**3. Discussion**

The present study identified a number of distinct features of collateral ECs and SMCs that may underlie or contribute to the unique characteristics of collaterals outlined in the Introduction. We found that despite the low and oscillatory shear stress present in collaterals at baseline in the absence of obstruction, their ECs are aligned with the vessel axis to the same degree present in distal-most arterioles and the descending aorta—vessels with high-velocity orthograde laminar flow. Endothelial cells of both collaterals and arterioles have primary cilia, and collaterals possess fewer of them. Smooth muscle cells of collaterals are continuous, unlike those of DMAs. Collaterals have higher levels of expression of genes associated with both pro- and anti-inflammatory and pro- and anti-proliferation pathways, compared to DMAs. A bias towards a higher rate of cell proliferation in collateral mural cells [18] is supported by the observation that collaterals begin to acquire tortuosity shortly after their formation in the embryo that increases through middle age (16 months, 49 human year-equivalents [47]). The above findings provide insights into structural and molecular specializations that may underlie the unique features and functions of collateral vessels.

Convergence of blood flow in collaterals at baseline imposes low and "disturbed" flow/shear stress forces on their mural cells, i.e., either absence of flow or very low flow that oscillates to-and-fro (~1–10 times per minute) and averages zero [6,15]. This results in increased wall stress according to the Bernoulli relationship. When present elsewhere in the arterial circulation, e.g., at bifurcations, the inner arch of the aorta or downstream of plaques, low and disturbed shear stress favors a non-aligned cobblestoned EC morphology that is associated with increased oxidative stress, inflammation, markers of aging, and low eNOS/NO activity (i.e., endothelial dysfunction) [28–31,44,45]. Surprisingly, however, we found that collateral ECs have the same alignment (and cell dimensions) as ECs in DMAs and the descending aorta. We did not investigate how this anti-inflammatory structural phenotype is specified. It is possible that one or more of the other unique features that we identified are involved (see below). On the other hand, the data in Figure 2 for flow and shear stress were obtained in anesthetized animals. During awake behaviors, collaterals may have periods of sustained flow in one or the other direction induced by changes in regional metabolic activity in the territory supplied by the arterial trees that they cross-connect. It is not known if collaterals contribute in this way to physiological metabolic regulation of blood flow and oxygen delivery, i.e., neurovascular coupling in brain and functional hyperemia elsewhere. However, such periods of sustained unidirectional flow could promote the EC orientation we observed. Irrespective of underlying cause, we speculate that the aligned phenotype of collateral ECs is part (or a marker) of a group of protective mechanisms that favor maintenance of collaterals and mitigate their rarefaction (Figure 9).

To our knowledge, this is the first report showing that ECs lining collaterals and arterioles have primary cilia. They are much more abundant than reported previously for conduit vessels from healthy individuals (i.e., absent or present on less than 1% of ECs [32,34]): 18% of collateral ECs have PrC compared to 28% for DMAs. Primary cilia on ECs were described in 1984 by Haust in aorta of rabbits and humans with atherosclerosis [48]. Subsequent reports described ciliated ECs in capillaries of the pineal gland of a 20 week-old human fetus [49], developing heart and aortic valve leaflets [33,50,51], and references therein], surrounding atheromas [52], ectopically in the common carotid artery of *ApoE*−/− mice, and in healthy individuals at bifurcations of conduit arteries and the inner curvature of the aortic arch; by contrast cilia are absent or nearly so in regions of arteries where shear stress is laminar [33,40,50]. Most other cell types express PrC during development, under certain conditions of cell culture [53], and in adults [35–37]. Depending on cell type, PrC participate in specification of embryo asymmetry, centriole disposition, proliferation/cell cycle regulation, autophagy, flow-sensing mechanotransduction, chemoception, and compartmentalization and tra fficking of signaling proteins among the cilioplasm, cytoplasm, and nucleoplasm (e.g., for Gli and PDGFR α) [35–37]. When present, cilia are most abundant in non-proliferating cells, with the proximal end anchored in an invagination of the plasmalemma (ciliary pocket) in ECs and other but not all cell types [33,34,36]. Primary cilia are complexed with the mother centriole of the basal body that links to the microtubule-organizing center (MTOC) [32,35–37,54]. Disassembly/resorption of the cilium during S-phase and centriole liberation are essential for cell division [36,37,54]. Since PrC are linked to the cytoskeleton via the MTOC, flow-induced ciliary bending in ECs [55] is capable of being transmitted throughout the cell, including to cell–cell and cell–matrix junctions [33,37]. Primary cilia transduce fluid shear stress in renal tubular epithelial cells and ECs through a pathway that that is exceptionally sensitive to shear stress and involves polycystin-1 and polycystin-2, which are encoded by *Pkd1* and *Pkd2* [35–37]. Polycystin-1 has mechanosensitive properties while polycystin-2 is a TRP calcium channel. Both proteins are required to sense shear stress and in turn release nitric oxide [33,37]. Defects in PrC are associated with many abnormalities. For example, mutations of *PKD1* and *PKD2* are causal for autosomal dominant polycystic kidney disease, with kidney ECs and tubular cells of patients evidencing deficient calcium and NO responses and increased proliferation [33,35–37].

