Vascular dimorphism ensured by regulated proteoglycan dynamics favors rapid umbilical artery closure at birth

The umbilical artery lumen closes rapidly at birth, preventing neonatal blood loss, whereas the umbilical vein remains patent longer. Here, analysis of umbilical cords from humans and other mammals identified differential arterial-venous proteoglycan dynamics as a determinant of these contrasting vascular responses. The umbilical artery, but not the vein, has an inner layer enriched in the hydrated proteoglycan aggrecan, external to which lie contraction-primed smooth muscle cells (SMC). At birth, SMC contraction drives inner layer buckling and centripetal displacement to occlude the arterial lumen, a mechanism revealed by biomechanical observations and confirmed by computational analyses. This vascular dimorphism arises from spatially regulated proteoglycan expression and breakdown. Mice lacking aggrecan or the metalloprotease ADAMTS1, which degrades proteoglycans, demonstrate their opposing roles in umbilical vascular dimorphism, including effects on SMC differentiation. Umbilical vessel dimorphism is conserved in mammals, suggesting that differential proteoglycan dynamics and inner layer buckling were positively selected during evolution.


Introduction
The umbilical cord, typically containing two arteries and one vein in humans, is a crucial fetal structure in placental mammals. Umbilical arteries carry fetal blood to the placental vascular bed, whereas the umbilical vein returns oxygenated blood to the fetus. Neonatal respiration at birth renders the maternal oxygen supply redundant. Umbilical arteries commence closure rapidly after delivery of the newborn whereas the veins remain open longer. The cord is routinely clamped following delivery and divided between the clamps in modern obstetric practice. Timing of cord clamping after birth, whether early or late, is extensively debated (Niermeyer, 2015;Tarnow-Mordi et al., 2017). A recent recommendation suggested clamping no earlier than 30-60 s after birth to facilitate the placental transfusion (TACoOaG, 2017). Although the necessity of clamping is rarely questioned, it appears to be a modern practice (Downey and Bewley, 2012). Cord clamping is rarely practiced in domesticated animals and certainly not in wild animals, yet all current mammalian species have survived evolutionarily. We hypothesized that intrinsic design characteristics of mammalian umbilical arteries prevent blood loss at birth without clamping.
Prior histological work revealed that umbilical arteries have a bilaminar structure (Meyer et al., 1978) but lack elastic lamellae, which endow large arteries with resilience during cyclic loading (Wagenseil and Mecham, 2009). However, the molecular mechanism underlying the bilayered structure and its relationship to arterial occlusion remains obscure. Here, we used a multi-disciplinary approach integrating a variety of morphologic approaches with mechanical testing, computational analysis and mouse mutants to demonstrate the molecular and biomechanical basis for rapid umbilical artery closure. The findings emphasize the dual importance of extracellular matrix proteoglycans in regulation of cell differentiation and conferment of desirable tissue mechanical characteristics.

Results
The umbilical artery has a bilaminar wall Three-dimensional imaging of term human umbilical cords, using synchrotron-based phase contrast micro-CT with effective pixel size 1.63 Â 1.63 mm 2 (Norvik et al., 2020) and histology, identified a much thicker tunica media (TM) in the umbilical artery than in the vein, with a visibly different structure ( Alcian blue, which binds sulfated glycosaminoglycans (GAGs), intensely stained the inner layer of the bilayered arterial TM but only the innermost three to four cell layers of the venous TM ( Figure 1c). SMCs in this GAG-rich region of the arteries were radially oriented and round, with nuclear-localized Sox9, a chondrogenic factor (Figure 1-figure supplement 1e; Ng et al., 1997).  (Figure 1c,d, f). Versican is a well-characterized vascular component (Wight and Merrilees, 2004), and aggrecan, which is known as a cartilage and neural proteoglycan (Lauing et al., 2014;Schwartz and Domowicz, 2014), is emerging as a significant CSPG in vascular disease (reviewed in Koch et al., 2020).

ADAMTS proteoglycanases are differentially expressed in the umbilical artery and vein
Aggrecan and versican are proteolytically cleaved by ADAMTS1, 4, 5, and 9 (Dancevic et al., 2016). ADAMTS1 and ADAMTS4 mRNAs had higher levels in the venous wall in microarrays (Figure 1e, Sup. Array data-1), and RNA-ISH demonstrated stronger expression of ADAMTS1, ADAMTS4, ADAMTS5, and ADAMTS9 in the veins (Figure 2a). ADAMTS1 was the most strongly expressed, localizing to venous endothelium and TM, with stronger umbilical artery expression seen in the outer than inner TM (Figure 2a). ADAMTS9 was similarly expressed as ADAMTS1, whereas ADAMTS4 and ADAMTS5 mRNAs were restricted to umbilical vein endothelium and some venous SMC (Figure 2a). Neo-epitope antibodies detecting ADAMTS-cleaved aggrecan and versican (anti-NITEGE and anti-DPEAAE, respectively) (Lark et al., 1995;Sandy et al., 1992;Sandy et al., 2001) showed strong staining throughout the venous TM and in the outer arterial TM, but not the inner arterial TM ( Figure 2b). Thus, proteoglycan accumulation in the inner TM of the umbilical artery may result from higher ACAN and VCAN expression and less proteolysis. In contrast, lower ACAN and VCAN expression and greater ADAMTS levels within the umbilical vein may preclude proteoglycan accumulation.

