Asymmetric Hapln1a drives regionalised cardiac ECM expansion and promotes heart morphogenesis during zebrafish development

The mature vertebrate heart develops from a simple linear cardiac tube during early development through a series of highly asymmetric morphogenetic processes including cardiac looping and chamber ballooning. While the directionality of heart morphogenesis is partly controlled by embryonic laterality signals, previous studies have suggested that these extrinsic laterality cues interact with tissue-intrinsic signals in the heart to ensure robust asymmetric cardiac morphogenesis. Using live in vivo imaging of zebrafish embryos we describe a left-sided, chamber-specific expansion of the extracellular matrix (ECM) between the myocardium and endocardium at early stages of heart morphogenesis. We use Tomo-seq, a spatial transcriptomic approach, to identify transient and regionalised expression of hyaluronan and proteoglycan link protein 1a (hapln1a), encoding an ECM cross-linking protein, in the heart tube prior to cardiac looping overlapping with regionalised ECM expansion. Loss- and gain-of-function experiments demonstrate that regionalised Hapln1a promotes heart morphogenesis through regional modulation of ECM thickness in the heart tube. Finally, we show that while induction of asymmetric hapln1a expression is independent of embryonic left-right asymmetry, these laterality cues are required to orient the hapln1a-expressing cells asymmetrically along the left-right axis of the heart tube. Together, we propose a model whereby laterality cues position hapln1a expression on the left of the heart tube, and this asymmetric Hapln1a deposition drives ECM asymmetry and subsequently promotes robust asymmetric cardiac morphogenesis.


Introduction 37
Congenital heart defects are the most common human birth abnormality, with an incidence of 38 approximately 1% of live births (van der Linde et al. 2011 ;Hoffman & Kaplan 2002). These 39 structural malformations arise due to abnormal morphogenesis and maturation of the heart 40 during embryonic development. A key stage in cardiac development is when the heart 41 transitions from a linear tube to an asymmetric organ, a process including initial looping 42 morphogenesis of the tube and subsequent ballooning of the cardiac chambers. Correct cardiac 43 morphogenesis is vital for ensuring normal blood flow through the heart, proper chamber and 44 vessel alignment, valve formation and septation. Therefore early cardiac morphogenesis is a 45 tightly controlled process, and requires coordination of heart-extrinsic signalling cues, cardiac 46 growth and tissue-intrinsic changes in cell shape, mediated by cytoskeletal rearrangements 47 (Desgrange et al. 2018). 48 The requirement for embryonic left-right signalling pathways in promoting directionality of 49 heart morphogenesis is well established, with asymmetric Nodal signalling playing a key role 50 7 regionalised synthesis of these proteins does not cause ECM asymmetry. We therefore 151 hypothesised that a protein required either for HA modification or cross-linking may be 152 regionally expressed in the heart tube prior to looping morphogenesis and promote regionalised 153 ECM expansion. 154 To identify candidate genes which modulate cardiac ECM expansion, we took a genome-wide 155 approach to identify genes expressed in the heart tube at 26hpf, prior to the onset of looping 156 morphogenesis. Since we observed the strongest left-sided ECM expansion in the putative 157 atrium, as well as a generally more expanded ECM at the venous pole of the heart compared 158 to the arterial pole, we used the previously described Tomo-seq technique to generate a 159 we defined a subset of tissue sections with atrial identity. We subsequently filtered genes that 165 are up-regulated in atrial sections compared to ventricular sections in both hearts and examined 166 this list for genes which may be implicated in ECM modification. 167 Using this approach we identified hyaluronan and proteoglycan link protein 1a (hapln1a, 168 formerly crtl1) as a candidate that might drive regionalised ECM expansion (Fig 2B). The 169 Hapln family of proteins are secreted into the ECM where they crosslink HA to proteoglycans 170 (Spicer et al. 2003), suggesting Hapln1a may act to modify the cardiac ECM environment. 