An eye on the dog as the scientist's best friend for translational research in ophthalmology: Focus on the ocular surface

Preclinical animal studies provide valuable opportunities to better understand human diseases and contribute to major advances in medicine. This review provides a comprehensive overview of ocular parameters in humans and selected animals, with a focus on the ocular surface, detailing species differences in ocular surface anatomy, physiology, tear film dynamics and tear film composition. We describe major pitfalls that tremendously limit the translational potential of traditional laboratory animals (i.e., rabbits, mice, and rats) in ophthalmic research, and highlight the benefits of integrating companion dogs with clinical analogues to human diseases into preclinical pharmacology studies. This One Health approach can help accelerate and improve the framework in which ophthalmic research is translated to the human clinic. Studies can be conducted in canine subjects with naturally occurring or noninvasively induced ocular surface disorders (e.g., dry eye disease, conjunctivitis), reviewed herein, and tear fluid can be easily retrieved from canine eyes for various bioanalytical purposes. In this review, we discuss common tear collection methods, including capillary tubes and Schirmer tear strips, and provide guidelines for tear sampling and extraction to improve the reliability of analyte quantification (drugs, proteins, others).


| INTRODUCTION
Preclinical animal models provide critical information to better understand human diseases' characteristics, identify biomarkers, develop diagnostic tools and novel therapeutics. Rabbits and laboratory rodents (mice and rats) are widely used for ophthalmic research as they are economical and easy to handle 1 ; however, serious drawbacks limit the translational usefulness of data obtained in these species, notably due to the need to artificially induce pathology in these animals (e.g., through genetic manipulation or experimental surgery), as well as apparent differences in ocular anatomy and physiology compared to humans. For instance, precorneal residence time of topically applied solutions is much prolonged in rabbits owing to their low blink rate, resulting in threefold overestimation of ocular drug exposure if findings were directly extrapolated from rabbits to humans. 2 Another example is topical nepafenac, a potent nonsteroidal antiinflammatory drug that reaches therapeutic levels in the posterior segment of mice (owing to their thin cornea and small globe size), inhibiting choroidal neovascularization by decreasing production of vascular endothelial growth factor (VEGF) 3 -in contrast, humans require intravitreal injections of anti-VEGF compounds to achieve the same outcome.
Multiple other examples exist in the scientific literature, together participating to the unacceptably low success rate of ophthalmic clinical trials to date, and resulting in substantial economic loss and burden for scientists, consumers, and society overall. 4 In fact, the main cause for clinical trial failure is either lack of safety or efficacy, 5 two components that are supposedly "validated" in initial preclinical animal studies.
Under the umbrella of the One Health Initiative, a growing number of investigations have integrated companion animals into preclinical studies to complement and expand the knowledge gained from studies in other animal models, accelerate and improve the framework in which research is translated to the human clinic, and ultimately generate discoveries that will benefit the health of humans and animals. 6 Over the last few years, several review articles have highlighted the benefits of using dogs for translational research in oncology, 7 neurology, 8 and other biomedical fields, 6 yet such information is not available in ophthalmology.
The present review provides a comprehensive comparison of key ocular parameters in humans, dogs, and traditional laboratory animals (i.e., rabbits, mice, and rats), highlighting selected strengths and important pitfalls that must be addressed when ocular research is conducted in animal models. This review is focusing on the ocular surface, a critical element of vision that includes the secreted tear film, lacrimal gland(s), eyelids, meibomian glands, cornea, conjunctiva, sclera, and nasolacrimal drainage apparatus. The ocular surface dictates the bioavailability of medications administered topically to the eye, 9 and is a common site of pathology in both human and veterinary medicine. Methods of tear fluid collection for bioanalytical purposes are also being discussed, with special consideration on the safety and efficiency of the collection technique at hand. Lastly, this review highlights on spontaneous and experimental ocular surface disorders in dogs, providing a tool for researchers to better model disease pathophysiology in clinical patients suffering from ocular surface disorders.

| Anatomy
The anatomy of the ocular surface is depicted in Figure 1 for dogs, and its parameters are being summarized in Table 1 for all species discussed in this review (i.e., humans, dogs, rabbits, mice, and rats).

