteric neurons up to 2 cm distal to submucosal neurons in 19/59 children and submucosal neurons distal to myenteric plexus in 1 of 59¹⁷. These observations confirm the value of looking for submucosal and myenteric neurons when evaluating for ganglion cells. Finally, in three children (HSCR2, HSCR7, and HSCR9) we found small patches of myenteric neurons in otherwise aganglionic bowel. Sections through these regions might be interpreted as “ganglion cells present”, leading to retained aganglionic bowel. Collectively these data suggest that whole mount imaging could provide substantial insight into HSCR symptoms after pull-through surgery.

Limitations:

It would be valuable to have data from a more pediatric organ donor colons to define variability in ENS structure in children with normal ENS anatomy. Pediatric donor colon specimens were rare during our period of tissue collection, but we avoided analyzing colon from children with significant medical problems (e.g., anorectal malformation) since this tissue may not be normal. We also need to define in more detail how ENS structure varies along the length of the bowel. For example, our adult surgical biopsies were all from “distal colon”, but we do not know if they were from descending or sigmoid colon. In addition, current data set do not allow us to correlate anatomy with post-surgical outcomes, which will require a larger set of HSCR specimens. Finally, although our approach is technically simple and reproducible, it cannot be used intraoperatively because staining takes three weeks and manual counting is time consuming. Our approach could be applied if HSCR surgery were performed in 2 stages (i.e., initial ostomy followed months later by pull-through surgery), but it would be ideal to have methods to visualize ENS anatomy in three dimensions while in the operating room. Confocal laser endomicroscopy or other new methods might solve this problem,¹⁸ if anatomic features predicting good outcomes could be unambiguously defined.

Conclusion:

Managing HSCR and other disorders of the enteric nervous system can be very challenging and current clinical pathology provides limited insight into ENS anatomy. Our new strategy evaluating the transition zone in HSCR without sectioning provides much more information about ENS structure. We hope that this approach can be leveraged to improve outcomes in HSCR and other bowel motility disorders.

What you need to know:

BACKGROUND AND CONTEXT

Clinical pathology relies on thin tissue sections that provide inadequate and unreliable information about enteric nervous system anatomy. Tissue clearing, immunohistochemistry and three-dimensional confocal imaging provide much better data.

NEW FINDINGS

Enteric nervous system anatomy changes throughout childhood. 3-D imaging facilitates visualization of anatomy and identification of neurons, which could be valuable for bowel motility disorders including Hirschsprung disease.

LIMITATIONS

Additional study is needed to define anatomic features that correlate with good or bad post-surgical outcomes in Hirschsprung disease. This technique requires 3 weeks and cannot be used intraoperatively.

CLINICAL RESEARCH RELEVANCE

The enteric nervous system controls most aspects of bowel function. Enteric nervous system defects cause Hirschsprung disease, chronic intestinal pseudo-obstruction, and other bowel motility disorders. Many features of enteric nervous system anatomy cannot be seen at all in thin tissue sections but are easily appreciated via a simple clearing, staining, and 3-dimensional imaging strategy that is ideal for clinical research.

BASIC RESEARCH RELEVANCE

For most bowel motility disorders, anatomy of cells that control motility remains poorly defined because these cells are buried in an opaque bowel wall. By making the bowel translucent, staining structures of interest, and confocal imaging, diverse enteric nervous system features are readily visualized. These features help define disease mechanisms, a necessary step for development of mechanism-based therapy.

Abbreviations:

HSCR

Hirschsprung disease

ENS

Enteric nervous system

HAEC

Hirschsprung-associated enterocolitis

HuC/D

Human HuC / HuD neuronal protein

nNOS

Neuronal Nitric Oxide Synthase

CHAT

Choline Acetyltransferase

ROIs

Regions of Interest

SEM

Standard error of the mean

MEN2A

Multiple endocrine neoplasia 2A