**2. Results**

### *2.1. Collateral Endothelial Cells Are Aligned with the Vessel Axis Despite their Chronic Exposure to Low and Oscillatory Shear Stress*

As a first-test of the hypothesis that collateral wall cells have unique phenotypes, we assessed the orientation of collateral ECs using scanning electron microscopy (SEM) and staining of the junctional protein zona occludens-1 (ZO-1). Despite the unique low and disturbed shear stress present in collaterals, ECs of collaterals were aligned with the vessel axis to the same extent, i.e., ~4 degrees off the longitudinal axis, which was present in nearby distal-most arterioles (DMAs) and in the descending thoracic aorta where blood flow is high-velocity and laminar (Figures 1 and 2). The area, perimeter, length, and width of collateral ECs were also comparable to ECs lining DMAs (Figure S1).

**Figure 1.** Pial collateral endothelial cells are aligned with the vessel axis. Scanning electron micrograph (SEM) of a corrosion cast of Batson's #17-filled cerebral pial arterial vessels and 2 collaterals, fixed after maximal dilation, which overlie the watershed zone between the anterior (ACA) and middle (MCA) cerebral artery trees. Upper inset, Microfil® cast of arterial vessels and collaterals (stars) in optically cleared brain. SEMs were obtained from six mice (see also Figure S1). Penetrating arterioles are evident branching from collaterals and pial arterioles.

**Figure 2.** Collateral endothelial cells are aligned with the vessel axis despite having low and oscillatory flow/shear stress in the absence of arterial obstruction. (**A**) Data were obtained in anesthetized mice via cranial window and previously published in reference [6]; unlike distal-most arterioles (DMAs) with diameters comparable to collaterals (COLs) and penetrating arterioles, COLs examined over 30 s intervals have either no flow or slowly oscillating, low velocity, to-and-fro flow with ~zero net-direction. After ligation of the MCA trunk (MCAO), flow to its territory reaches that evident in DMAs within 10–30 s. (**B**–**E**) Collateral endothelial cells (ECs) have the same "anti-inflammatory" alignment (~4 degrees from horizontal) as ECs of DMAs and the descending thoracic aorta, despite having low/disturbed shear stress at baseline. This is in contrast to the "pro-inflammatory" non-alignment present in the inner curvature of the aortic arch. In this and subsequent figures, data are means ± SE for "*n*" number of mice. Data in D determined from SEM images, *n* = 6 mice. Panel E magnification bar is 25 μm. Panel A 2-sided t-tests for shear stress followed by Bonferroni correction for \*\* *p* < 0.01 vs distal arterioles; Δ *p* < 0.05 vs penetrating arterioles; 2-sided t-test for ttt *p* < 0.001 vs before MCAO. ZO-1, zona occludens-1 immunohistochemistry.

#### *2.2. Endothelial Cells of Collaterals and Distal-Most Arterioles Have Primary Cilia; Collaterals Have Fewer*

While examining EC orientation using SEM we noticed the presence of casts of channels penetrating into a fraction of the ECs lining collaterals (Figure 3). Based on studies in other cell types including bovine aorta and human umbilical vein ECs [32–34], these are invaginations of the plasmalemma that abut the ciliary membrane to form the ciliary pocket, which houses the proximal end of luminal primary cilia (PrC) that were removed by shearing forces ("depilated/de-ciliated") during infusion of Batson's #17. The Batson's-filled ciliary pockets (i.e., PrC) were commonly located near the nucleus, in accordance with the base of the cilium being associated with the basal body [34–40]. Distal arterioles also have cilia (Figure S1). One, two, or rarely three cilia were present in ECs that expressed them (Figure 3, Figure S2). More than one cilium per EC has not been found in previous studies, to our knowledge; however, previous reports have only examined large conduit vessels. The above method, which serendipitously identified PrC, is indirect since ciliary pockets were what were detected. We therefore confirmed their presence using immunofluorescence (Figure 4). Eighteen percent of collateral ECs expressed PrC, as compared to 28% of DMAs; thus ECs lining collaterals had 34% fewer cilia. We have not found other reports that ECs within vessels of the microcirculation express primary cilia.

