Dlx1/2 mice have abnormal enteric nervous system function

24 Decades ago, investigators reported that mice lacking DLX1 and DLX2, 25 transcription factors expressed in the enteric nervous system (ENS), die with possible 26 bowel motility problems. These problems were never fully elucidated. We found that 27 mice lacking DLX1 and DLX2 (Dlx1/2-/mice) had slower small bowel transit and 28 reduced or absent neurally-mediated contraction complexes. In contrast, small bowel 29 motility seemed normal in adult mice lacking DLX1 (Dlx1-/-). Even with detailed 30 anatomic studies, we found no defects in ENS precursor migration, or neuron and glia 31 density in Dlx1/2-/or Dlx1-/mice. However, RNA sequencing of Dlx1/2-/ENS 32 revealed dysregulation of many genes, including vasoactive intestinal peptide (Vip). Our 33 study reveals a novel connection between Dlx genes and Vip and highlights the 34 observation that dangerous bowel motility problems can occur in the absence of easily35 identifiable ENS structural defects. These findings may be relevant for disorders like 36 chronic intestinal pseudo-obstruction (CIPO) syndrome. 37


Introduction
this hypothesis, we stained P0 bowel with antibodies that distinguish many types of 244 enteric neuron. We observed no differences in ratios of neurons expressing myenteric 245 plexus neurotransmitter markers nitric oxide synthase (NOS), calretinin, γ-aminobutyric 246 acid (GABA), or vasoactive intestinal peptide (VIP; Figure 6A-E, G-K, M-Q) or the 247 submucosal plexus neurotransmitter VIP ( Figure 6F, L, R). Since we lacked reliable 248 antibodies for choline acetyl-transferase (ChAT), we bred and analyzed Dlx1/2;ChAT-249 EGFP-L10a mice but found no difference in EGFP+ neuron number ( Figure 6A Since our detailed anatomic analyses failed to reveal obvious structural problems 271 of the ENS of Dlx1/2-/-mice, we next hypothesized that dysmotility in Dlx1/2-/-mice is 272 due to altered gene expression not discernable at the level of ENS anatomy. We tested 273 this hypothesis by RNA-seq of pooled ENCDC isolated from E14.5 and P0 bowel using 274 fluorescence activated cell sorting (FACS). We chose these ages for theoretical and 275 technical reasons. E14.5 was selected because Dlx1 and Dlx2 are expressed in murine 276 ENS by E12.5 (10, 17), and we wanted to evaluate early effects of mutations, rather than 277 secondary effects that occur as development proceeds. Furthermore, at E14.5 there are 278 many ENCDC for analysis, and neuron subtype markers are starting to appear as enteric 279 neural networks are established. We also analyzed P0 ENS to match our functional 280 experiments performed using neonatal bowel. For our E14.5 analysis, we bred Dlx1/2 281 mice to an Ednrb-L10A-GFP reporter line that expresses a fluorescent protein in ENCDC 282 at E14.5 (32) and performed RNA-seq on FACS-sorted ENCDC from E14.5 Dlx1/2-/-283 ;Ednrb-L10A-GFP Gfp/wt (mutant)  We identified 22 dysregulated genes in Dlx1/2-/-E14.5 ENCDC using q < 0.1 as 288 a statistical threshold after filtering out genes with low expression (average expression < 289 1) and fold change (|log 2 fold change| < 1; Figure 7A-B). Given that DLX2 regulates 290 Zfhx1b in CNS, we had hypothesized that Zfhx1b might be dysregulated in the Dlx1/2-/-291 ENS, but we found no evidence of altered Zfhx1b levels. Remarkably, however, we 292 observed ~84% reduction in mRNA levels for the ENS neurotransmitter vasoactive 293 intestinal peptide (Vip) and a 3-fold increase in abundance of mRNA for the 294 neurotransmitter proenkephalin (Penk) in small bowel ENCDC from Dlx1/2-/-mice 295 compared to WT animals. We also observed changes in receptors (Lifr), extracellular 296 matrix genes (Mmp2, Col3a1), cytoskeletal regulators (Dnm3, Capn6, Fmn1), neurite 297 growth factors (Ptn), and glia-associated genes (Plp1, Lrp1b; Figure 7B). Since 298 neurotransmitter dysregulation might most directly explain bowel dysmotility in Dlx1/2-299 /-mice, we attempted to validate altered levels of Vip and Penk mRNA using quantitative 300 RT-PCR on independently isolated E14.5 mutant and control samples ( To further explore the VIP-expressing