ADAMTS-mediated differential proteoglycan abundance in the umbilical artery and vein is evolutionarily conserved
We postulated that abundant hydrated proteoglycans in the inner arterial TM provided compressive stiffness that could not only prevent kinking and premature occlusion but could potentially facilitate rapid umbilical artery closure at birth. If so, similar adaptations should be present in other mammals. Analysis of umbilical cords from nine large primate and non-primate mammals disclosed similar dimorphism, namely, umbilical arteries were occluded and had thicker walls with similar infolding of the inner arterial TM ( Figure 3a) and strong Alcian blue and CS-staining, contrasting with veins ( Figure 3a,b). Anti-aggrecan and anti-NITEGE stained several animal species, confirming aggrecan abundance in the inner arterial TM and robust aggrecan cleavage resulting from ADAMTS protease activity in the outer TM of the artery and the TM of the vein (Figure 3c

Aggrecan and ADAMTS1 are necessary for normal umbilical cord morphogenesis
Mouse umbilical cords also demonstrated vascular dimorphism (Figure 4a), suggesting that genetically modified mice would provide mechanistic insights into proteoglycan dynamics and its impact. Aggrecan and versican immunofluorescence showed strong staining in the mouse umbilical artery inner TM and adventitia, with weaker staining in the veins (Figure 4b). Acan, Vcan and Adamts1,4,5,9 RNA-ISH at early (E12.5) and late (E18.5) gestational stages showed that Acan and Vcan were strongly expressed in the umbilical arteries ( Figure 4-figure supplement 1a). Adamts1 was the most highly expressed proteoglycanase in the mouse umbilical vein just prior to parturition (E18.5) (Figure 4-figure supplement 1a), evidenced by strong b-gal staining in venous TM, adventitia and endothelium; the inner umbilical artery TM and endothelium of Adamts1 lacZ/+ embryos lacked b-gal staining (Figure 4c). Although Adamts9 mutant embryos were previously observed to have short umbilical cords, abnormal umbilical artery development, and to die by 14.5 days of gestation (Nandadasa et al., 2015), umbilical cord development was not previously investigated in Figure 1 continued indicate minimum and maximum values. ***, p<0.001). (c) Alcian blue, eosin (pink) and nuclear fast red staining of umbilical vessel cross-sections shows a proteoglycan-rich (blue) inner TM in the umbilical artery but not the vein. Quantified staining intensity is shown on the right (n = 6 umbilical cords, whiskers indicate minimum and maximum values, ***, p<0.001). (d) Chondroitin sulfate (CS), heparan sulfate (HS), aggrecan and versican immunofluorescence (n = 4 cords for each antibody) showing that CS staining corresponds with aggrecan and versican staining and alcian blue in (c). (e) Volcano plots illustrating differential gene expression between human umbilical artery (red) and vein (green) (top, n = 4 umbilical arteries and veins) and differential gene expression between human umbilical artery inner TM (red) and the outer TM (green) (bottom, n = 2). (f) RNA in situ hybridization shows robust ACAN and VCAN expression (red signal) in the inner artery TM and weak expression in the vein (n = 3 umbilical cords for each in situ probe). * marks the vessel lumen. Brackets in c,d,f mark the TM. Wj, Wharton's jelly. Scale bars = 100 mm in c,d,f. The online version of this article includes the following video and figure supplement(s) for figure 1:    and ADAMTS9 expression in umbilical vein endothelium and tunica media (TM) and in outer arterial TM. Robust ADAMTS4 and ADAMTS5 expression was confined to the venous endothelium, with moderate ADAMTS4 expression and minimal ADAMTS5 expression in SMC (n = 3 umbilical cords for each probe). (b) ADAMTS-cleaved aggrecan (anti-NITEGE, red) and versican (anti-DPEAAE, red) both showed strong ADAMTS proteolytic activity Figure 2 continued on next page mutants of the two genes implicated here as potentially critical for umbilical cord vascular dimorphism, Acan and Adamts1.
The Adamts1 -/mutant is an insertion of an IRES lacZ-bearing cassette into intron 1 of the gene (Oller et al., 2017). This insertion reveals Adamts1 expression via staining for ß-galactosidase activity, and eliminated expression from the targeted allele but in its hemizygous state, the insertion led to reduction in both mRNA and protein (Oller et al., 2017). The Acan cmd-Bc allele is a spontaneous mutation found in a BALB/C colony (Bell et al., 1986) and resulted from deletion of exon 2 through exon 18 (Krueger et al., 1999). Acan -/embryos do not survive past birth (Krueger et al., 1999;Lauing et al., 2014) and few surviving Adamts1 -/mice were identified at the time of weaning (Oller et al., 2017). Acan mutants are thought to succumb to respiratory failure resulting from soft tracheal cartilages and ribs, whereas the cause of Adamts1 -/lethality is unknown. At E18.5, Acan -/and Adamts1 -/mutants each had significantly short umbilical cords ( Figure 4d) demonstrating their requirement for proper umbilical cord development. Umbilical cord histology showed thinner vascular walls in Acan -/umbilical vessels, and conversely, thicker vascular walls in Adamts1 -/umbilical vessels (Figure 4e-g). At earlier developmental stages (E12.5 to E14.5), lack of aggrecan did not affect either umbilical cord length or circumferential SMC reorientation (Figure 4-figure supplement 1bd), which occurs around E13.5 and is defective in Adamts9 mutants (Nandadasa et al., 2015). Furthermore, lack of aggrecan did not impair the survival of mouse embryos until parturition, since Acan -/embryos were observed at the expected Mendelian ratio at E18.5 ( Figure 4-figure supplement 1c). Thus, Acan and Adamts1 appear to be involved in umbilical vessel development from early gestation, but their functions manifest near parturition.