| Lacrimal glands
Four types of lacrimal glands can be distinguished in mammals: (i) the orbital lacrimal gland (glandulae lacrimales superior), located in the dorsolateral orbit just caudal to the orbital rim, with secretory ducts that intraorbital and ventromedial to the globe (rabbits) or extraorbital and caudal to the globe (rodents), with a single secretory duct that opens into the lower conjunctival fornix; and (iv) the Harderian gland, or Harder's gland, extending from the base of the third eyelid into the caudal orbit, with secretory ducts opening at the nictitating membrane (rabbits and rodents). 11,13 The histomorphology of lacrimal glands varies with age and sex of the individual. 87 In dogs, an orbital lacrimal gland and gland of the third eyelid contribute to 60%-70% and 30%-40% of the overall tear secretion, respectively. 88 The morphological and histological features of the canine glands resemble the human lacrimal gland including distinct lobules and acini that provide serous and mucous secretions, as well as intralobular ducts that drain into small excretory tubules. 12,89 Likewise, an Harderian gland is not present in the canine or human orbit. 12,19,89 However, two notable differences exist between species: (i) the combined volume of the two canine glands is smaller than the main lacrimal gland in humans (0.24 vs. 0.60 cm 2 ) 16,17 ; and (ii) the accessory lacrimal glands of Krause and Wolfring are absent in dogs (or not yet reported), presumably being consolidated through evolution into the single gland of the third eyelid. 90 These accessory glands account for 10% of the total lacrimal secretory mass in humans but their contribution to the overall tear secretion is negligible (1-2%). 10,91 In rabbits, the histoarchitecture of the main lacrimal gland is comparable to humans with loosely packed acini and round/oval lumen; in contrast, mice and rats have densely packed acini with small pleiomorphic lumen and numerous intercellular tight junctions. 14 Like humans, rabbits also possess accessory lacrimal glands of Wolfring in the tarsal portion of the palpebral conjunctiva. 15 However, the Harderian gland present in rabbits and rodents is a unique anatomical feature that has important repercussions for comparative studies; in fact, the gland's lipid secretions in the tear film have profound effects on the ocular surface physiology (e.g., tear composition, tear film dynamics, blink rate) and pharmacology of topically applied medications (see Sections 2.2 and 2.3).

Highest in fornix
Low in bulbar conjunctiva [40][41][42][43] Low in bulbar conjunctiva 39,44 Relatively dense in bulbar conjunctiva [45][46][47] Low in bulbar conjunctiva 48 The morphology of the canine lacrimal drainage system is remarkably similar to that of humans, except for a longer nasolacrimal duct (notably in long-nosed dogs), and the presence of accessory duct openings into the nasal cavity. 92 In both species, tear drainage begins with the lower and upper nasolacrimal puncta and canaliculi in the medial canthus, joining into a lacrimal sac in the bony lacrimal fossa, and extending into the nasolacrimal duct that runs through an osseous channel towards the nasal cavity. 20 Species similarities are also evident on a microscopic level, including an epithelial lining with microvilli and mucin-secreting goblet cells, subepithelial seromucous glands, and mucosal-associated lymphoid tissue. 92 In contrast, the nasolacrimal apparatus of rabbits has distinct differences compared to humans. Rabbits only have a single nasolacrimal punctum/canaliculus (medial lower eyelid) and the nasolacrimal duct has two very distinct flexures due to the ventral deflection of the snout, a unique feature that results in a convoluted path for tear drainage. 20,93 The fetal development of the rabbit's nasolacrimal apparatus is also unique in mammals, more closely resembling reptiles versus humans. 21 At an ultrastructural level, the epithelium lining the duct is double-layered (similar to humans) but there are no goblet cells or subepithelial seromucous glands. 94 Nonetheless, the use of the rabbit is still recommended as a practical model to characterize the nasolacrimal apparatus, 95 albeit this choice is described as "less than ideal" by the authors. Mice and rats have a well-developed nasolacrimal apparatus that shares similar ontogenetic origin to humans, 21 although the histological features are different. The duct lining is covered by a multilayered stratified squamous epithelium with goblet cells but without subepithelial seromucous glands. 20

