Deep tissue analysis of distal aqueous drainage structures and contractile features

Outflow resistance in the aqueous drainage tract distal to trabecular meshwork is potentially an important determinant of intraocular pressure and success of trabecular bypass glaucoma surgeries. It is unclear how distal resistance is modulated. We sought to establish: (a) multimodal 2-photon deep tissue imaging and 3-dimensional analysis of the distal aqueous drainage tract (DT) in transgenic mice in vivo and ex vivo; (b) criteria for distinguishing the DT from blood and lymphatic vessels; and (c) presence of a DT wall organization capable of contractility. DT lumen appeared as scleral collagen second harmonic generation signal voids that could be traced back to Schlemm’s canal. DT endothelium was Prox1-positive, CD31-positive and LYVE-1-negative, bearing a different molecular signature from blood and true lymphatic vessels. DT walls showed prominent filamentous actin (F-actin) labeling reflecting cells in a contracted state. F-actin co-localized with mesenchymal smooth muscle epitopes of alpha-smooth muscle actin, caldesmon and calponin, which localized adjacent and external to the endothelium. Our findings support a DT wall organization resembling that of blood vessels. This reflects a capacity to contract and support dynamic alteration of DT caliber and resistance analogous to the role of blood vessel tone in regulating blood flow.

Glaucoma is the leading cause of irreversible blindness worldwide and intraocular pressure (IOP) is a major risk factor for the disease. The trabecular meshwork (TM) is a key resistance point in the aqueous humor drainage tract and important determinant of IOP. But drainage pathways downstream to the TM may be nearly as important as up to half of total outflow resistance has been attributed to the intrascleral distal aqueous drainage tract (DT) [1][2][3][4][5][6][7][8][9][10] . The DT comprises Schlemm's canal (SC), collector channels (CC), intrascleral plexus (ISP) and aqueous veins connecting to episcleral veins.
The clinical importance of the DT has been brought into fresh focus by newer glaucoma surgeries that bypass TM resistance by incising the TM or inserting drainage stents through it. Following these procedures, IOP may be lowered but generally not to the anticipated level of episcleral venous pressure (EVP). This raises a suspicion that significant outflow resistance must remain in the DT after the TM is bypassed.
The extent to which the DT is able to modulate outflow resistance is unknown, although a capacity to contract has been suggested 8,11,12 . If so it may be that the DT has a contractile apparatus analogous to blood vessels lying downstream from it. Capacity of blood vessels to contract permits alteration of vessel caliber and modulation of flow resistance. Perhaps a similar mechanism is relevant to the DT apparatus lying upstream and in continuity with the episcleral veins. If such a mechanism applies to distal aqueous drainage it would suggest that resistance here is neither static nor passive, but rather that it is dynamic and regulated.
We postulate that the walls of the DT lying beyond the TM incorporate contractile cells with a smooth muscle identity in an organization mimicking conventional blood vessels. In blood vessels, smooth muscle forms a middle coat (tunica media) of the wall. It is sandwiched between an inner endothelial layer next to the lumen and a strong outer adventitial coat of collagen structural extracellular matrix (ECM). Episcleral veins that are conventional blood vessels into which the DT empties would have such an organization. More proximally, ciliary muscle (CM) that participates in uveoscleral aqueous drainage is itself smooth muscle.