Contrasting SMC phenotypes in Acan and Adamts1-deficient umbilical cords
The arterial and venous lumina were smaller in Adamts1 -/mice relative to wild-type, indicating that their thicker vascular walls compromised luminal diameter, and larger in Acan -/mice (Figure 4e-g). Phospho-histone H3 staining revealed fewer proliferating cells in Acan -/umbilical cords at E18.5 ( Figure 4h). Immunostaining for SMC markers smooth muscle a-actin (SMA), smooth muscle myosin heavy chain (SMMHC), and phosphorylated myosin light chain (pMLC) showed weaker intensity in Acan -/umbilical arteries compared to wild-type (Figure 5a,b). In contrast, Adamts1 -/umbilical vessels showed stronger SMA, SMMHC and pMLC staining than wild-type littermates and apparent overgrowth of the arterial and venous walls ( Figure 5c). Intriguingly, endomucin, a venous endothelium-specific marker (dela Paz and D'Amore, 2009), also stained Adamts1 -/umbilical arterial endothelium ( Figure 5c) suggesting that ADAMTS1 may have a role in specifying artery/vein identity. Immunostaining of E17.5 Adamts1 -/umbilical cords indicated a crucial role for ADAMTS1 in regulating proteoglycan dynamics in the mouse umbilical cord. Specifically, we observed robust aggrecan and versican accumulation in the Adamts1 -/umbilical vein and in the outer TM of the Adamts1 -/umbilical artery (Figure 6a

Differential SMC contraction in the bilayered umbilical arteries
Despite uniform staining with SMC markers in human umbilical vascular SMC, co-staining with serine 20 -phosphorylated myosin light chain (pMLC) marking contractile SMCs (Dougherty et al., 2014) revealed that human umbilical arteries had more contractile SMCs than the vein, predominantly in the outer TM (Figure 7a,b). This suggests that outer umbilical artery SMCs are principally responsive to vasoconstriction stimuli at birth, whereas inner SMCs are relatively non-contractile. We hypothesized that an outer ring of contracting SMCs could drive the CSPG-rich inner arterial TM centripetally, occluding the lumen, and addressed this possibility initially using ex vivo biomechanical testing of late-gestation mouse umbilical vessels (Figure 7c, Figure 7-figure supplement 1). Mouse umbilical arteries had a smaller lumen, as expected at E18.5, and deformed less when loaded   imaging, which was not observed in the umbilical veins ( Figure 7c). Cross-sectional area measurements at fixed lengths revealed wall volume reductions during vasoconstriction (Figure 7-figure supplement 1e), less in the umbilical vein (~35%) than the umbilical artery (~50%), suggesting fluid exudation from the wall under forceful SMC contraction.

Computational modeling of arterial occlusion
These biomechanical tests of mouse umbilical cords, together with histological and immunostaining findings from human cords, motivated and informed a novel computational model of the umbilical artery incorporating its complex bilayered, multi-constituent structure (GAG-rich inner layer and contractile SMC-rich outer layer; Figure 7d) and multiaxial mechanical loading: axial extension, luminal pressurization, active contraction by SMCs, and intramural swelling of the inner layer that regulates tissue volume locally based on GAG content. Nonlinear regression of biaxial mechanical data from passive tests of the murine vessels identified best-fit values of the material parameters in the baseline constitutive model, while data from active contraction studies guided the selection of the associated active constitutive parameters ( Table 1). Model-based parametric studies examined combinations of different levels of GAG-driven swelling and SMC-generated active stress to identify their roles in umbilical artery closure at different levels of fixed luminal pressure. Increasing inner layer swelling in the absence of active outer layer stress narrowed the lumen at a fixed pressure, as expected given the constraining effect of the outer stiff passive matrix (Figure 7-figure supplement 2a). This trend reversed in the presence of active stresses, with increasing inner layer GAGs able to oppose vasoconstriction if overall wall volume remained constant (Figure 7-figure supplement 2b). Thus, increased inner layer swelling attenuates the ability of SMC contraction to prematurely reduce luminal radius, as revealed by varying the active stress parametrically for different fixed values of inner layer swelling. Importantly, the model predicted a sharp transition from a widely patent to a narrowed lumen due to small changes in active stress at lower values of swelling whereas radius changes were more gradual at higher values of swelling (Figure 7-figure supplement 2b). This transition, at which a decrease in volume of the inner layer associates with larger or smaller luminal radii for values of active stress below or above T act ffi 50 kPa (a key parameter of active stress generation) appears to be close to the in vivo value. Hence, consistent with ex vivo findings (