enteric neuron subpopulations, we bred to 308 generate Dlx1/2;Vip-IRES-Cre;R26R-TdTomato mice, which produce TdTomato in cells 309 expressing Vip. Although we had already counted VIP immunoreactive enteric neurons, 310 14 VIP is primarily in neurites, making cell counting suboptimal. We discovered that 311 TdTomato+ neurons in mid-SI are significantly reduced in absolute cell density (18.4%) 312 and relative to total neuron number (13.2%) in myenteric plexus of Dlx1/2;Vip-IRES-313 Cre;R26R-TdTomato mice ( Figure 8A-D). Furthermore VIP reporter (TdTomato+) 314 neuron density was dramatically reduced in absolute (51.5%) and relative terms (41.0%) 315 in the submucosal plexus of Dlx1/2;Vip-IRES-Cre;R26R-TdTomato mice ( Figure 8F-I). 316 As before, we saw no difference in total neuron density ( Figure 8E, J). We also 317 confirmed that most TdTomato+ neurons appeared to express VIP protein at P0 318 To determine if the relative density of VIP+ neurites projecting into circular 328 smooth muscle (CSM) is altered in Dlx1/2-/-mice, we quantified neurite density by 329 drawing a line perpendicular to the axis of CSM neurite projections and counting thin 330 fibers crossing that line. Total density of CSM TuJ1+ neurites was significantly reduced 331 in Dlx1/2-/-mice (22.3% reduction, p=0.0077; Figure 9K-Q). However, the ratio of 332 TdTomato+ neurites to total was not reduced ( Figure 9R). TdTomato+ neurite density did 333 not reach our statistically threshold, but mean neuron density was also reduced 334 (p=0.1006; Figure 9S). To determine if the altered neurite density was consistent across 335 strain backgrounds, we quantified density of proximal SI CSM-projecting TuJ1+ neurites 336 in Dlx1/2-/-mice on a CD1 background, but we found no difference in neurite density 337 between mutants and controls (p=0.3193; Figure 9T). These data suggest Dlx1/2 may 338 affect targeting or growth of some neurites in CSM. 339 To further investigate how Dlx1 and Dlx2 affect neurite growth, we cultured 340 ENCDCs from unselected E12.5 bowel in the presence of glial cell-line derived 341 neurotrophic factor (GDNF) and labeled neurites using TuJ1. For this experiment, we 342 used Dlx1/2;ChAT-EGFP-L10a mice on a mixed CD1xC57BL/6 background, because 343 we initially hoped to use EGFP to distinguish ChAT+ and ChAT-neurons in culture; 344 however, EGFP was not expressed at this young age. Cultured neurons exhibited highly 345 variable morphology (Supplementary Figure 6A To determine if VIP is reduced in myenteric nerve fibers, we immunostained P0 353 mouse SI with antibodies against TuJ1, HuC/D, and VIP and imaged at high power. 354 Using CellProfiler image analysis software, we created a mask defined by TuJ1, but not 355 HuC/D (i.e. myenteric plexus nerve fibers, with cell bodies excluded) and computed 356 average VIP intensity within the masked region ( Figure  disease where the ENS is missing from distal bowel (33) and neuropathic CIPO where 371 ENS is present, but dysfunctional (1). Human CIPO causes abdominal distension and 372 bowel dysmotility that resembles bowel problems in Dlx1/2-/-mice, but anatomic defects 373 are often difficult to identify in CIPO. While it is possible that more careful analysis of 374 ENS structure would identify defects in human CIPO bowel, functional defects can also 375 occur in the absence of anatomic abnormalities. For Dlx1/2-/-mice, we did, eventually 376 uncover anatomic defects, but only after RNA-seq identified reduced Vip levels in mutant 377 ENS. RNA-seq is possible in human ENS, and might also be valuable in analysis of 378 bowel from people with CIPO (34,35). 