We conducted tissue-based studies to characterize the fine structure and aspects of the molecular identity of the DT and explore whether it has contractile features. We studied this in mice as its aqueous drainage tract is similar to that of primates in structure and function [13][14][15][16][17][18][19] . We used transgenic fluorescent reporter mice expressing specific endothelial tags to improve identification of the DT. The DT has a unique partial lymphatic identity [20][21][22][23] , with endothelium here expressing Prospero homeobox protein 1 (Prox1) but not lymphatic vessel endothelial hyaluronal receptor-1 (LYVE-1) or other lymphatic endothelial markers typically expressed in true lymphatic vessels 22,[24][25][26] . While blood vessel endothelium and DT endothelium share in common expression of typical endothelial markers such as vascular endothelial growth factor receptor 2 (VEGFR2), platelet endothelial cell adhesion molecule 1 (PECAM-1 or CD31) and vascular endothelial cadherin (VE-cadherin), blood vessels do not express lymphatic endothelial markers such as Prox1 [27][28][29][30] .  We used trans-scleral multimodal 2-photon (2 P) imaging combining 2 P excitation fluorescence (TPEF) and second harmonic generation (SHG) [31][32][33][34][35][36][37][38][39][40][41][42][43][44][45] in vivo and ex vivo. This approach recognizes the DT structural and functional continuity with the systemic circulation and importance of maintaining this context. It was necessary to confirm the structures under study were DT channels rather than blood or lymph-carrying vessels, which also Reconstruction was based on 2-photon serial optical sections, merging signals from scleral collagen SHG (cyan; a), SHG signal voids (b) and siGLO-related intracellular fluorescence (green; c) in eyes from Balb/c (a-c) and VE-cadherin-td-Tomato (d-l) mice. siGLO was taken up by trabecular meshwork (TM) and intrascleral plexus (ISP) endothelium after mouse intracameral perfusion. Isosurface-mapped negative imprints of SHG signal voids were used to reconstruct collector channels (CC) and ISPs and assigned as yellow. Schlemm's canal (SC) was a slit-like opening between scleral SHG and TM siGLO signals. Continuation of SC openings into CCs (yellow) were seen. (a) Orthogonal reconstruction of SHG signal voids showing lumen of ISP and CC connecting to SC. CC ostium (CCO) and CC (arrow) connect SC to an ISP (arrowhead). (b) Sclera (SHG) was rendered transparent to reveal the relationships of ISPs with CC (yellow), collector channel openings (CCO; dashed circle), SC (gap) and TM (green) in 3D space. (c) The 3D model is rotated, showing a trans-scleral view from the ocular surface of interconnected ISPs, CC, CCO (dashed circle), SC (gap) and TM (green). (d-i) Isosurface-mapped, 3D-reconstructed ISPs derived from (i) negative imprints of SHG signal voids (d-f); yellow; representing ISP lumen) and (ii) VE-cadherinexpressing endothelium (g-i; red) of a VE-cadherin-td-Tomato mouse. The same branching ISPs were rotated to provide different views within sclera (cyan; rendered transparent). (j-l) merge of (d-f) (lumen) and (g-i) (endothelium) showing endothelium wrapped around ISP lumen. Arrows: ISP connecting to CC (f). traverse the limbus. This led us to establish criteria to confirm that DT structures (a) were continuous with SC and (b) showed the unique endothelial molecular signature expected of the system.

Results
Tissue-based identification of distal aqueous drainage tract structures. We first identified intrascleral DT structures and confirmed they were traceable to SC. Mouse sclera is thin and its DT apparatus is near the ocular surface, making it amenable to trans-scleral visualization by deep tissue imaging techniques. We used TPEF to view autofluorescence and transgenic reporter fluorescence, and SHG, a label-free 2 P technique to view structural collagen 32,37,41,45 . Imaging of hybrid cell membrane-localized td-Tomato (mTomato or mT) and cytosolic green fluorescent protein (GFP) co-expressing (mT/GFP) transgenic reporter mice revealed an interconnected network of intrascleral channels. The channels were evident as voids in the scleral collagen SHG signal deep to the ocular surface, as shown in Fig. 1 (in vivo) and 2 (ex vivo). The channel voids of the DT could be seen extending distally from SC toward the ocular surface (Figs 1a and 2A). Thus it was possible to trace DT channels nearer the ocular surface such as ISPs and CCs back to SC in 2 P serial image sections.
The hybrid transgenic mosaic expression profile of GFP (cytosolic) and td-Tomato (membrane) indicated the location of cells within the collagenous structural ECM and walls of channels and vessels in the sclera. The collagenous structural ECM was revealed by SHG to an extent that wavy collagen bundles could be resolved (e.g., Fig. 2B-E). SHG signal voids indicating absence of structural collagen correlated with lumen and cell complexes in the walls of intrascleral channels and vessels. The location of cell complexes at the edge of SHG signal voids (Fig. 1b) was revealed by GFP (Fig. 2C) and/or td-Tomato (Fig. 2D) mosaic fluorescence typical of the GFP/mT hybrid mice we used 46 . Thus simultaneous visualization of lumen SHG signal voids and wall cell-based fluorescence allowed delineation of the configuration of intrascleral channels and vessels ( Fig. 2A-I).
Branching ISPs in optical sections were orientated in a plane parallel to the ocular surface ( Fig. 2B-C). CCs, having a more radial orientation to the ocular surface, appeared in cross-section as rounder structures (Fig. 2D,E). The CCs were connections between ISPs (Fig. 2F,G) and SC (via collector channel openings [CCO]; Fig. 2H,I).