| Third eyelid
The nictitating membrane (third eyelid) is a large fold of the conjunctiva that protrudes from the medial canthus over the anterior surface of the globe in many animals, including dogs, rabbits, and rodents. The counterpart in humans is the plica semilunaris, a vestigial remnant in the form of a crescent-like conjunctival fold in the medial canthus. 11,40 Despite gross differences, both structures have important physio-morphological similarities such as the presence of goblet cells and lymphoid follicles, contributing to the lubrication and immune protection of the ocular surface. 40 Nonetheless, the presence of a third eyelid should be considered in comparative studies as it could impact ocular examinations (e.g., third eyelid protrusion from ocular irritation) or ocular drug delivery (e.g., altered retention time of a contact lens), 96  The main anatomical difference is the tarsal plate, which is comprised of dense fibrous tissue and cartilage-specific components in humans 102 -providing a rigid internal support to the eyelids-compared to a much thinner and poorly developed fibrous tissue in dogs. 22 Also, the interpalpebral fissure area is approximately 20% larger in dogs (2.2 vs. 1.8 cm 2 ), 26,31 although the measurements of the palpebral fissure width depend on the dog's size and body weight. 26 The palpebral opening in the rabbit is relatively small (10-16 mm), 22,23,28 albeit much larger than mice (3.7-5 mm) 29 and rats (6-9 mm), 30,56 with a shorter and thicker upper eyelid compared to the inferior palpebrae; consequently, the interpalpebral fissure area is 20% smaller in rabbits than in man (1.44 vs. 1.8 cm 2 ). 31 The meibomian gland ducts and acini are also larger in rabbits than mice and rats, 77 but the overall volume and distribution of meibomian glands is different than in humans: the total meibomian gland volume in the human (39.5 mm 3 ) is twice that of the rabbit (18.8 mm 3 ), with a larger volume in the upper eyelid (man) compared to similar volumes in the upper and lower eyelids (rabbit). 31

| Conjunctiva
The conjunctiva is a thin mucous membrane that serves important roles on the ocular surface including mucin secretion and immune surveillance. The anatomical subdivision of the conjunctiva is the same in humans, dogs, and common laboratory species (rabbits and rodents): the palpebral conjunctiva-lining the inside of the eyelids-reflects back at the level of the conjunctival fornix to form the bulbar conjunctiva, a region that covers the anterior portion of the sclera and attaches to the corneoscleral limbus. 19 However, the amount of bulbar conjunctiva exposed ("scleral show") is notably larger in humans compared to animals given differences in eyelid opening and/or corneal diameter. Another important species difference is the presence of a nictitating membrane in animals (but not man), as the third eyelid is covered by conjunctiva on its anterior and posterior surfaces. As such, animals have two conjunctival fornices in the inferonasal region-one on each side of the third eyelid-and the overall conjunctival surface is generally larger in animals compared to humans. In dogs, the conjunctival area is supposedly larger than in humans given the depth of the canine conjunctival fornices and the amount of conjunctiva covering the canine nictitating membrane, 76  Conjunctival goblet cells are distributed individually in humans, dogs and rabbits, in contrast to clustered organization in mice and rats. 38,39 The distribution of goblet cells is overall similar in dogs and humans, with high density in the canine third eyelid and human plica semilunaris, relatively high density in the conjunctival fornices and palpebral conjunctiva, and lower density in the bulbar conjunctiva. [39][40][41][42][43][44] In rabbits, the highest density is noted at the lid margin of both upper and lower palpebral conjunctivae, 45,46 while the density in the bulbar conjunctiva is generally higher than in humans (399-1576 vs. 7-979 cells/mm 2 ). 43,47 In addition to mucin-secreting goblet cells, the conjunctiva also contains an organized immune network termed conjunctiva-associated lymphoid tissue (CALT), a structure that plays a key role in protecting the ocular surface by initiating and regulating immune responses. 78 The presence of lymphoid follicles was confirmed in the conjunctiva of most mammals studied by Chodosh and colleagues-including humans, dogs, rabbitswith the exception of mice and rats, 49 although a later report detected lymphoid tissue in the nictitating membrane of BALB/c mice. 50 At an ultrastructural level, specialized M cells are present in the epithelium overlying the conjunctival follicles in dogs 103 and rabbits, 79 similar to humans. 86