Folding of the arterial inner layer is necessary for vascular occlusion
Given the consistent histological finding of inner arterial TM infolding following birth, we modeled the biomechanics of superimposed inner layer buckling in the bilayered arterial model. Buckling can release energy stored in the inner layer during vasoconstriction-induced compression, thereby reducing the structural stiffness and resistance to SMC contraction. This analysis parametrically considered possible perturbations to the cylindrical geometry achieved at various levels of fixed luminal pressure and different values of swelling and actively generated wall stress. Examining the influence of the number of inner layer folds for different values of swelling disclosed higher inward buckling probability with more folds (Figure 7d). Since T act needed to cause buckling tended to plateau at 7 folds, we used 7 folds subsequently for illustrative purposes. T act needed to cause buckling decreased for a more swollen inner layer Figure (7dFigure 7-figure supplement 2c) and increased exponentially with inner layer volume loss. This finding was likely due to the less negative values of circumferential wall stress in the inner layer occurring with shrinkage ( Figure 7-figure supplement 2d). We found that an arterial wall consisting solely of SMCs and uniform matrix maintained a mean positive circumferential stress in the inner layer during contraction, that prevented buckling. Thus, a delicate biomechanical balance exists -reduced inner layer volume allows a smaller radius to be achieved via SMC contraction prior to buckling (Figure 7-figure supplement 2b), thus aiding closure, yet excess volume reduction of the inner layer increases the active stress requirement for buckling and achieving complete vessel closure ( Figure 7d). The umbilical artery can reduce its cylindrical radius dramatically at T act near a basal value of~50 kPa, progressing to buckling and closure via a subsequent near-maximal contraction ( Figure 7e). In agreement with the computational modeling, 20/25 of human umbilical arteries had 4 or more buckles, whereas those with no buckles in the area analyzed by histology were patent (Figure 7f). Other large mammalian species analyzed showed a similar phenomenon (Figure 7f). Thus, buckling of the proteoglycan-rich inner tunica media may be a crucial and evolutionarily conserved mechanism employed by all mammals for rapid arterial occlusion at birth.

Discussion
We report two distinct umbilical cord blood vessel specializations that may facilitate rapid umbilical artery occlusion at birth: a distinct proteoglycan-rich inner arterial TM, generating a bilayered arterial wall, and selective contraction of SMCs in the outer layer ( Figure 8). The rounded SMCs of the inner TM may be specialized for CSPG production rather than contraction, consistent with nuclear Sox9 staining, a function they exert prior to delivery. During delivery, lack of pMLC staining suggests that despite differentiated SMC marker expression, the inner cells play a passive role. Biomechanical testing and computational analysis confirm that selective proteoglycan enrichment in the inner arterial TM ensures that contracting SMCs in the outer TM can effectively occlude the arterial lumen at birth (Figure 8). Histologic and computational analysis showed buckling of the inner TM and fluid redirection into the resulting TM protrusions as critical mechanisms resulting from specialization of the inner and outer arterial TM. By in silico simulations of umbilical arteries with modulation of the contractile outer layer and proteoglycan-rich inner core, we demonstrate that complete occlusion can be achieved.
Our work suggests that a principal mechanism governing umbilical cord vascular dimorphism resides in extracellular matrix, namely, differentially regulated dynamics of aggrecan and versican, which may then modulate SMC development and differentiation. Other mammalian umbilical vessels examined, from animals as large as the walrus and elephant to as small as the mouse, showed similar CSPG and aggrecan modulation as in humans. Although Vcan mutant mice die before umbilical cord development is completed (Yamamura et al., 1997) and could not be studied, Acan and Adamts1 mutants demonstrated their mechanistic contributions to the observed dimorphism. Our emphasis on aggrecan in the inner layer, with its abundant GAGs and their high fixed charge density, was relevant to the computational findings of the importance of inner layer swelling and buckling in response to SMC contraction. The umbilical vein contains fewer contractile SMCs and has scant aggrecan and versican. Hence, SMC activation in the umbilical vein does not occlude the lumen to the same degree as in the arteries, a contention supported by computational analysis. Given the evolutionary pressure to achieve hemostasis urgently in the artery rather than the vein, these findings suggest a highly evolved mechanism for preventing exsanguination of the newborn that is potentially relevant to other embryonic shunts that occlude rapidly at birth.
The computational model was built on a long history of studying murine arteries and veins (Ferruzzi et al., 2013), but was specialized to the GAG-rich inner layer and SMC-rich outer layer of the umbilical artery. Modeling the dynamics of associated volume changes would have required a mixture or poroelastic model and significantly more experimental data, including measurement of layer-specific permeabilities and fixed-charge densities. Instead, we modeled the quasi-equilibrated states using a well-accepted approach wherein the degree of swelling can be adjusted for each simulation (Demirkoparan and Pence, 2007;Szafron et al., 2017). Interestingly, prior results by others Figure 4 continued umbilical cords). (d) Short umbilical cords in E18.5 Adamts1 -/and Acan -/embryos compared to wild type. Red arrowhead indicates an omphalocele in Adamts1 -/embryos. (e) H & E staining of E18.5 wild type, Acan -/and Adamts1 -/cord cross-sections showing thinner walls in Acan -/umbilical vessels and thicker walls in Adamts1 -/umbilical vessels. (f-g) Cord length, TM thickness and vessel luminal area quantifications for Adamts1 -/-(f) and Acan -/mice (g) at E18.5 compared to wild-type littermates. Acan -/umbilical cords show larger lumens and Adamts1 -/vessels show smaller lumens in (n = 7-11 umbilical cords each, whiskers indicate minimum and maximum values, *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001). (h) Phospho-histone H3 (pHH3) staining shows significantly fewer proliferating cells (white arrowheads) in Acan -/umbilical vessels. Dotted white lines mark the boundaries of vessel lumens (n = 4 cords each, whiskers indicate minimum and maximum values, **, p<0.001; *, p<0.05). Scale bars = 100 mm in (a), 25 mm in (c), 100 mm and 50 mm in (e). The online version of this article includes the following figure supplement(s) for figure 4:  showed that swelling of an initially unloaded, unilayered cylindrical tube consisting of a neo-Hookean material (which we used to model GAG-rich tissue) increases luminal diameter if the tube is unconstrained (Demirkoparan and Pence, 2007). The tube must be constrained, such as by a stiff outer layer surrounding the swollen layer if swelling is to decrease luminal diameter (Szafron et al., 2017). The abundance of contractile SMCs in the stiffer outer layer and their basal tone may in fact enhance outer layer stiffness contributed by extracellular matrix and hence buckling appears to be essential to augment contraction-induced closure of the umbilical artery (cf [Moulton and Goriely, 2011]).
This observed architecture of the umbilical cord is likely to have supported survival of mammalian species, humans included, well before formal obstetric involvement in labor. Cord clamping is the default practice today and has the sanction of convention, offering the option of immediate neonatal resuscitation if needed. In regard to the umbilical artery, it would replicate the effect of a natural and apparently conserved physiologic process that interrupts its blood flow during transition from fetal to neonatal life. The latest recommendation to clamp the cord later rather than immediately after birth appears to align better with the delayed closure of the umbilical vein. The present studies and SMMHC (red, right-hand panels) staining shows blunted dimorphism of Adamts1 -/umbilical artery and vein with stronger expression of differentiated SMC markers in Adamts1 -/umbilical vessels and acquisition of endomucin, a venous endothelium marker, by arterial endothelium (n = 3 umbilical cords each genotype) Scale bars = 100 mm in (a-c).  in humans and other mammals show that evolution has devised an umbilical cord-intrinsic mechanism that facilitates rapid arterial occlusion at birth, leaving the vein patent and permitting a placental infusion. This unfailing sequence ensures continuation of mammalian species without other formal intervention, since evolutionary success is not about minimizing poor outcomes, but ensuring survival of a significant majority. Continued on next page