By confirming that intrascleral channels had their origins in SC we were able to verify they were a part of the distal aqueous drainage tract. Isosurface mapping and 3D reconstruction of distal aqueous drainage tract. Having verified that intrascleral structures under study were part of the DT, we used their luminal SHG signal voids to map and 3-dimensionally reconstruct the intrascleral course of the tract using software-based techniques. This would permit better characterization of the 3-dimensionally complex DT and its relationships with endothelial and contractile markers.
siGLO delivered intracamerally was used to localize cells along the DT of wild-type Balb/c and C57BL/6 mouse eyes. siGLO was taken up primarily by TM cells and ISP endothelium and used to map the location of TM relative to the adjacent sclera (SHG; cyan) and distal tract ( Fig. 3a showing SHG voids and ISP (green); yellow in Fig. 3b-c). With the scleral SHG signal rendered transparent (Fig. 3b,c), the complex 3D configuration of the DT and its relationships with surrounding sclera and proximal TM were evident. Reconstructions showed ISP branching, interconnection and coursing at different scleral depths, as was suggested by Fig. 2B and F. The trans-scleral vertical connection of ISPs, CC and SC in the trabecular outflow pathway was also evident.
SHG signal voids of the DT were further investigated for the presence of VE-cadherin using TPEF imaging of VE-cadherin-td-Tomato expressing mouse eyes ( Fig. 3d-i). VE-cadherin is an archetypal endothelial marker and its expression signified presence of endothelium ( Fig. 3g-i). SHG signal voids ( Fig. 3d-f) represented mainly lumen. By rendering the scleral SHG signal transparent, branching ISPs and their relationship with a CC could be viewed from different angles in 3D space. VE-cadherin-positive endothelium was arranged around a lumen (e.g., Fig. 3i) corresponding to SHG signal voids (Fig. 3f). There was some overlap between the endothelial td-Tomato signal and SHG signal void, however (Fig. 3l), as the SHG signal void also represented non-collagenous elements (e.g., cells, ECM) in the ISP wall. These findings provided validation for our use of SHG signal voids as surrogates to locate structures of the DT.
Characterizing endothelium of the distal aqueous drainage tract in situ. We further characterized the DT by profiling endothelial markers to verify the presence of a unique endothelial molecular signature expected of the system. Prox1 is strongly expressed in the endothelium of SC 24,25,28,[47][48][49] . Prox1 is also expressed in the DT distal to SC 42 but past reports define this less clearly. We sought to localize regions of Prox1-positive endothelium in the DT of Prox1-GFP reporter mice by TPEF (Figs 4 and 5). Prominent Prox1 expression was seen in the SC endothelium ( Fig. 4a,b, green). Here, Prox1 co-localized (Fig. 4d, yellow) with VEGFR2 ( Fig. 4c, red), an endothelial marker. Prox1 expression was more prominent in the inner wall (IWSC) than in the outer wall (OWSC) of SC (Fig. 4b).
VEGFR2 but not Prox1-GFP expression was seen in blood vessels such as episcleral veins or ciliary body vasculature (Fig. 4d).
True lymphatic vessels expressed Prox1-GFP as well (Fig. 4e). In Prox1-GFP mice, focal regions of hyperfluorescence with regular periodicity representing Prox1-GFP expression within lymphatic valves were seen in lymphatic vessels but not in SC 25 . While Prox1-GFP was expressed in the endothelia of both SC and lymphatic vessels (Fig. 4e,f), only true lymphatic vessels expressed LYVE-1 (Fig. 4e). Blood vessels expressed CD31 but not Prox1 (red; Fig. 4g). CD31 expression was very low and appeared almost imperceptible in lymphatic vessels ( Fig. 4g) but it was clearly expressed in SC (Fig. 4g) where it co-localized with Prox1 (Fig. 4h).
A fuller characterization and comparison of lymphatic vessels and DT channels was performed (Fig. 5). True lymphatic vessels at the limbus were identified by anti-LYVE-1 antibody labeling wherein the positive LYVE-1 signal co-localized with Prox1-GFP. Unlike lymphatic vessels, Prox1 co-localization with LYVE-1 was not seen Endothelial Prox1-GFP expression was analyzed with reference to scleral collagen SHG signal voids and found to be present in the DT from SC to ISP (Fig. 6). In the inner sclera, isosurface mapping localized Prox1-GFP expression to a discrete broad, shallow trench in the scleral collagen SHG signal contour corresponding to SC ( Fig. 6a-d). A posterior ridge of collagenous condensation bordering SC corresponded to the mouse equivalent of a scleral spur. Similar to SC, Prox1-GFP expression localized to the region of scleral SHG signal voids corresponding to the location of CC and ISP, as revealed by direct visualization (Fig. 6f,j-l,m-o), isosurface mapping (Figs 6e,h,i) and 3D reconstructions (Fig. 6g).