Primary cilia are absent in human umbilical vein ECs maintained under laminar shear stress and proliferative quiescence in cell culture. In human umbilical veins less than one percent of ECs have cilia that protrude into the lumen, while in a larger fraction the cilia are located intracellularly [34,55]. Embryonic aorta and ECs cultured from it have a single cilium that projects into the lumen [38–40]. Endothelial cilia are present in regions of high shear stress during embryonic development, along with expression of the shear stress-sensitive transcription factor, KLF2, which transactivates *eNOS* and other anti-inflammatory and anti-proliferative genes [44,45]. In regions with low or disturbed shear stress PrC are disassembled/absent and expression of *Klf2* and *eNOS* are abolished and reduced, respectively [33,56]. Expression of *Klf2* is also inhibited in non-ciliated ECs isolated from embryonic arteries, and chemical removal of PrC from ECs in culture has a similar e ffect, i.e., abolishing *Klf2* expression [57]. Interestingly, in ECs of adult *ApoE*−/− mice, which have endothelial dysfunction but do not develop plaques, cilia are expressed ectopically in the common carotid artery despite the presence of laminar flow, compared to wildtype mice that are devoid of cilia [33]. Cilia were lost when high shear stress was induced via implantation of a flow-restricting cast around the vessel. Casting of common carotids in wildtype mice induced ciliogenesis only in regions of low and disturbed shear stress. These findings sugges<sup>t</sup> that expression of PrC on ECs in vivo in adults is limited to regions of low/disturbed shear stress but can occur ectopically in arteries with laminar flow in the presence of endothelial dysfunction caused by hyperlipidemia [33] and possibly other vascular risk factors. Of note, in highly ciliated, disturbed-flow areas such as the inner curvature of the aortic arch or downstream of plaques, approximately 25% of ECs have a single cilium while the rest are devoid [33], a percentage similar to what we observed in arterioles and collaterals.

Our findings that PrC are also present on arterioles and collaterals in healthy young adult mice highlight the need for studies examining cilia function in these vessel types. This includes determining whether our observation of fewer cilia on collateral ECs than arterioles has functional significance. Endothelial cells are coupled mechanically, electrically, and di ffusionally to adjacent ECs and SMCs [58], thus only a fraction of ECs may need to express cilia for transduction of mechano-sensitive or other signals. High shear stress causes disassembly of cilia in cultured ECs, while oscillatory flow reversal induces their expression [33]. We speculate that cilia on arteriole and collateral ECs may reflect the lower flow/shear stress in arterioles and very low and disturbed flow in collaterals and that fewer PrC on collateral ECs may serve to reduce their sensitivity to the prevailing disturbed shear stress environment. In other words, fewer cilia on collateral ECs may be part of a repertoire of adaptations that balance against or oppose—through maintained or increased expression of KLF2/4, eNOS, and other anti-inflammatory/anti-proliferative factors—the low-grade inflammatory, oxidative, proliferative, and apoptotic signals promoted by the disturbed hemodynamic environment present in collaterals (Figure 9). Egorova et al. [36] proposed something similar, i.e., that since presence of PrC on ECs is associated with KLF2 expression, endothelial cilia may signal a brake on EC activation in regions of low and disturbed flow. A protective role for cilia in these regions [36,37] is supported by the recent report that removal of endothelial cilia using conditional deletion of *Ift88* increased atherosclerosis and inflammatory gene expression, and decreased eNOS activity in *Apoe*−/− mice fed a high-fat diet [59], and

that the endothelium becomes sensitized in athero-prone regions to undergo osteogenic differentiation in Tg737 (*orpk*/*orpk*) cilium-defective mice [60]. It is also possible that if PrC are protective, fewer of them in collaterals could contribute to the high susceptibility of these vessels to rarefaction from aging and other vascular risk factors. However, fewer cilia on collaterals might simply reflect a secondary or bystander effect, being a consequence, for example, of a higher inherent proliferation rate of collateral ECs as evidenced by the progressive increase in collateral tortuosity (discussed below), since presence of PrC and their association with the basal body is believed to favor removal of cells from the cell cycle [35–37]. Future studies will be required to determine if our finding of multiple cilia on ECs reflects ECs that have undergone proliferative senescence and associated failed cytokinesis and nuclear polyploidy [39].