379 We chose to investigate bowel motility in Dlx1 and Dlx1/2 mutants because Dlx1 380 and Dlx2 are expressed in developing ENS at ages relevant for ENCDC migration, 381 differentiation, and subtype specification, and because the original description of Dlx2-/-382 mice suggested they die from bowel dysmotility (10,(17)(18)(19)(20). We showed that neonatal 383 Dlx1/2-/-mice have profound bowel function defects, including delayed gastric emptying 384 and slow small bowel transit via an in vivo FITC-dextran assay. FITC-dextran was found 385 in the stomachs of all Dlx1/2-/-mice tested, confirming that despite cleft palate, Dlx1/2-/-386 mutants were capable of swallowing and that stomach empties slowly. When we 387 maintained bowel in an oxygenated organ bath and treated with TTX, which blocks nerve 388 cell activity, we found Dlx1/2-/-bowels had severely reduced TTX-sensitive contraction 389 patterns. Interpreting these findings is challenging because a complex interplay of 390 excitatory and inhibitory neural input regulates bowel smooth muscle contraction and 391 relaxation. A naïve interpretation is that neuron-mediated smooth muscle excitation 392 (contraction) is defective in Dlx1/2-/-mice. Another possibility is that Dlx1/2-/-bowel 393 has reduced smooth muscle inhibition at baseline, leading to a mildly contracted 394 phenotype approaching that of control bowel treated with TTX. Consistent with this 395 hypothesis, Dlx1/2-/-bowel did not seem to contract significantly upon TTX treatment, 396 unlike control bowel. Moreover, bowel videos suggest that neurally-mediated 397 "contraction complexes" actually consist of complex motor patterns, where bowel first 398 straightens (possibly reflecting longitudinal smooth muscle relaxation), then contracts 399 rapidly, and finally returns to baseline tortuosity (Supplementary Video 1). TTX-treated 400 bowel, which is highly contracted, never undergoes this pattern, and Dlx1/2-/-mouse 401 bowel does so only rarely. Since Dlx1/2-/-mice had a normal pattern of TTX-insensitive 402 (i.e. non-neuronal) contractions, and Dlx1 and Dlx2 are prominently expressed in 403 developing ENS, we interpret these findings to mean that Dlx1/2 mutations cause 404 neuropathic CIPO-like disease in mice. 405 Surprisingly, immunohistochemical studies showed Dlx1/2-/-bowel had largely 406 normal ENS structure at birth, with normal neuron and glia density and normal ratios of 407 many enteric neuron subtypes, even though changes in these parameters have been 408 identified in many mouse lines (3, 36-42). We note, however, that our extensive anatomic 409 studies were not exhaustive, that some antibodies gave weak signals in P0 ENS (e.g., 410 tyrosine hydroxylase, substance P), and that we did not perform enough replicates to 411 define sex-restricted defects. Furthermore, although Dlx2-/-mice had a small ENCDC 412 migration delay at E12.5, this may occur because of different strain backgrounds for 413 Dlx2-/-(C57BL/6) and Dlx1/2-/-mice (CD1). Strain effects on ENCDC migration have 414 been reported in the ENS (29). It is, however, unlikely that this minimal transient 415 ENCDC migration delay impairs bowel function at P0. 416 To identify defects causing CIPO-like disease in Dlx mutant mice, we performed 417 RNA-seq. Unexpectedly, we identified a novel connection between Dlx genes and Vip. 418 The Dlx1/2-/-ENS had decreased Vip mRNA at P0 (53% reduction) and a greater 419 reduction in Vip mRNA at E14.5 (64% reduction by qPCR, 84% reduction by RNA-seq). 420 Consistent with RNA data, the intensity of VIP immunostaining in cell bodies and in 421 neurites was significantly reduced in Dlx1/2-/-myenteric plexus. Moreover, Dlx1/2-/-422 ;VIP-IRES-Cre;TdTomato reporter mice had significantly fewer TdTomato+ neurons in 423 myenteric (18.4%) and submucosal plexus (51.5%), even though we could not appreciate 424 neuron loss via VIP immunohistochemistry. We attribute this difference to limited VIP 425 protein in nerve cell bodies, which may make neuron counting less reliable with antibody 426 staining than using Vip-lineage reporter mice. 427 In the ENS, VIP acts as a critical ENS neurotransmitter with multiple roles. In 428 myenteric plexus, VIP is co-expressed with nitric oxide synthase and PACAP in We suspect this is because we quantified both brightly-(~11%) and dimly-labeled 467 (~42%) populations of GABA+ neurons ( Figure 6X), whereas prior studies may have 468 only quantified brightly stained neurons). We also note that loss of Dlx1 causes apoptosis 469 in subsets of CNS interneurons in 1-2-month-old mice (15) Table 2) at room 559 temperature with gentle rocking (1 hour). After 3 additional 5-minute washes in PBS, 560 samples were mounted serosal side up in 50% glycerol/50% PBS on glass slides. 