Contractile cells with smooth muscle identity in distal aqueous drainage tract walls.
To explore the possibility that the DT is contractile, we sought to locate contractile cells that may be associated with the tract and determine whether they possessed a smooth muscle identity. A similar organization is seen in blood vessels in which smooth muscle contractility plays important roles in modulating caliber and flow resistance.
Filamentous actin (F-actin), the polymerized, contractile form of actin, was labeled in situ to locate tissues in the DT that may be contractile. Fluorescence dye-conjugated phalloidin that allows stabilization and preferential detection of contractile states and stress fibers was used for labeling 44,50,51 .
Phalloidin-Alexa-568 labeling of C57BL/6 mouse eyes ex vivo showed prominent F-actin in the walls of ISP, CC and SC (Figs 7 and 8) in a pattern reflecting a multilayered cell organization (Fig. 7j,k). The wall distribution of F-actin labeling was similar to that seen of m-Tomato expression (Fig. 2), with both yielding a brighter and broader signal band than that expected of endothelium based on the diaphanous Prox1-GFP signal ( Fig. 6j-k, m-n; arrowhead). F-actin-positive labeling was continuous throughout the walls of ISP, CC and SC, as seen in serial optical sections (Figs 7b-i; 8a-f) and orthogonal reconstructions (Figs 7a; 8f,g).
We validated our in situ F-actin findings obtained by en face TPEF by comparing them with phalloidin-Alexa-568 labeling in immunohistochemistry (IHC) of radial histological sections of C57BL/6 mouse eyes. IHC findings mirrored those of TPEF wherein prominent F-actin labeling was seen in ISP walls ( Fig. 9b,h,n). Furthermore, the pattern of F-actin labeling in ISP walls was similar to that seen in the walls of choroidal arteries (Fig. 9e,k,q).
We postulated that the positive F-actin labeling in ISP walls corresponded in large part to the location of a smooth muscle coat as would typically be seen in blood vessels such as arteries. The mouse eye tissues were antibody-labeled for classic smooth muscle epitopes of alpha smooth muscle actin (ASMA), calponin and caldesmon. These proteins are involved in smooth muscle contractility and expressed in the smooth muscle of CM 16 and arterial walls [52][53][54][55] . IHC analysis showed that all three smooth muscle proteins localized to ISP walls (Fig. 9a,g,m) in a profile similar to that seen in choroidal arterial walls (Fig. 9d,j,p). All three smooth muscle epitopes of ASMA, calponin and caldesmon co-localized with F-actin in ISP walls (Fig. 9c,i,o). The pattern of smooth muscle epitope co-localization with F-actin mirrored that seen in choroidal artery walls (Fig. 9f,l,r).
We explored further the idea that cells with a smooth muscle identity form a layer external to endothelium in the distal aqueous drainage tract. High magnification TPEF analysis of C57BL/6 mouse eyes (Fig. 10) showed cells with flat nuclei (Hoechst 33342-positive) associated with thin F-actin-positive cytoplasm in an endothelial monolayer bordering the lumen of ISP (Fig. 10a). Immediately external to the endothelium were cells with plump nuclei and more diffuse F-actin-positive cytoplasm in a multilayer arrangement (Figs 7j,k; 10a). Location of the endothelial layer was consistent with a Prox1-positive layer bordering ISP and CC lumen of the Prox1-GFP mouse eyes (Fig. 10b-d). The plump nuclei in a multilayer arrangement immediately external to the endothelium (Fig. 10a) correlated with ASMA-positive labeling in a layer immediately external to the Prox1-positive region (Fig. 10b-d). The diameter of ISP lumen had a range of 9-15 μm (Fig. 10a).
Profiling contractile markers in proximal and distal aqueous drainage tract. We then compared the DT profile of contractile markers with that of the more proximal aqueous drainage tissues of TM (conventional) and CM (unconventional) in which contractility is of functional importance.
Balb/c albino mice lacking pigment in the otherwise heavily pigmented TM and ciliary body were used to optimize imaging of the deeper proximal aqueous drainage structures. Scleral autofluorescence (AF), having its source in structural collagen and elastin 35,39,43 , was detected by trans-scleral en face TPEF imaging (Fig. 11). The AF signal revealed similar scleral features to those seen with SHG: AF signal voids revealed a pattern of ISP connecting with CC and SC (Fig. 11b,f,j,n,r vs. Figs 1b, 2, 7, 8, 10).; and AF signals revealed septae within SC so that the SC lumen appeared loculated (Fig. 11j,n). Serial optical sections showing F-actin-positive signals (Fig. 11a,e,i,m,q) and merged F-actin and AF signals (Fig. 11c,g,k,o,s) revealed F-actin-positive CM fiber bundle extensions intermingling with F-actin-positive cells in the TM/IWSC region overlying SC.