171 Furthermore, Hapln1 mutant mice exhibit heart malformations including septal defects and 172 perturbations of the inflow and outflow tracts, consistent with abnormal heart morphogenesis 173 (Wirrig et al. 2007). 174 8 mRNA in situ hybridisation analysis revealed that hapln1a is expressed in the posterior of the 175 heart disc and the cardiac cone (Fig 2C, D). At 26hpf hapln1a expression is upregulated on the 176 left side of the cardiac tube with elevated levels of expression of hapln1a in the future atrium 177 compared to the future ventricle, recapitulating the regionalised ECM expansion we observe 178 in the heart (compare Fig 2E and Fig 1K). This is in line with recent studies demonstrating that 179 the posterior compartment of the cardiac disc is re-positioned to the left side of the heart tube 180 (Guerra et al. 2018). By 50hpf hapln1a expression is restricted to very low levels in the 181 atrioventricular canal, the precursor to the atrioventricular valve ( Fig 2F). Fluorescent in situ 182 hybridization reveals hapln1a is expressed in the myocardium (Fig 2G-I), while analysis of 183 Hapln1a protein localisation confirms it is deposited in the ECM (Fig 2J-L). Despite the 184 absence of hapln1a expression in the heart at 50hpf (Fig 2F), Hapln1a protein is maintained in 185 the ECM at 50hpf (Fig 2M-Q), suggesting that the ECM environment established during early 186 stages prior to heart tube formation is maintained during heart development and may be 187 important for continued cardiac morphogenesis. 188

189
Hapln1a is required for heart morphogenesis and promotes ECM expansion. 190 To determine whether Hapln1a is required for cardiac morphogenesis we used CRISPR-Cas9-191 mediated genome editing to generate hapln1a mutants. We injected a pair of guide RNAs 192 targeting approximately 200bp upstream of the translation-initiating ATG and immediately 193 downstream of the ATG, allowing us to excise the putative promoter of hapln1a ( Fig S4). We 194 recovered F1 adult fish carrying two deletions; a 187bp deletion (hapln1a D187 ) and a 241bp 195 deletion (hapln1a D241 ), both of which remove the initiating ATG and upstream sequence, and 196 established both as stable lines at F2. To confirm the deletions removed the hapln1a promoter 197 and prevented transcription, hapln1a expression was analysed at 26hpf in F3 mutant embryos 198 for each allele. Homozygous hapln1a promoter mutants of either allele exhibit a complete loss 9 of hapln1a expression at 26hpf compared to wild type embryos (Fig 3C, E, Fig S4), 200 demonstrating successful deletion of the hapln1a promoter, and confirming the promoter 201 mutants as loss of function models. Furthermore, embryos heterozygous for a hapln1a 202 promoter mutation also exhibit a reduction in levels of transcript compared to wild type siblings 203 ( Fig 3D, Fig S4). Analysis of heart development in hapln1a D241 mutants at 50hpf did not reveal 204 striking abnormalities in cardiac morphogenesis (Fig 3F-H), however we did observe 205 occasionally mispositioned and malformed atria. To investigate this further, we examined 206 morphology of individual chambers at 50hpf by in situ hybridization analysis of the ventricular 207 marker myh7l (myosin heavy chain 7-like) and the atrial marker myh6 (Fig 3I-N). 208 Quantification of either whole heart size, or individual chamber size revealed a significant 209 reduction in whole heart size in hapln1a D241 mutants when compared to wild type siblings (Fig  210   3R), and a mild reduction in atrium size ( Fig 3S). We observed a similar significant reduction 211 in atrial size in the second hapln1a D187 allele ( Fig S4). To quantify cardiac morphology at 50hpf 212 we defined the looping ratio, a quotient of the looped and linear distances (Fig S4 and  213 Methods). Although homozygous hapln1a mutants exhibit a reduction in cardiac size, we 214 observed no significant reduction in looping ratio ( Fig 3U, Fig S4). 215 To assess the impact of loss of hapln1a on continued morphogenesis of the heart we used live 216 light sheet imaging of Tg(myl7:lifeActGFP); Tg(fli1a:AC-TagRFP) transgenic embryos, 217 acquiring 3D images of the heart at 72hpf. We found that hapln1a D241 mutant hearts appear 218 dysmorphic at 72hpf, with abnormally positioned atria and disrupted heart looping compared 219 to wild type (Fig 3O-Q). Together this demonstrates that similar to mouse, hapln1a is required 220 for cardiac morphogenesis. 