Human and large mammal cords
Twenty-five human umbilical cords were obtained from uncomplicated term pregnancies either after vaginal birth (n = 13) or Cesarean section for obstetric indications (n = 12), that is, malpresentation or repeat Cesarean section. The samples were collected under an IRB exemption from Cleveland Clinic (EX-0118) for use of discarded tissue without patient identifiers. These cords were used for histological/immunohistologic analysis, synchrotron imaging, RNA in situ hybridization, and transcriptomics of inner versus outer umbilical artery TM. Animal cord sections were provided by Disease Investigations, Institute for Conservation Research, San Diego Zoo Global from the Benirschke archive.

Mutant mice
The Adamts1 transgenic allele (Adamts1 tm1Dgen ), referred to herein as Adamts1 -/-, was produced by insertion of an IRES-lacZ cassette into intron 1 of Adamts1 using homologous recombination in mouse embryonic stem cells (Oller et al., 2017). The Acan cmd-Bc allele was previously described (Krueger et al., 1999) and is referred to herein as Acan -/-. Mice were handled under standard conditions under approved IACUC protocols at the Cleveland Clinic (IACUC protocol nos. 18-1996 and18-2045) and University of Chicago (IACUC protocol no. 43751). Mutant mouse embryos were collected by timed matings of heterozygous mice by the detection of copulation plugs (taken as day 0.5 of gestation). Embryos were dissected out immediately following CO 2 mediated euthanasia and cervical dislocation of pregnant mice. Dissected whole embryos, with umbilical cords and placentas attached, were fixed in 4% paraformaldehyde at 4˚C overnight. Umbilical cords were dissected out the following day and washed thrice in PBS and embedded in paraffin or in 4% agarose for vibratome sectioning as previously described (Nandadasa et al., 2015).

Biomechanical and computational analysis
The umbilical artery and vein were obtained at E18.5 from mouse embryos (n = 4) following approval by the Yale University IACUC (protocol no 2018-11508), then mounted within a custom computercontrolled biaxial device designed specifically for biomechanical testing of murine vessels (Gleason et al., 2004). Vessel maintenance, pre-conditioning, biaxial loading protocols and data collection are described in Appendix 1. The umbilical artery was modeled computationally as a thickwalled, bilayered cylindrical tube subjected to swelling of the GAG-rich inner layer and active contraction of the smooth muscle-rich outer layer; the model also included a passive contribution of extracellular matrix as revealed by biomechanical tests.