In the TM (just deep to SC; Fig. 11m-t), F-actin-positive labeling and AF signals had a similar pattern. This indicated coincidence of F-actin-positive cells (Fig. 11s) with trabecular-like beams (Fig. 11r) surrounding tissue pores (AF signal voids; Fig. 11r-s). This beam-pore arrangement was reminiscent of the human TM.
The notion that F-actin tissue distribution is similar in mouse and human TM was explored further (Fig. 12). Similar to human TM, gaps between F-actin-positive labeling in mouse TM increased in size with tissue depth (i.e., further internal from SC; Fig. 12a-c) until a transitional region consistent with CM was reached just deep to the TM (at tissue depths seen in Fig. 12d-f).  We explored the possibility that F-actin-positive cells in mouse TM and CM expressed classic smooth muscle epitopes of ASMA, calponin and caldesmon, which if present would indicate a smooth muscle identity ( Fig. 13; Balb/c) 16 . In histological sections, the mouse TM is a cellular layer just adjacent and internal to SC, while the CM lies adjacent and internal to TM as a muscle layer sandwiched between the TM and pigmented ciliary body 16 . Our IHC analysis showed positive labeling of ASMA, caldesmon and calponin in the CM but not TM. ASMA, caldesmon and calponin labeling also co-localized with F-actin in the CM but not TM. The CM labeling profile was reproducible for all smooth muscle markers tested (i.e., ASMA, caldesmon and calponin; Fig. 12b,c) and mirrored the labeling profile and pattern of co-localization with F-actin for the same markers in the DT walls and choroidal arteries (Fig. 9). Neither smooth muscle markers nor their co-localization with F-actin was seen in the TM, a contractile aqueous drainage tissue that is not considered to be smooth muscle.

Discussion
We have assembled tools for tissue-based and in vivo studies of the mouse to address goals to: (1) identify and verify the DT in deep tissue imaging; (2) digitally map the 3-dimensionally complex DT and its relationships; (3) better characterize the unique endothelial molecular signature of the DT relative to lymphatic and blood vessels; (4) explore the possibility that the DT has capacity to contract and carries a smooth muscle identity mimicking that of blood vessels; and (5) compare the DT profile of contractile markers with that of proximal drainage tissues of the TM and CM. We exploited transgenic reporter mice and multi-modality trans-scleral 2 P imaging incorporating TPEF and SHG. Endogenous and exogenous fluorophores were visualized at a subcellular-level in the DT. We were able to study the DT in a way that preserved its 3D organization and connection to the systemic vasculature, which is important functionally and could provide a context for better understanding its specific physiology and molecular regulation.
The first step we took to verify that structures under study were part of the DT was to trace them to their origin in SC. Lumen of the DT were evident as voids in the scleral SHG or AF signal. These voids representing lumen could be traced and mapped using software reconstruction techniques we have previously described 40,[43][44][45] . The primate SC connects with downstream vessels via CC at intervals along the length of SC. It has been assumed that mice also have such an organization but this had not directly been shown prior to this study. Downstream from CC, ISP mapped as a large interconnected endothelial-lined collector system within the mid to superficial sclera. ISP connected via CC to its deep scleral origins in SC. CC varied in shape, size and orientation in their intrascleral course between ISP and SC.
A second step in identifying the DT was to detect endothelial expression of Prox1. Prox1 co-localization with LYVE-1 (for true lymphatics) and/or global endothelial markers (e.g., CD31) reveals differential selectivity for true lymphatic vessels and blood vessels and was used to help distinguish the different vascular systems by their respective endothelial signatures at the limbus where imaging was performed. It was based on an understanding that Prox1 is expressed in both the DT and lymphatic vessels but not in the blood vessels; CD31 is strongly expressed in the DT and blood vessels but not in lymphatics; and LYVE-1 is expressed in lymphatics but not DT or blood vessels 26,27 . Our findings are summarized in Fig. 14.