221 Since Hapln1 functions as an ECM binding protein and its localisation recapitulates the 222 regionalised ECM expansion in the heart tube prior to heart morphogenesis, we hypothesized 223 that Hapln1a promotes cardiac morphogenesis by driving regionalised ECM expansion in the 224 heart. Since both hapln1a promoter deletion alleles carry the Tg(myl7:lifeActGFP) transgene, 225 this prevented analysis of ECM width throughout the heart tube of hapln1a mutants using the 226 ssNcan-GFP HA sensor. Therefore, to examine ECM asymmetry in the heart tube upon loss 227 of hapln1a, we injected a morpholino (MO) against hapln1a into zebrafish embryos at the 1- Hapln1a is a member of a family of ECM binding proteins which crosslink hyaluronan with 245 proteoglycans in the ECM (Spicer et al. 2003). Since hapln1a is transiently expressed at cardiac 246 disc and early tube stage, this suggests that the cardiac ECM that drives continued 247 morphogenesis of the heart is established at early stages of heart development and requires the 248 interaction of Hapln1a with hyaluronic acid. To interrogate the temporal requirements for HA 249 in heart looping, we applied the HA synthesis inhibitor 4-Methylumbelliferone (4-MU 250 (Nakamura et al. 1997;Ouyang et al. 2017)) to embryos prior to the onset of heart tube 251 formation at 18hpf, and either washed the drug off at 22hpf, or left the embryos to develop to 252 50hpf, when we assessed heart looping morphology. Inhibiting HA synthesis from cardiac disc 253 stage (18hpf) until 50hpf often arrested heart development mid-way during tube formation (Fig  254   S6), a more profound phenotype than that observed in has2 zebrafish morphants or Has2 mouse 255 mutants (Smith et al. 2008;Camenisch et al. 2000). However, inhibition of HA synthesis during 256 the short time window between cardiac disc (18hpf) and cardiac cone (22hpf) stage, prior to 257 tube formation, resulted in normal tube formation but a specific disruption to heart looping 258 morphogenesis ( Fig S6). This supports the hypothesis that HA synthesised prior to formation 259 of the heart tube is required for looping morphogenesis of the heart. 260 Having demonstrated a requirement for HA synthesis for heart morphogenesis during early 261 cardiac development when hapln1a expression is initiated, we wanted to confirm the 262 interaction of hapln1a and HA in heart looping morphogenesis. Injection of sub-phenotypic 263 doses of morpholinos targeting either has2 or hapln1a did not result in significant defects in 264 cardiac morphology at 48hpf (Fig S6). However, co-injection of both has2 and hapln1a 265 morpholinos results in profound defects in heart morphology at 48hpf (Fig S6), including a 266 reduction in heart looping ratio, and abnormal atrial morphology. This more profound 267 phenotype than that observed by either injection of hapln1a MO + tp53 MO, has2 MO + tp53 268 MO, 4MU treatment or deletion of the hapln1a promoter, suggesting that while timely HA 269 signaling drives heart morphogenesis subsequent to tube formation, hapln1a is an important 270 regional modulator of this process. 271

272
While analysis of hapln1a mutants demonstrates a requirement for Hapln1a in heart 273 development (Fig 3), we wished to investigate whether the regionalization of hapln1a 12 expression is important for cardiac morphogenesis. We generated a DNA construct in which 275 the full length hapln1a coding sequence is driven by the pan-myocardial myl7 (myosin light 276 chain 7) promoter, flanked by Tol2 transposon sites to allow integration into the genome (Fig  277   5A). We co-injected myl7:hapln1a DNA with tol2 transposase mRNA at the 1-cell stage and 278 analysed both myl7 and hapln1a expression at 50hpf, allowing us to visualize heart morphology 279 alongside assessing the extent of hapln1a misexpression (Fig 5B,C). We quantified the looping 280 ratio (Fig S4), as well as the percentage coverage of hapln1a expression in the whole heart and 281 plotted percentage coverage against looping ratio ( Fig 5D). We found that increasing the 282 domain of hapln1a expression in the heart results in a reduction