It has recently been shown that ECs in the developing mouse retina rely on PrC to stabilize vessel connections during remodeling of the vascular plexus in regions with low to intermediate shear stress [61]. Endothelial cilia sense flow in zebrafish embryos, participate in recruitment of mural cells to arterial fated vessels, and are required for normal vascular morphogenesis [62,63]. The number and diameter of collaterals declines beginning in middle age [17]. This age-induced rarefaction is strongly accelerated by genetic or pharmacologically induced eNOS/NO deficiency or the presence of vascular risk factors [16,19]. Increased shear stress induces collateral outward remodeling following acute or slowly developing arterial obstruction [1,2,5]. *Pkd1*+/− mice and patients with autosomal dominant polycystic kidney disease have endothelial eNOS/NO dysfunction [64]. It will be important to examine in future studies whether collateral PrC participate in one or more of the above functions using EC-specific knockdown of polycystin-1, since: 1) deficiency in it leads to altered ciliary function, 2) polycystin-1 together with polycystin-2 participate in flow-sensing by PrC, 3) mutant forms of either protein cause polycystic kidney disease [35–37], 4) there is evidence that VHL, independent of its role in the degradation of Hif1<sup>α</sup>, together with GSK3β are required for structural maintenance of the cilium [65], and 5) the protein Rabep2, which is required for collaterogenesis [11], is a novel substrate of GSK3β [66], localizes at the cilium–basal body complex, and knockdown of it leads to defective ciliogenesis [67]. Other approaches to interfere with cilia presence and function, e.g., with knockdown of other ciliary proteins such as *Pkd2* and *Ift88*, also will need to be examined.

In contrast to distal arterioles, which have sparse and discontinuous SMCs in various tissues including retina (we were unable to identify studies in brain) [41–43], SMCs were continuous on collaterals. We speculate this may be an adaptive increase in wall thickness to balance the increase in circumferential wall stress caused by the Bernoulli-specified conversion of kinetic energy of flow to increased potential energy (transmural pressure) as a consequence of flow-convergence in collaterals. It would be interesting to examine whether the composition and amount of extracellular matrix, which could assist SMCs in balancing the increased wall stress in collaterals, differs in collaterals versus arterioles. Of note, despite their increased SMC coverage, collaterals have less rather than more tone compared to similarly-sized arterioles, and lack myogenic responsiveness—additional unique features of collateral vessels [26,27].

The disturbed hemodynamic, pro-oxidative environment in which collateral mural cells reside led us to examine whether expression of genes involved in inflammation, cell proliferation, aging, and angiogenesis differ for collaterals versus distal arterioles. Collaterals displayed increased mRNA levels for the pro-inflammatory, pro-apoptosis inflammasome gene, *Pycard*, the pro-proliferative genes, *Ki67, Pdgfb*, and *Angpt2*, the anti-proliferative gene, *Dll4*, and the differentiated arterial-type EC marker gene, *Ephrinb2*. However, expression of cell cycle inhibitor genes, *p21, p27*, and *p53*, were not different, nor were other genes associated with proliferation, cell cycle arrest, and aging (*p<sup>16</sup>Ink4a, Ampk, Sirt1, telomerase*). Neither were there differences in expression of other genes associated with EC and/or SMC proliferation (*Vegfa, Flk1, Clic4, Pdgfa, Flt1*) and that are required (in the case of the first three genes [7,8]) for formation of collaterals during development or involved with specifying EC and SMC differentiation and quiescence (*Tgfb, Angpt1*). Increased expression by collaterals of the above pro-proliferation genes is consistent with our tortuosity measurements, which sugges<sup>t</sup> that collateral

mural cells have a higher proliferation rate, compared to other arterial vessels: collateral tortuosity was evident by the first day after birth, continued to increase through middle age, and then declined. The latter occurred at the same time that collaterals experience a decline in number and diameter with advanced aging [17]. These findings support the hypothesis that age-associated collateral rarefaction is caused by proliferative senescence and subsequent apoptosis of collateral ECs and SMCs due to a lifetime elevation in proliferation rate caused by the disturbed hemodynamic and low blood oxygen content environment in which collaterals reside (Figure 9).