Prox1-expressing endothelium bordered lumen of the DT 20-23 . In SC, Prox1 expression was more prominent in the IWSC than in the OWSC, as has been reported 24 . Prox1 endothelial expression was also present in CC and ISP, which has been suggested 25 but not widely recognized. Prox1-GFP expression in CC and ISP appeared as a diaphanous signal bordering the lumen (e.g., Fig. 6j,k,m,n) that could also appear patchy depending on the plane of capture (Fig. 5g,h). Perhaps this subtlety explains why endothelial Prox1 expression in more distal parts of the DT is not better recognized. Our 2 P imaging had deeper tissue penetration than conventional 1 P imaging that has previously been used to study this system. The 2 P imaging permitted more sensitive detection of fluorophores deep in tissue. Our finding of widespread Prox1 expression in the DT contributes to a fuller characterization of the system. It is also important because Prox1 has a putative role in mechanical biosensing 25 that is relevant to the regulation of IOP and glaucoma.
F-actin is the polymerized, contractile form of actin. F-actin-positive labeling was used to generally localize cells in a contracted state within the tissues. The profile of F-actin-positive labeling in the DT of WT mice (Fig. 7) was similar to that of the cellular membrane-targeted td-Tomato signal in mT/GFP mice (Fig. 2). Both yielded a broader and more prominent signal bordering the lumen compared with the diaphanous Prox1-GFP-positive endothelial signal in Prox1-GFP mice (Fig. 6m,n). This led us to postulate that the F-actin signal encompassed a further cellular complex external to the endothelial monolayer. Higher magnification images indicating a multi-layered configuration of F-actin-positive cells external to endothelium (Figs 7j,k, 10) supported this notion. That F-actin primarily co-localized with mesenchymal markers of ASMA, caldesmon and calponin (Fig. 9)which is not expected in true endothelium -was also in line with this.
We propose that the F-actin signal that co-localized with ASMA, caldesmon and calponin in DT walls originated in mural contractile cells with a smooth muscle identity. These contractile cells reside in a DT wall region analogous to the tunica media smooth muscle wall layer of blood vessels. A mesenchymal identity is not expected in true endothelium. That the labeling profiles of ASMA, caldesmon and calponin in DT walls were similar to those seen in arterial smooth muscle walls and that the proteins also co-localized with F-actin lends further credence to this idea. Furthermore, the smooth muscle marker labeling profile in the DT wall mirrored that of CM, which itself is smooth muscle, as illustrated in Fig. 14. ASMA association with human SC and CC has been reported before 11,12 .
(c,g,k,o,s) (red-green). Merged orthogonal reconstructions: (d,h,l,p,t) (red-green). Vertical white line in orthogonal image: z-location of optical section. Curved dashed lines: outline a lumen that continued from ISP to collector channel (CC; dashed circle) to CC opening (CCO) to SC.
We measured mouse CC ostia to have lumen diameters in the scale of 10 μm, which is comparable to lumen diameters of human metarterioles (5-10 μm) 56 . Human metarterioles have a three-layered stratification: (1) tunica intima featuring attenuated endothelial cells on a basement lamina with scarce elastic and collagen fibers; (2) tunica media of smooth muscle cell layers; and (3) tunica externa of collagen mixed with autonomic nerve fibers, fibroblasts, histiocytes, and mast cells. The mouse DT appears to share features of this wall organization. We postulate that the mouse analogue of the tunica externa is the sclera itself, or maybe a fine collagenous condensation of the sclera that was imperceptible by our SHG imaging. That the DT and arteries have a common wall design likely reflects a common need of different vascular systems to constrict to modulate drainage resistance and intermittently distend and recoil to accommodate pulsatile flow.
The extent to which the intrascleral DT is able to distend within the relatively rigid eye coat is unclear. If the DT has significant compliance it would perhaps align it more with a venous than arterial analogue, providing a parallel paradigm for conceptualizing the tract's function. Alternatively, as the DT connects to the episcleral veins, one could conceive of an episcleral venous compliance system directly downstream from the DT in which the DT serves primarily as a resistor. Compared with arteries, veins have a less prominent contractile layer and more prominent lumen, a design that supports compliance and responses to changes in hemodynamic pressure. Venous features are not necessarily static, however, as they may adapt over time to dynamic forces. A classic example is the clinical observation of arterial features developing in episcleral veins in response to prolonged elevated episcleral venous pressure -so called 'arterialization of veins' . These questions conceptualizing the DT as vascular analogues could stimulate new ways to understand IOP regulation and seem worthy of further attention.