in looping morphogenesis (Fig  283   5D), suggesting that regionalised expression of hapln1a in the heart is important for proper 284 cardiac morphogenesis. Since hapln1a is expressed at higher levels in the atrium than the 285 ventricle and ECM asymmetry is more robust in the atrium, we hypothesized that hapln1a 286 misexpression in each chamber may impact differently on heart morphogenesis. We quantified 287 hapln1a misexpression in each chamber by calculating the percentage of the chamber which 288 expresses hapln1a (Fig 5E-I), and found that while misexpression of hapln1a in the ventricle 289 did not impact upon heart morphogenesis ( Fig 5J), misexpression of hapln1a in the atrium 290 resulted in abnormal cardiac morphogenesis ( Fig 5K). This suggests that spatially-restricted 291 hapln1a expression in the atrium drives cardiac morphogenesis. 292 293 Finally, since Hapln1a is asymmetrically expressed on the left side of the heart tube, and is 294 required for heart morphogenesis, we hypothesized that it may contribute to a previously-295 described tissue-intrinsic mechanism of heart looping morphogenesis (Noël et al. 2013 in left-right asymmetry, including a disruption to normal leftward displacement of the heart 301 tube (Schottenfeld et al. 2007). We hypothesised that induction of hapln1a expression occurs 302 independent of embryonic laterality cues, but that asymmetric positioning of hapln1a-303 expressing cells in the heart tube may be tightly linked to the direction of heart tube position, 304 and therefore dictated by embryonic left-right asymmetry. We analysed hapln1a expression 305 in an incross of pkd2 hu2173 heterozygotes and found that consistent with our hypothesis hapln1a 306 is always expressed in the posterior of the heart disc in pkd2 hu2173 mutants at 22hpf (Fig 6D), 307 demonstrating that initiation of hapln1a expression is laterality independent. Importantly, at 308 26hpf we observed that positioning of hapln1a-expressing cells is dependent upon cardiac 309 position -in embryos where the heart is positioned to the right, hapln1a is upregulated on the 310 right side of the tube, whereas if the heart remains midline, hapln1a does not exhibit a clear 311 left-right asymmetry in up-regulation ( Fig 6A-C, E). These data support a model where 312 laterality cues do not initiate hapln1a expression but are required for its subsequent position in 313 the heart tube. To further investigate our model we analysed Hapln1a protein localisation in 314 spaw mutant embryos, which lack asymmetric Nodal expression prior to asymmetric organ 315 morphogenesis (Noël et al. 2013). spaw mutant embryos exhibit midline hearts at 26hpf and 316 Hapln1a is no longer positioned on the left side of the heart tube as observed in sibling embryos, 317 but instead is secreted into the cardiac ECM on the ventral face of the heart ( Figure 6F-M). 318 Together, we propose a model where initiation of hapln1a expression in the posterior cardiac 319 disc is independent of KV-based laterality cues, but the subsequent cell movements which 320 occur during heart tube formation reposition this population of cells to the left side of the heart, 321 dictating the axis of ECM asymmetry in the heart tube (Guerra et al. 2018). Therefore, 322 embryonic laterality positions the regionally specialized ECM in the heart tube, ensuring that 323 directionality and growth of the heart are tightly coordinated to fine tune cardiac 324 morphogenesis. ( Figure 6N). 325 326

Discussion 327
Our data show that prior to heart looping morphogenesis in zebrafish, the heart tube exhibits 328 regionalised ECM expansion that is dependent upon localised expression of the ECM binding of meis2b mutants revealed defects in atrial morphology at juvenile and adult stages, supporting 344 our conclusion that early anterior-posterior asymmetry in the heart disc/left-right asymmetry 345 in the heart tube are important for continual cardiac morphogenesis. However, contrary to our 346 study which reveals a reduced atrial size in hapln1a mutants, meis2b mutant adult zebrafish 347 exhibit an enlarged atrium (Guerra et al. 2018) suggesting that while they are expressed in the same domain, these two genes regulate atrial morphogenesis differently. hapln1a mutants are 349 adult viable (data not shown), therefore it would be interesting to determine whether the atrium 350 remains underdeveloped in hapln1a mutants, or whether they also develop a hyperproliferative 351 atrial hypertrophy phenotype by adulthood. 