Collaterals also displayed increased activity of eNOS, which previous studies have shown opposes rarefaction of collaterals caused by aging and other vascular risk factors [16,18,19]. eNOS-derived NO inhibits oxidative stress, inflammation, proliferation, leukocyte adhesion, platelet aggregation, and cellular aging and promotes SMC relaxation [58,68]. Since shear stress is a proximate stimulus for eNOS-derived NO, increased eNOS/NO in collaterals with their low and disturbed shear stress environment may lessen the effect of factors promoting collateral rarefaction (Figure 9). Likewise, maintained expression of the EC shear stress-sensitive transcription factors, *Klf2* and *Klf4*, in collaterals despite their low and oscillatory flow, which inhibits expression of these factors elsewhere in the arterial vasculature with disturbed flow [44], may act as additional "balancing" factors or collateral specializations, along with increased eNOS, aligned ECs, fewer cilia, robust SMC coverage, and increased ephrin-B2 and Dll4. Expression of KLF2 and KLF4, which negatively regulate proliferation, inflammation, and angiogenesis, upregulate eNOS, and are sharply downregulated at sites of low and disturbed shear stress, were not different in collaterals versus DMAs. Interestingly, PrC promote *Klf2, Klf4*, and *eNOS* expression [35–37,44,69,70].

A limitation of the above studies is that RNA was obtained from dissected vessels, which are composed of ECs, SMCs, and, although less so, pericytes, fibroblasts, and resident myeloid cells. Studies are needed that employ separation of cell types and examination of a wider array of genes and their respective protein levels. However, the difficulty in manually dissecting collaterals and distal arterioles in the required numbers, the effect of cell dissociation techniques on baseline levels of RNA and protein, the absence of cell culture models of "collateral" ECs and SMCs, and the as of ye<sup>t</sup> lack of any collateral-specific marker gene, preclude the use of these approaches. Of note, however, expression of several of the genes examined are specific or enriched for ECs, e.g., *Flk1, Angpt1, Angpt2, Ephrinb2, DLL4, eNOS, Clic4, Klf2*, and *Klf4*. However, analysis of gene transcription does not always reflect changes in protein level or function; thus the investigation of oxidative stress, inflammatory, proliferation, or senescence markers at the protein level may better reflect differences in cellular characteristics.

### **4. Materials and Methods**

Three to 4 month-old male C57BL/6 (B6) wildtype (lab colony, Jackson Laboratories breeders, Bar Harbor, ME, USA) or genetically modified mice were studied. Scanning electron microscopy was performed on the cerebral arterial vasculature that was filled with Batson's #17. Angiography and morphometry were performed after filling with yellow Microfil® [18]. Prior to infusion of the casting agents, and also for immunohistochemistry, the vasculature was cleared of blood and perfusion-fixed after maximal dilation with 10−<sup>4</sup> M nitroprusside. Filling was confined to the pre-capillary vessels by adjusting the viscosity and input pressure of the casting material. All pial collaterals between the anterior cerebral artery (ACA) and middle (MCA) trees of both hemispheres were identified; tortuosity (axial length of the collateral ÷ scalar length connecting the collateral's endpoints [17]) and lumen diameter, EC orientation, and other morphometrics were obtained at midpoint for each collateral and an average value obtained for each animal unless indicated otherwise in the figure legends. Permanent MCA occlusion was by electro-cautery occlusion of the M1-MCA just distal to the lenticulostriate branches [18]. Right femoral artery (FA) occlusion was achieved by ligating the superficial FA immediately proximal and distal to the superior epigastric artery, which was also ligated [18]. Hindlimb perfusion was measured by laser Doppler perfusion imaging of the plantar foot and the adductor thigh regions. Values reported are means ± SE, with significance at *p* < 0.05; *n*-sizes (number of animals studied) and statistical tests are given in the figure legends.

All applicable international, national, and/or institutional guidelines for the care and use of animals were followed, including the National Institutes of Health Guide for the Care and Use of Laboratory Animals (UNC IACUC #18-123.0-A, 1 April 2019). This article does not contain any studies with human participants performed by any of the authors.