Statistical analysis
Statistical analyses were carried out using GraphPad Prism analytical software (versions 6-8, Graph-Pad, San Diego, CA) in determining statistical significance using two-tailed Student's t-test. Statistical details including N and p values are provided in each corresponding figure legend. Statistical analyses for microarray gene expression were performed using Affymetrix's Transcriptome Analysis Console (TAC 4.0) through the RMA-SST sketch algorithm and R version 3.5.2. Fold changes were calculated by an empirical Bayes ANOVA method through the TAC 4.0 software. Details of these and additional methods are provided in Appendix 1. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. . Supplementary file 2. Microarray comparison of the transcriptome of the inner umbilical artery tunica media with the outer umbilical artery tunica media.

Author contributions
. Transparent reporting form

Data availability
All data generated or analysed during this study are included in the manuscript and supporting files.
The following datasets were generated: Microarray analysis of human umbilical cords NCS samples to evaluate changes in expression in artery versus vein were run on the Affymetrix Hu-Gene U219 microarray Peg-plate; 150 ng of input RNA from each sample was labeled using an Affymetrix 3' IVT labeling protocol. Samples hybridized to the PEG arrays were washed, stained, and scanned by the Affymetrix Gene Titan. RNA from two of the four cords was also evaluated on the Affymetrix Hu-Gene 1.1 ST Peg-plate microarray, using the whole transcriptome (WT) labeling protocol also starting with 150 ng of input total RNA. The hybridized PEG Arrays were washed, stained, and scanned by the Affymetrix Gene Titan according to standard protocols. RNA (150 ng) from inner and outer TM samples was labeled using Affymetrix's WT PLUS protocol. Labeled samples were hybridized overnight to Affymetrix Hu-Gene 2.0 ST microarray cartridges. The sample was removed, then the microarray cartridges were washed and stained on the Affymetrix GeneChip Fluidics Sta-tion450 and scanned by the Affymetrix GeneChip Scanner 3000. Total RNA was labeled using Affymetrix's FLASH Tag Protocol. Labeled samples were hybridized overnight to Affymetrix GeneChip miRNA 2.0 Array, which interrogates all the mature human miRNA sequences in miRBase Release 20microRNAs. The sample was then removed and the microarray cartridges were washed and stained on the Affymetrix GeneChip Fluidics Station450 and scanned by the Affymetrix GeneChip Scanner 3000. Gene expression analysis for each microarray study was performed using Affymetrix's Transcriptome Analysis Console (TAC 4.0) through the RMA-SST sketch algorithm and R version 3.5.2. Fold changes were calculated by an empirical Bayes ANOVA method through the TAC 4.0 software. Parameters for gene expression changes include a p-value 0.05 and a fold change value !1.5 and À1.5. Data collected from samples labeled by different labeling protocols, obtained using different scanners or hybridized to different arrays were analyzed as independent data sets.

Additional details of transgenic mice
The Adamts1 transgenic allele used here (referred to as Adamts1 -) was generated by Deltagen Inc (San Carlos, CA; Deltagen identifier T1288; MGI:5427602 (B6;129P2-Adamts1 < tm1Dgen>/H;)) by inserting an IRES-lacZ-neomycin resistance gene cassette into intron 1. RT-PCR using PCR primers bridging exon 1 and exon two showed that the insertion eliminated gene expression (Deltagen, unpublished data). The allele was deposited in the MRC Harwell, Frozen Embryo and Sperm Archive (Harwell, UK). Frozen Adamts1 +/embryos harvested at the two-cell stage of development from matings of wild-type C57BL/6 females with hemizygous males at MRC Harwell were obtained under an academic use license signed by the Cleveland Clinic with Deltagen. Frozen embryos were implanted into pseudo-pregnant female recipient mice at the Case Transgenic and Targeting Facility (Cleveland, OH). Subsequently, the Adamts1allele was crossed into the C57BL/6 strain for at least 10 generations and is maintained in this strain. The Acan cmd-Bc allele (referred to here as Acan +/-) was backcrossed to C57BL/6 for over 20 generations at the University of Chicago (Lauing et al., 2014) and subsequently transferred to the Cleveland Clinic Lerner Research Institute. b-Galactosidase staining of Adamts1 +/umbilical cords was done as previously described (McCulloch et al., 2009). Both bgalactosidase staining and genotyping were used to identify Adamts1 +/mice obtained from hemizygous with wild-type matings, whereas PCR genotyping was used to distinguish the three genotypes possibly arising from crosses of Adamts1 or Acan hemizygous animals which were used to generate homozygous embryos. For genotyping, genomic DNA was isolated from clipped toes between 7-10 days after birth or from embryo tails using 100 ml or 50 ml Direct PCR Tail Lysis reagent, respectively (Viagen, catalog number 102 T) containing 1 ml of proteinase K (Milipore Sigma, catalog number 3115879001) followed by digestion overnight at 55˚C. Alternatively, tail DNA was extracted using QuickExtract DNA (Lucigen) at 65˚C for 15 min and reaction was stopped by incubation at 100˚C for 5 min.

RNA in situ hybridization
Six-micron-thick paraffin sections were probed with RNAscope probes according to manufacturer's guidelines. The following probes were used: ACAN (Advanced Cell Diagnostics ( Cat. No. 400441). The probes were detected using RNAscope 2.5 HD Red reagent kits (ACD Cat. No. 322350) essentially as recently described (Mead and Apte, 2020).