352 A major role of the ECM in tissue morphogenesis is to provide structural or biomechanical 353 cues to neighbouring tissues. While hapln1a is expressed prior to tube formation and during 354 very early stages of looping morphogenesis only, Hapln1a protein persists in the cardiac jelly 355 even after the heart has undergone initial looping morphogenesis (Fig 2). Together with our 356 data demonstrating that HA synthesis is required prior to heart tube formation to promote 357 cardiac morphogenesis (Fig S6), this suggests that the ECM environment generated early 358 during heart development is crucial for continual and/or maintenance of chamber 359 morphogenesis. Interestingly, recent studies have demonstrated that Hapln1 is the key element 360 required for tissue folding in the human neocortex, and that HA is required to maintain the 361 architecture of the tissue after folding has occurred (Long et al. 2018). In light of this, we 362 propose that formation of the specific ECM environment at cardiac disc stage is required to 363 ensure the heart maintains correct shape as it undergoes early looping morphogenesis. 364 Alternatively, Hapln1a-mediated cross-linking may modulate regional stiffness of the cardiac 365 ECM. Differential matrix stiffness has been shown to regulate a wide variety of cellular observe that in pkd2 mutants where asymmetry of the heart is reversed and the heart tube 406 extends to the right, due to the cellular movements required for tube formation hapln1a-407 expressing cells are positioned on the right side of the heart tube instead of the left (Fig. 6). We 408 propose a model in which while antero-posterior patterning of the heart disc is laterality-409 independent, since laterality signals promote cardiac disc rotation and heart tube displacement, 410 this results in hapln1a asymmetry along the left-right axis of the heart tube. This generates 411 lateral ECM asymmetry in the heart, promoting robust looping and chamber ballooning 412 morphogenesis ( Figure 6N). This would begin to explain previous observations that heart tube 413 position prior to looping predicts the direction of looping morphogenesis (Baker et al. 2008;414 Chen et al. 1997), since the direction of heart tube extension will dictate lateralised ECM 415 asymmetry in the tube. In addition, we show that in embryos lacking Nodal signalling, Hapln1a 416 is positioned on the ventral face of the heart (Fig 6). Together with our previous observations 417 that the heart disc of spaw mutants undergo a very mild level of rotation and that the direction 418 of this limited rotation is consistent with the final outcome of looping direction (Noël et al. 419 2013), it is possible that even slight rotation in the heart disc is sufficient to set up mild 420 asymmetry in the ECM that can go on to promote directional looping, supporting the 421 hypothesis that these pathways may act together to robustly ensure directional looping 422

morphogenesis. 423
Together this study elaborates upon our previously proposed model where laterality-based 424 extrinsic cues feed into a tissue-intrinsic mechanism of heart looping to promote robust 425 directional cardiac morphogenesis.   regression of the data (n=194). Spearman's correlation coefficient (r) deviates significantly 541 from zero demonstrating that increased coverage of hapln1a in the myocardium results in 542 reduced heart looping morphogenesis. E-I: Illustration of quantification approach to analyse 543 pan-cardiac or chamber-specific hapln1a overexpression at 55hpf. Embryos are stained for 544 myl7 and hapln1a mRNA (E), and images are split into red, blue and green channels. The green 545 channel (F) highlights the myl7 expression in the heart and is used to manually draw round the 546 ventricle and atrium. The red channel highlights the hapln1a expression (G), and is first 547 processed to subtract the background, before the chamber ROIs are applied (G). Each chamber 548 ROI is isolated, the surrounding image cleared, and a threshold is applied to the hapln1a 549 staining within the ROI (H-I). The number of pixels