561 NADPH diaphorase staining was performed as described (57). Briefly, NADPH (Sigma-562 Aldrich) and Nitro Blue Tetrazolium were dissolved in PBS with 0.2% Triton-X-100. 563 Samples were incubated in this solution (7-15 minutes, 37°C), rinsed in PBS, then stained 564 with antibodies as described above. For tissue labeled with ChAT antibody, samples were 565 fixed in 100% methanol on ice (30 min) and placed in Dent's bleach (1:4:1 100% DMSO: 566 100% Methanol: 30% H 2 O 2 ) for 1 hr at RT prior to blocking. 567 568

Immunofluorescent staining of cultured cells 569
Slides were rinsed in PBS, blocked (1 hour in PBS + 0.1% or 0.5% Triton X-100 (PBST) 570 with 5% NDS), and incubated with primary antibody (Supplementary Table 2 Enteric neuron subtype proportions were quantified by counting the number of HuC/D+, 587 GFP+,nNOS+,calretinin+,GABA+,VIP+,or somatostatin+ cell. 588 Neurite fiber density was quantified drawing a line perpendicular to circular smooth 589 muscle and counting the number of neurites that crossed the line. Observers were blinded 590 to genotype for all quantitative analyses. 591 For measuring VIP intensity, we used 5-10 randomly selected 63x confocal z-592 stacks per sample. Only slices capable of being accurately thresholded were analyzed. 593 Using CellProfiler (58) Pregnant mouse mothers (E18.5) were injected subcutaneously with 2 mg of progesterone 609 (Sigma-Aldrich, #P3972-5G). The next morning (E19.5), mice were euthanized via 610 cervical dislocation. Pups were removed from the mother as previously described (59)  E14.5 dams were euthanized with CO 2 . P0 pups were delivered by Cesarean section as 628 described above. Pups were rapidly removed from the mother. E14.

Analysis of bowel motility patterns 666
Videos were converted to .wmv format using Movie Maker and saved at 1x and 16x 667 speeds. In-house MatLab (MathWorks, Natick, MA) scripts 668 (https://github.com/christinawright100/BowelSegmentation) were used to threshold the 669 movies, generate kymographs, and perform Fourier transform analysis. Peak frequency 670 was determined by calculating the Fourier transform at each vertical slice of the 671 kymograph, averaging these, and plotting the averages to determine the peak frequency. 672 Investigators blinded to genotype and condition quantified neurally-mediated 673 contractions in kymographs and recorded if contractions were rhythmic or non-rhythmic. 674 To quantify bowel width and tortuosity, 10 random frames were saved from each video 675 and analyzed in ImageJ. For each frame, we measured 5 random widths (randomly 676 generated using a MatLab script) along the proximal bowel (50 widths per bowel). For 677 tortuosity, total bowel length was divided by straight-line distance. 678

679
Quantitative PCR (qPCR) Analysis 680 Quantitative real time-polymerase chain reaction (qRT-PCR) was performed using 681 previously described primers which we validated (Supplementary Table 3 We used Prism 7.03 (GraphPad Software, San Diego, CA) and SigmaPlot 11.0 (Systat 699 Software, Chicago, IL) for statistical analysis. A two-sided Student's t-test or Mann-700 Whitney Rank Sum Test (MWRST) was used when comparing two groups. When 701 comparing multiple groups, a one-way ANOVA with post hoc multiple comparisons tests 702 (Tukey) was used unless assumptions were not met, in which case we used a Kruskal-703 Wallis test with Dunn's post hoc multiple comparisons test. When comparing weight 704 across multiple time points and groups, a two-way repeated measures ANOVA was used 705 with post hoc multiple comparisons tests (Tukey). A cutoff of p < 0.05 was considered 706 significant. For RNA-seq analyses, a false discovery rate-adjusted p-value (i.e. q-value) 707 of 0.10 was considered significant. Data represent mean ± standard error of the mean 708 (SEM). Investigators were blinded to condition for all quantifications. 709 710

Study approval 711
All animal experiments were approved by the Children's Hospital of Philadelphia 712 IACUC. 713 714