**Figure 3.** Collateral endothelial cells have primary cilia. (**A**) SEM of a corrosion cast of pial arterial vessels and a collateral, fixed at maximal dilation. (**B**) Stars identify casts of plasmalemmmal invaginations that contain the proximal end of the primary cilia (PrC) filled with Batson's #17 after removal of the PrC by shearing during infusion of the casting agent; each EC has 0–3 PrC. (**C**) Higher magnification SEM of a ~2 μm long PrC invagination.

**Figure 4.** Collateral endothelial cells have fewer primary cilia than distal-most arterioles (DMA). Immunofluorescent-stained collaterals (COL) with focal plane set within the lumen above the far-wall, showing primary cilia. Figure S2 shows cilia on DMAs. Inset, higher magnification. Right panel, *n* = 6 mice, 2-sided t-test.

### *2.3. Collaterals Are Invested with a Continuous Layer of Smooth Muscle Cells, unlike Distal-Most Arterioles Whose Smooth Muscle Cells Are Discontinuous*

Smooth muscle cells (SMCs) become sparse and discontinuous on distal arterioles in many tissue types [41–43]. Unlike arterioles that have orthograde flow, collateral blood flow converges from opposite directions, resulting in an average flow of near or at zero in the center-most segmen<sup>t</sup> of the collateral in the absence of occlusion (Figure 2A). The kinetic energy of flow is therefore converted to potential energy, which increases the circumferential wall stress of collaterals. Accordingly, we reasoned that SMC investment of collaterals might be greater than that present on DMAs to balance this increased wall stress. Figure 5 supports this hypothesis.

### *Int. J. Mol. Sci.* **2019**, *20*, 3608

**Figure 5.** Collaterals are invested with a continuous layer of smooth muscle cells (SMCs), unlike distal-most arterioles (DMAs) that lack or have discontinuous SMCs. (**A**), Immuno-fluorescent staining of SMCs (αSM-actin) and ECs (IB4-lectin). Representative image of pial collaterals, DMAs and penetrating arterioles (PA). (**B**), Brightfield image of αSMA-stained collateral and PA filled with yellow Microfil®, then freed from surrounding pial membrane.

### *2.4. Gene Expression Di*ff*ers for Collaterals Versus Distal-Most Arterioles*

Given the disturbed hemodynamic, pro-oxidative stress (i.e., low blood oxygen content) environment in which collateral mural cells reside, plus the high susceptibility of collaterals to undergo rarefaction with aging, vascular risk factor presence, and eNOS/NO deficiency [16–21,25] compared to nearby DMAs [17], we postulated that expression of genes involved with inflammation, cell proliferation, aging, and angiogenesis differ for collaterals and DMAs. To test this hypothesis, we measured transcript levels of 22 such genes (Figure 6), as well as eNOS immunofluorescence (Figure 7). Collaterals had increased expression of *Pycard, Ki67, Pdgfb, Angpt2, Dll4, Ephrinb2*, and eNOS, whereas 16 other genes were not significantly different. Of note, expression of the EC shear stress-sensitive transcription factors Klf2 and Klf4, which promote anti-proliferative, -inflammatory, and -angiogenic processes, upregulate eNOS, and are sharply downregulated at vascular sites of low and disturbed shear stress [35–37,44,45], were, in contrast, not downregulated in collaterals compared to DMAs with laminar high-velocity flow.