within the chamber ROI is then measured, 550 alongside the total area of the ROI, quantifying percentage of the chamber expressing hapln1a. Live zebrafish embryos were imaged on a ZEISS Lightsheet Z.1 microscope at 72hpf. To 711 assess cardiac morphology at 72hpf embryos were anesthetised with tricaine before mounting 712 in 1% low melting point agarose in E3 with 8.4% tricaine, using black capillaries. To stop the 713 heart the imaging chamber was filled with E3 plus tricaine (8.4%) and the temperature 714 maintained at 10°C. All samples were imaged using a 20X lens and 1.0 zoom. Dual side lasers 715 with dual side fusion and pivot scan were used for sample illumination. Image stacks were 716 initially processed using Vision4D (Arivis AG, Germany) and Fiji. Processing steps included 717 noise removal, background correction, and subsequent application of individual morphological 718 filters to each channel to sharpen the edges of the myocardial and endocardial tissue layers. 719 Maximum intensity z-projections of the composite channels were used to visualise cardiac 720 morphology the centreline of the heart was manually traced. Optical transverse sections 721 perpendicular to the centreline were generated at regularly-spaced intervals originating from 722 the venous pole into the atrium to visualise the cardiac ECM. 723 Embryos injected with ssNcan-GFP mRNA were fixed overnight in 4% PFA, and 724 immunohistochemistry was carried out to amplify the GFP signal. Embryos were dissected 725 and imaged using a Zeiss Airyscan microscope, and z stacks were obtained with a z-step size 726 of 1µM. Images were Airyscan processed using Zen Black software (Zeiss), and the resulting 727 image stacks were optically resliced using Fiji. ECM width was manually measured in Fiji. 728 ECM measurements were aligned between samples at the venous pole of the heart for plotting. 729 Looping ratio was calculated from images of myl7 expression detected by ISH. All samples 730 from one experimental set were blinded using the ImageJ Blind_Analysis plugin 731 (https://github.com/quantixed/imagej-macros/blob/master/Blind_Analysis.ijm). The linear 732 distance from arterial to venous poles of the heart was measured as a straight-line distance, and 733 looped distance was drawn from the same positions at each pole through the centre of each 734 chamber, down the centreline of the looped heart. Looping ratio was determined by dividing 735 looped distance by linear distance. Statistical testing of average looping ratio between 736 experimental conditions was carried out using Kruskal-Wallis with Dunn's multiple 737

comparisons. 738
Heart, ventricle or atrium area at 50hpf was quantified from in situ hybridisations by manually 739 drawing round either myl7, myh7l or myh6 staining area in Fiji. Statistical testing of heart or 740 chamber size between genotypes was carried out using Kruskal-Wallis with Dunn's multiple 741

comparisons. 742
Quantification of hapln1a overexpression was performed by imaging overexpression embryos 743 where myl7 expression was detected using INT/BCIP and hapln1a expression detected using 744 NBT/BCIP. All embryos were imaged using the same microscope settings, and individual 745 images combined into a composite of all experiments. Using the composite, channels were 746 split in Fiji, resulting in the blue and green channels carrying the myl7 stain, and the red channel 747 carrying the hapln1a stain. Background was subtracted in the red channel. The myl7 staining 748 was used to calculate looping ratio for each heart. In addition, each myl7 signal was used to 749 manually trace the whole heart, atrium, AVC and ventricle for each heart, which was saved as 750 a region of interest (ROI) and the area of each chamber was measured in pixels. Next, the 751 hapln1a image was thresholded to generate a binary image. The whole-heart, or chamber-752 specific ROI was applied to the thresholded hapln1a channel, and the number of positive pixels 753 in each ROI recorded. Number of positive pixels are as a % of total number of pixels 754 comprising the heart or the specific chamber could then be calculated and plotted against 755 looping ratio for each heart. Spearmann's correlation coefficient (r) was calculated in 756 GraphPad Prism, with 95% confidence intervals.