We found that the mouse TM showed prominent F-actin labeling but lacked the smooth muscle phenotype seen in the DT, CM and choroidal artery walls. While the TM is regarded to be contractile 44,57,58 it seems to be a different type of contractile tissue to that of the CM 16 or DT. The TM typically does not exhibit smooth muscle features except perhaps in a context of presumed mesenchymal transition associated with ASMA upregulation and a heightened myofibroblastic contractile state 36,[59][60][61][62] . CM fibers inserting in the region of the TM could themselves contribute to contractile forces exerted on and within the TM. CM elements intermingling within the TM may explain why the TM, even in a physiological state, could appear to have histochemical mesenchymal features reminiscent of muscle. It may be that resident TM cells displaying non-muscle actomyosin contractility play roles in physically organizing the extracellular matrix to counter extraneous muscular and hydrodynamic forces exerted on the tissue. Somehow the system must find balance. This is necessary to promote outflow homeostasis by the conventional (TM and DT) and unconventional (CM) tracts, which together determine IOP 57,58 . By what means homeostasis is orchestrated by the different drainage tissue components and what are sources of homeostatic malfunction leading to disease remain important questions for future study. Our data indicates that the walls of DT channels extending from SC to ISP maintain certain common features across their different parts, namely Prox1-expressing endothelium adjacent to an external contractile cellular compartment with smooth muscle identity. Clues to this arrangement were evident in in vivo and ex vivo imaging, with the techniques cross-validating each other. The agreement suggests that 2 P imaging ex vivo may suffice to capture data representative of the in vivo context in appropriate cases. Ex vivo imaging is simpler to perform than in vivo imaging, which is subject to additional challenges such as motion artifact. The limitation of ex vivo imaging, however, is that the aqueous physiological outflow dynamic is absent.
We speculate that the DT is not merely a system of passive channels within the sclera. Rather it has capacity to contract, with potential to modulate caliber and outflow resistance in analogous fashion to the modulation of vascular resistance by blood vessel tone. The interplay of systemic vascular resistance and compliance plays critical roles in maintaining hemodynamic stability and it is not inconceivable that the DT plays comparable roles in IOP homeostasis. The existence of a pulsatile component to aqueous outflow under the influence of cardiovascular hemodynamics 63-68 further argues for a DT with capacity to dynamically modulate tone and flow resistance as blood vessels do. The tissue-based and live mouse approaches we have described provide useful options for dynamic in vivo studies that are appropriate next steps in characterizing the specific cell biology, physiology and pharmacology of the DT.
VE-cadherin-td-Tomato expression was induced by tamoxifen treatment over 48 h. Briefly, tamoxifen was dissolved in DMSO to a concentration of 20 mg/mL and further diluted 1:2 with sunflower oil. 1 mg of tamoxifen was injected intraperitoneally twice over a 24 h period.
Two-photon imaging of ex vivo eyes. Whole enucleated mouse eyes, either fresh or fixed with 4% PFA and labeled for F-actin were imaged by 2 P microscopy as previously described 45 (Leica TCS SP5 AOBS MP (Leica Microsystems, Heidelberg, Germany) or Zeiss 710NLO (Carl Zeiss AG, Oberkochen, Germany systems with multiphoton laser (Coherent, Santa Clara, CA, USA)).
Eyes were placed on a glass bottom dish (P35G-1.5-14-C; MatTek Corporation, Ashland, MA, USA), along with a drop of PBS, with the axis of the optic nerve and central cornea perpendicular to the direction of the beam. A 63X/1.3NA glycerol objective was used to focus an 850 nm wavelength laser to excite AF and Phalloidin-Alexa-568 fluorescence, as previously described 45 . The following multiphoton bandpass filters were used to collect endogenous or exogenously-labeled TPEF signals: 415-435 nm (for collagen SHG), 500-550 nm (for elastin AF and green fluorescence), and 590-680 nm (for red fluorescence). The location of SC and the TM (just deep to SC) was identified by the scleral collagen SHG void characteristic of SC lumen [34][35][36][37][43][44][45] . Some imaging was performed in Balb/c mice as their non-pigmented eyes allowed for relatively unimpeded trans-scleral views and higher quality images of deeper tissue regions such as ciliary muscle 16 as previously described 16 . Table 2 summarizes imaging modalities used.
Images (1024 × 1024 pixels and 16-bit grayscale resolution) were captured in LAS AF (Leica) and analyzed with Volocity 5.4.1 (PerkinElmer; Waltham, MA, USA), LAS AF Lite 2.2.1 (Leica), Imaris 7.3.0 (Bitplane; Zurich, Switzerland) or Image J (NIH; Bethesda, MD, USA), and edited for publication with Photoshop CS5 (Adobe), as previously described 45 . Two-photon imaging of live mice. An objective inverter (LSM Technologies) 73 was used to image live mice following anesthesia as described previously 45 . Contact between the objective and the mouse eye was mediated with an optical coupling gel (0.5% Carbomer 940, 300 mM D-sorbitol, adjusted to pH 7.3 with triethanolamine; from Snowdrift farm, Tucson, AZ, USA).