Biomechanical characterization of mouse umbilical arteries and veins
Umbilical cords were excised and separated from the placenta from wild-type C57BL/6J mice at embryonic day E18.5 (n = 4). The umbilical arteries and veins were identified by opening the abdomen of the embryo and locating an artery by its connection to the iliac artery and a vein by its connection to the inferior vena cava through the ductus venosus. After separating and cleaning the umbilical arteries and veins from excessive adipose tissue, the intact vessels were cannulated on custom-drawn glass micropipettes, secured with silk sutures at each end, and mounted within a custom computer-controlled biaxial test device designed specifically for testing murine vessels. These cylindrical specimens were then immersed in a Krebs-Ringer bicarbonate solution (Krebs) and oxygenated with 95% O 2 and 5% CO 2 while maintained at 37˚C. The lumen of the umbilical artery closed immediately after separating the umbilical cord from the placenta but was relaxed by acclimation within the testing chamber for 5-15 min at 10 mmHg and wash-out with Krebs solution up to three times. The vessels were then subjected to two isobaric (luminal pressure of 5 mmHg, then 10 mmHg) -axially isometric (fixed specimen-specific in vivo axial stretch) contractions by adding 100 mM KCl to the bath to ensure viability of the SMC. The transmural organization of the vessel wall was monitored during different stages of contraction using an optical coherence tomography (OCT) system having an axial (depth) resolution <7 microns and lateral resolution of 8 microns (Callisto Model, Thorlabs, Newton, NJ).
For the subsequent passive tests, the normal Krebs solution was replaced with a Ca 2+ -free Krebs solution. The vessels were preconditioned via four cycles of pressurization (artery: 0-40 mmHg and vein: 0-25 mmHg) while held fixed at their individual in vivo axial stretch. Subsequently, the vessels were subjected to a series of seven biaxial testing protocols: cyclic pressurization over ranges noted above at three different fixed values of axial stretch and cyclic axial stretching at four different fixed values of luminal pressure (artery at 10, 20, 30 or 40 mmHg and vein at 1, 2, 10 or 20 mmHg). Data collected online included outer diameter, luminal pressure, axial length, and axial force, which facilitated robust parameter estimation of the biaxial biomechanical behavior (Ferruzzi et al., 2015). Toward this end, we used a validated 'four-fiber family' type constitutive relation (see below) that has proved useful in characterizing murine arteries (systemic and pulmonary) and veins.
Note that these biaxial data were used to build a baseline, bilayered cylindrical model for the purposes of parametrically exploring possible effects of different levels of smooth muscle contractility in the outer tunica media, GAG-induced swelling of the inner tunica media, and possible buckling of this innermost layer in silico. Hence, rather than focus on specimen-to-specimen differences, we sought 'mean' properties of the normal murine umbilical artery. Data from all seven passive biaxial testing protocols from all four mice were grouped as a single data set and best-fit values of the associated eight material parameters were determined simultaneously by minimizing the sum-of-thesquares differences between predicted and measured pressures and axial forces, as described previously (Ferruzzi et al., 2015). To facilitate a global, rather than local, minimization, we used multiple randomly generated initial guesses for the parameter estimation, accomplished using the MATLAB routine (lsqnonlin function). It is important to note that best-fit parameters in exponential constitutive relations need not be unique, hence what is most important is that together they yield correct values of biaxial stress (verified by comparison to measured values) and predict appropriate levels of material stiffness and stored energy. Finally, it is noted that initial parameter estimation was based on data from a representative specimen rather than all data combined; the subsequent model simulations were similar in both cases, suggesting that the baseline model captured salient features as desired.