**Figure 6.** Gene expression differs for collaterals versus distal-most arterioles. Upper left panel, pial arterial vasculature perfusion-fixed at maximal dilation then filled with PU4ii polyurethane. Stars identify penetrating arterioles, including three that bifurcate and descend into the cortex immediately below the arteriole or collateral (green stars). Ten collaterals and 10 nearby similarly-sized distal-most arterioles DMAs were dissected from each of 36 mice and pooled into six samples for extraction of RNA. Transcript abundance was determined by Nanostring n-Counter® for 22 genes, each normalized to one of six housekeeping genes (*Gapdh,* β*actin, Tubb5, Hprt1, Ppia, Tbp*) selected for comparable level of expression [46]. \* *p* < 0.05, \*\* *p* < 0.01 by 2-sided t-test for collaterals versus DMAs.

**Figure 7.** Collaterals express increased levels of phospho- and total eNOS compared to arterioles. Top panels, immunohistochemistry for phospho- (**A**, red) and total (**B**, green) eNOS, *n* = 4 mice, 1-sided t-tests, 163x magnification. \* *p* < 0.05, \*\* *p* < 0.01.

### *2.5. Changes in Tortuosity over a Collateral's "Lifetime" Suggests Accelerated Proliferative Senescence of their Mural Cells*

We previously reported that aging, hypertension, and other vascular risk factors cause a loss of collateral number and a smaller diameter in those that remain [17–19,25]. Aging-induced rarefaction becomes evident in late middle-age in mice (16 months of age), is not seen in DMAs, is accelerated in onset by the presence of other vascular risk factors, and is associated with increased cellular markers of oxidative stress, inflammation, proliferation, and aging, as well as increased vessel tortuosity [17–19,25]. Tortuosity is a hallmark of collaterals. We postulated that it arises from a persistently increased rate of proliferation of collateral wall cells (confirmed in [18]) that is driven by the disturbed hemodynamic conditions present in collaterals. The above studies did not examine tortuosity at time points earlier than 3 months of age. To determine whether collaterals begin to acquire tortuosity from formation onward, we examined mice on embryonic day E15.5 (collaterals form between E15.5 and E18.5 [6,7]) and on post-natal day-1 and day-21. Tortuosity was absent at E15.5 but became evident by P1 and was significant by P21 (Figure 8). When viewed in context with our previous data for tortuosity at later ages (Figure 9; human year-equivalents for the latter five bars are approximately 13, 19, 49, 69, 84 years [47]), these findings support the hypothesis that collateral rarefaction, which becomes evident in late middle-age in mice [17], is caused by proliferative senescence of collateral ECs and SMCs due to a lifetime elevation in proliferation in excess of apoptosis that is caused by the disturbed hemodynamic conditions in which collaterals reside.

**Figure 8.** Tortuosity increases progressively with time after formation of collaterals. Images at right, ephrin-B2LacZ reporter mouse (construction of mutant described in [6]) showing embryonic collaterals. Bar graph, tortuosity index = l/L (axial length of collateral ÷ scalar length connecting collateral endpoints). E, embryonic day, P, post-natal day. Data for last four bars from Faber et al. [17] who also showed that collateral diameter and number begin to decline at or after 16 months of age. Number of mice (C57BL/6, B6) for bars 1–7: 8,9,8,9,10,7,8. For each mouse tortuosity, was determined for all MCA–ACA collaterals and averaged. ANOVA followed by 1-sided Bonferroni t-tests, \*, \*\*, \*\*\*, *p* < 0.05, 0.01, 0.001, respectively.

**Figure 9.** Persistence/maintenance versus rarefaction/pruning of collaterals "hangs in the balance". Proposed model whereby collateral (COL) mural cells reside in an environment of low/disturbed shear stress, high circumferential wall stress, and low blood oxygen content. This favors a pro-inflammatory, pro-proliferative, pro-apoptotic, and accelerated aging EC phenotype, leading to loss of collateral number and diameter (rarefaction). Compared to distal arterioles, collaterals have specializations and differential gene expression (left box) that provide adaptations that mitigate against factors that promote collateral rarefaction (right box). Vascular risk factors, e.g., aging, hypertension, EC dysfunction, and oxidative stress, disturb the balance. Collaterals are more sensitive than other vessels to these environmental risk factors, like "canaries in a mine-shaft".