### *4.1. Angiography and Morphometry*

As previously described in detail [20] animals were anesthetized deeply with ketamine and xylazine and heparinized. The distal thoracic aorta was cannulated, right atrium perforated, and the mouse was perfused with PBS containing freshly prepared sodium nitroprusside (10−<sup>4</sup> M, for maximal dilation) and Evans blue dye (for light staining of brain and the endothelial surface) at ~100 mmHg. One ml of yellow Microfil (Flowtech Inc, Carver, MA, USA) was infused at a viscosity adjusted to fill the entire pial arterial and collateral circulations with su fficient pressure (~100 mmHg) and duration to cause limited capillary transit and venous filling to assure complete filling of all precapillary vessels and collaterals. After the Microfil had set for 20 min, brains were kept in 4% PFA and collaterals were imaged the next day using a Leica fluorescent stereomicroscope. All collaterals between the anterior cerebral artery (ACA) and middle cerebral artery (MCA) trees of both hemispheres were counted and divided by 2 to give the average number per hemisphere. Images were subsequently analyzed (ImageJ, NIH, Bethesday, MD, USA): Collateral lumen diameter was determined at midpoint at 50X for all ACA-MCA collaterals and averaged for each mouse.

### *4.2. Permanent Middle Cerebral Artery Occlusion (pMCAO)*

As previously described [20], mice were anesthetized with ketamine and xylazine (100 and 10 mg/kg, ip, respectively) and rectal temperature was maintained at 37 ± 0.5 ◦C. The temporalis muscle between the eye and ear on one side was retracted after a 4 mm incision. After a 2 mm craniotomy (18000-17 drill, FST, Foster City, CA, USA), the dura was incised with a 27 gauge needle tip and reflected to reveal the main trunk of the MCA which was cauterized (18010-00, FST, modified) distal to the lenticulostriate branches. The incision was closed with suture and Vetbond (3M, Minneapolis, MN, USA), intramuscular cefazolin 50 mg/kg and buprenorphine were administered, and the animal was monitored in a warmed cage during recovery from anesthesia to maintain the above rectal temperature. Mice were euthanized 4 days after pMCAO. Pial collateral morphometry was then performed as described above.

### *4.3. Laser Doppler Perfusion Imaging and Hindlimb Ischemia Model*

Femoral artery ligation (FAL) was performed and hindlimb perfusion was measured using a laser Doppler perfusion imager (Model LDI2-IR, Moor Instruments, Wilmington, DE, USA, ~2 mm penetration and high resolution) as described previously [71]. Briefly, the anterior thigh and adductor regions were depilated followed by 24 h to recover from any erythema. Mice were then anesthetized with 1.25% isoflurane/O2, rectal temperature was maintained at 37 ± 0.5 ◦C. A 3 mm incision was made overlying the femoral vessels 5 mm proximal to the knee. The femoral artery was gently isolated and ligated twice with 7-0 suture immediately distal to the lateral caudal femoral artery and 1 mm further distally, followed by transection between the ligatures. The superficial epigastric artery was also ligated. The skin was closed using 5-0 silk. Mice were placed in a heated chamber to block ambient light, and rectal temperature was maintained at 37 ± 0.5 ◦C. Less than 5 min after ligation, the plantar and adductor regions of both legs were laser-scanned. Average perfusion within the region of interest, drawn to outline the plantar surface of the paws, was calculated using Moor LDI Image Processing V5.0 software and reported as the ratio of ligated to the non-ligated hindlimb, where the region of interest was defined according to anatomic landmarks as described previously [71].

### *4.4. Quantitative NanoString Expression Analysis*

As described previously [46], mice placed under deep anesthesia (ketamine and xylazine, 100 and 10 mg/kg, ip, respectively) were perfused with 1 ml RNAlater ®(Sigma-Aldrich Corp, St. Louis, MO, USA) premixed with 10% Evans blue dye through the thoracic aorta. The brain was then removed and immersed in RNAlater ®. Approximately 10 pial collaterals and 10 nearby similarly sized distal-most arterioles per mouse (total 36 mice) were micro-dissected and pooled in RNAlater ®as 6 samples. Samples were homogenized (TH, Omni International, Marietta, GA, USA) in Trizol Reagent (Invitrogen, Carlsbad, CA, USA). Total RNA was purified using the RNeasy Micro Kit according to the manufacturer (Qiagen, Valencia, CA, USA). RNA concentration and quality were determined by NanoDrop 1000 (Thermo Scientific, Wilmington, DE, USA) and Bioanalyzer 2100 (Agilent, Foster City, CA, USA), respectively. Measurement of transcript number was conducted for 22 selected genes by the genomics facility at UNC using NanoString custom-synthesized probes (NanoString, Seattle, WA, USA). Transcript number for each gene was normalized to one of the following 6 housekeeping genes selected to be in range of the target gene under analysis: *Gapdh*, β*actin*, *Tubb5*, *Hprt1*, *Ppia* and *Tbp*.