Computational modeling of the umbilical artery Swelling and contraction
The traction-free, non-swollen configuration R; Q; Z ð Þ is treated as the reference with inner radius, interfacial radius, and outer radius denoted A; B; C ½ , respectively. Swelling is accounted for via previously outlined methods (Szafron et al., 2017;Demirkoparan and Pence, 2007), which leads to a residually stressed, traction-free configuration with corresponding material points mapped to The final loaded configuration, pressurized and axially stretched, is then characterized by a; b; c ½ in r; ; z ð Þ. Consider a deformation from R; Q; Z ð Þ to R Ã ; Q Ã ; Z Ã ð Þ via the deformation gradient tensor With n Ã ¼ 1, there is no swelling and the vessel is considered to have no residual stresses; in contrast, n Ã >1 (expansion) and n Ã <1 (shrinkage) yields self-equilibrating wall stresses in the absence of external loading. Note that qR Ã =qR ¼ Rn Ã =L Ã z R Ã . Further deformation to any loaded configuration r; ; z ð Þ from the swollen configuration R Ã ; Q Ã ; Z Ã ð Þ is then described by F P ¼ diag qr=qR Ã ; r=R Ã ; L z ð Þ, where the assumption of incompressibility during transient external loading requires det The matrix component F 11 can also be expressed in terms of the original derivatives, such that qr=qR Ã ð Þ qR Ã =qR ð Þ ¼ Rl z n Ã =r, which allows us to apply the chain rule and integrate R r a rqr ¼ R R A n Ã R=l z qR to find any radial point r within the vessel wall. The vessel is assumed to be quasi-equilibrated in any state, such that linear momentum balance requires div t ¼ 0, where t is the Cauchy stress tensor. Circumferential and axial equilibrium is satisfied identically at each r; ; z ð Þ. Radial equilibrium requires qt rr =qR þ t rr À t ð Þ=r ¼ 0. The Cauchy stress is specialized as t ¼ t ex À pI with p the Lagrange multiplier enforcing incompressibility during transient loading, I the identity tensor, and t ex the 'extra' part of the stress due to deformation and the constitutive response. Integration yields where P is the transmural pressure across the vessel wall, with P>0 indicating internal pressurization. We also calculate the overall axial load required for overall equilibrium, See Figure 7d. The equilibrium problem is solved iteratively for the loaded inner radius a for each luminal pressure P and axial extension l z . Constitutively, the extra part of the Cauchy stress can be computed from a stored energy density function W for the vessel, with t ¼ 2F qW=qC ð ÞF T = det F À pI and C ¼ F T F the right Cauchy-Green tensor. Due to the microstructure of the umbilical vessels, the GAG-rich inner layer is modeled as a neo-Hookean matrix that can swell (Demirkoparan and Pence, 2007) with 1 a shear modulus for this inner layer. As there is no evidence of collagen with a preferred orientation or smooth muscle cells capable of contraction within the inner layer, it is considered isotropic and passive. Fewer GAGs are present in the outer layer, but we include the possibility of a swellable matrix for illustrative purposes and to provide radial stiffness. The outer layer is then modeled using a modified four-fiber family model for a passive nonlinear stress-stretch behavior and a Rachev-type model for SMC contractility (i.e. active behavior), with a potential function B 1;3 f 000 þ B 1;2 f 00 þ B 1;1 f 0 þ B 1;0 f ¼ 0 h i r¼a and B 2;2 f 00 þ B 2;1 f 0 þ B 2;0 f ¼ 0 h i r¼a ; with coefficients B n;m given below. As there is little evidence of buckling in the outer layer, and the Rachev-type contractility model generally yields tensile stresses that would inhibit buckling, we specify that the incremental deformation vanishes at the interface of the two layers with f b ð Þ ¼ 0 and f 0 b ð Þ ¼ 0 and that the incremental shear traction is zero (Yang et al., 2007), which gives C 1;2 f 00 þ C 1;1 f 0 ¼ 0 h i r¼b and C 2;2 f 00 þ C 2;0 f ¼ 0 h i r¼b ; with coefficients C n;m given below. We used the compound matrix method (Haughton and Orr, 1997;Lindsay and Rooney, 1992), a modification of the determinantal method commonly used for linear bifurcation analysis (Haughton and Ogden, 1979), to solve the differential equation numerically. The fourth order equation above was rewritten as a system of four, first order differential equations, y 0 ¼ Ay, where y ¼ f ; f 0 ; f 00 ; f 000 Â Ã T and A is the corresponding coefficient matrix. Boundary conditions at the inner surface and interface were similarly re-written as By ¼ 0 ½ a and Cy ¼ 0 ½ b , respectively. We define two linearly independent initial conditions at r ¼ a, which can be integrated to r ¼ b to create linearly independent solutions y 1 ð Þ and y 2 ð Þ such that y ¼ k 1 y 1 ð Þ þ k 2 y 2 and similarly, f 3 ¼ 1; 4 ð Þ, f 4 ¼ 2; 3 ð Þ, f 5 ¼ 2; 4 ð Þ, f 6 ¼ 3; 4 ð Þ. We evaluate C k ¼ jC k j with C 1 ¼ C 1;1 C 2;1 C 1;2 C 2;2 ¼ 1; 2 ð Þ; C 2 ¼ C 1;1 C 2;1 C 1;3 C 2;3 ¼ 1; 3 ð Þ; and similarly, C 3 ¼ 1; 4 ð Þ, C 4 ¼ 2; 3 ð Þ, C 5 ¼ 2; 4 ð Þ, C 6 ¼ 3; 4 ð Þ, where n; m ð Þ C 1;n C 2;n C 1;m C 2;m : To evaluate the new bifurcation condition, we create a system of equations f 0 ¼ Af and identify the components of f 0 as, for example, The components of A can thus be conveniently defined in terms of the original components of A (Haughton and Orr, 1997), as listed below. This new system is then integrated from a to b using a fourth-order Runge-Kutta method, and T act is varied iteratively until the boundary condition at b is satisfied, namely Loading conditions, including the volume change n Ã , luminal pressure P, and axial stretch l z , can be varied parametrically to understand their effects on the critical value of T act needed to induce buckling. Note, one may also fix the value of T act and identify the critical value of a different loading variable of interest.

Bifurcation analysis -specific functions
For the fourth order governing differential equation for the incremental displacement: rr þ rB 0 rr þ n 2 À 1 ð Þ Ã B rr þ n 2 B À B rrrr ð Þ A 1 ¼ r r 2 B 00 rr þ rB 0 rr À B rr À n 2 B þ B rrrr ð Þ À n 2 r B 0 þ B 0 rrrr ð Þ A 2 ¼ r 2 r 2 B 00 rr þ 7rB 0 rr þ 5B rr À n 2 B þ B rrrr ð Þ Note: These coefficients include only the non-zero components of B for the considered stored energy density function.