A modified IX81 inverted 2 P confocal microscope (Olympus, Melville, NY, USA), coupled to a tunable Ti:Sapphire femtosecond laser (Chameleon Ultra II; Coherent) was used to image the live mouse eyes. The following bandpass filters were used to collect TPEF signals: 400-480 nm (for Hoechst fluorescence and collagen SHG), 505-560 nm (for FITC, Alexa-488, GFP/EGFP), and 590-650 nm (for Texas red and td-Tomato). All live mouse images and movies were acquired using a UPLSAPO 60X NA 1.2 water immersion objective in 2 P configuration 74 , as previously described 45 . Volumetric 3-dimensional (3D) tissue reconstructions. Imaris was used to generate "isosurface volumes" to recreate polygonal, volumetric renderings of tissue signals such as SHG, SHG signal voids, or endogenous and exogenously-labeled fluorescence markers (eg., AF, F-actin-phalloidin-Alexa568, Prox1-GFP, siGLO) as previously described 36,[43][44][45] and summarized in Table 2. Predetermined thresholds for size and fluorescence intensity were used to set minimum value requirements for volume rendering. Predetermined optimum threshold values 40,43 were voxels of 0.24 × 0.24 × 1 μm in the X, Y, Z axes respectively. Where fluorescence dynamic range (especially of deep structures) was narrow relative to the dynamic range of shallower regions, reconstructions were cropped to a maximum optical slice thickness of 25 μm in the z-axis prior to isosurface volume rendering.
Volumetric 3D reconstructions from signal voids. Imaris was used to reconstruct the SHG signal into "isosurface volumes" with the following parameters: 2 μm surface grain size; 1.803 μm largest sphere diameter; 3,600 minimum fluorescence intensity cutoff; and 10 μm minimum voxel size, as previously described 45 . The SHG isosurface volume was subtracted from a generic signal mask (all voxels of generic z-stack set to 20,000), leaving a negative SHG impression representing SC and ISP lumen. Isosurface volumes of SC lumen and ISP channel voids were reconstructed using the following parameters: 0.481 μm surface grain size; 1.803 μm largest sphere diameter; automatic (default) fluorescence intensity minimum cutoff value; and 10 μm minimum voxel size.
Intrascleral plexus lumen diameter measurements. ISP cross-sectional diameters were measured at intervals of 25 microns from the center of the CC ostium. Measurements were obtained at each interval for every 2D frame of a z-stack containing a portion of the ISP lumen. The largest diameter measured among all frames at that interval was selected to represent the diameter of the ISP at that interval.
Immunohistochemistry. 7 µm-thickness frozen sections of enucleated eyes were fixed, permeabilized, and blocked in 4% PFA and blocking solution (5% BSA/0.3% Triton X-100) for 1 h at room temperature (RT), then incubated overnight at 4 °C with combinations of primary antibodies (single, dual, or triple) in the blocking solution. The sections were further incubated for 1 h at RT with a combination of secondary antibodies (Alexa Fluor ®  488, 568, and 633-conjugated anti-rabbit, mouse, and/or goat antibodies) and then mounted using ProLong Gold Anti-fade reagent with DAPI. To visualize polymerized actin, secondary incubation of slides included Alexa-568-phalloidin along with the appropriate combination of secondary antibodies. Negative, non-specific labeling was established with normal IgG isotypes. Sections were analyzed with a Leica SP5 high-speed spectral confocal laser-scanning microscope or Zeiss LSM 710NLO confocal microscope. Immunofluorescence staining for single or double contractile markers was performed in randomly selected slides (4-5 slides per each eye) containing 4 sections per slide. Hematoxylin and Eosin (H&E) staining in paraffin sections of C57BL/6 mouse eyes was performed as described previously 16 . Whole-mount immunofluorescence staining. Enucleated eyes were fixed with 4% PFA/PBS at 4 °C for 2 h prior to microsurgically isolating the whole cornea (i.e., including the anterior segment) from the posterior segment and lens. Whole corneas (and attached anterior segments) were washed 3X with PBS for 5 min then incubated with rabbit polyclonal anti-LYVE-1. After several washes the samples were incubated for 4 h at RT with Cy5-conjugated anti-rabbit antibody. Specimens were mounted using VectaShield mounting medium with DAPI (Vector Labs, Burlingame, CA, USA). Whole mount immunofluorescence images were captured using a Leica M165 fluorescent stereomicroscope (Leica). Data Availability. The datasets generated dGlaucuring and/or analysed during the current study are available from the corresponding author on reasonable request.