| Cornea
The anatomy of the cornea is unique to each species with important differences in corneal dimensions and ultrastructural features (e.g., thickness, collagen arrangement, nerve supply). 11,22,24,37,60,65,104,105 First, the cornea is generally larger in dogs and rabbits compared to humans, while the dimensions are much smaller in mice and rats. 23,[51][52][53][54][55][56][57][58] As such, the relative amount of cornea and conjunctiva exposed on the ocular surface varies among species, an anatomical fact that has important implications in ocular pharmacology and other research fields; for instance, the surface area ratio of conjunctiva to cornea is two times smaller in rabbits (8.6-8.9) than humans (17.2), a finding that could largely explain species differences in drug penetration into the anterior chamber. 37 Second, the corneal thickness varies among mammals and is generally correlated to the size of the animal. 60 From highest to lowest, the mean central corneal thickness is 497-594 µm in dogs, 60,61 505-563 µm in humans, 59 354-407 µm in rabbits, 23,54,59,62 159-170 µm in rats, 60,62 and 90-137 µm in mice. 55,60,62 The average canine cornea is only slightly thicker than in humans. In contrast, the thinner cornea in rabbits and rodents can limit the use of these laboratory species for selected experiments; for instance, cross-linking is discouraged in corneas thinner than 400 µm due to potential damage to the corneal endothelium or intraocular tissues. 106 On a structural level, the main layers of the cornea are the same in humans and animals (epithelium, stroma, Descemet membrane, and endothelium) with the notable exception of the Bowman's membrane. 11 membrane is present in nearly all primates (including humans) and selected animals (e.g., sheep, deer, and giraffe), 107 but is absent in dogs and common laboratory species. 60 63,109 and mice (13 layers, 37-46 µm). 55 The corneal stroma, comprising nearly 90% of the total corneal thickness in most mammals, is primarily composed of collagen fibrils arranged in lamellae. While extensive collagen intertwining is noted in the majority of the corneal stroma in humans, it is only present in the anterior most-aspect of the cornea in dogs and rabbits. 71,73 Differences in collagen intertwining, along with the absence of Bowman's membrane in laboratory species, explain the vast disparity in stiffness of the anterior stroma (16.2, 1.3, and 1.1 kPa) and posterior stroma (2.5, 0.5, and 0.4 kPa) in humans, dogs, and rabbits, respectively. 71,73 The elastic modulus of the cornea is reportedly higher in rodents, although the methodology used was different. 74,75 Corneal rigidity should be considered in comparative studies in which the biophysical attributes of the cornea are important (e.g., wound healing, keratoprosthesis). The corneal endothelium shares a similar morphological blueprint among species (single cell layer, honey-comb pattern), while the cellular density varies from 3233 cells/mm 2 in rabbits, 2875 cells/mm 2 in mice, 2818 cells/mm 2 in dogs, 2732 cells/mm 2 in humans, and 2242 cells/mm 2 in rats. 60,64 The mammalian cornea is the most densely innervated tissue in the body. Corneal nerves play important roles to maintain ocular surface health and homeostasis, including sensory functions (touch, pain, and temperature), release of trophic neuropeptides, maintenance of the limbal stem cell niche, and activation of brainstem circuits to promote reflex blinking and lacrimation. From highest to lowest, the sensitivity of the cornea to mechanical stimulus is as follows: humans (0.2-1.0 g/mm 2 ), rats (0.42-0.47 g/mm 2 ), mice (0.59 g/mm 2 ), dogs (2.16-2.9 g/mm 2 ), and rabbits (6.21-10 g/mm 2 ). 2,[66][67][68][69][70] The murine model is the most extensively studied of all laboratory species given gross similarities between mice and humans in corneal sensitivity and nerve architecture. 65,110 The canine model is also studied in detail given shared features with humans in several spontaneous diseases such as diabetes mellitus, herpetic keratitis, and nonhealing corneal ulcers [111][112][113][114] ; importantly, investigators should account for the canine breed selected for the experiment as corneal sensitivity depends on the dog's cephalic conformation, that is, dolichocephalic (long-headed with long nose, e.g., Greyhound), mesocephalic (medium-headed with medium nose, e.g., Beagle) or brachycephalic breeds (short-headed with short nose, e.g., Pug). 67 In regard to rabbits, two striking species differences exist: (i) Corneal sensitivity in rabbits is much lower than in humans, dogs and rodents 2,66 ; and (ii) Morphology of the rabbit subbasal plexus is unique, with nerve fibers sweeping horizontally across the corneal surface in a temporal-to-nasal direction compared to a typical whorl-like or spiraling pattern in other species. 65

| Sclera
Humans have a widely exposed white sclera, a feature that is unique when compared to other primate species. In contrast, the scleral exposure is minimal in dogs and routine laboratory species. The thickness of the sclera also differs among species: at the ocular surface (limbal sclera), recorded measurements vary from 0.8 mm in dogs, 84 0.5 mm in humans, 115 0.29 mm in rabbits, 116 0.1 mm or less in rats, 80 and 0.05-0.06 mm in mice. 117

| Tear film dynamics
Effective tear dynamics, combined with well-balanced composition of the tear film (See Section 2.3), are critical for the maintenance of ocular surface homeostasis and physiology. Tear fluid dynamics-or the balance between tear secretion, distribution, absorption, evaporation, and drainage-are closely regulated by the lacrimal functional unit.
The lacrimal functional unit is unique to each species (see aforementioned anatomical differences), comprised of secreting glands (orbital, accessory, third eyelid, Harder's, meibomian), eyelids, conjunctival goblet cells, corneoconjunctival surface, and their interconnecting innervation. 118 Key physiological parameters provide insight into the complex tear dynamics-highlighted in Figure 2 and Table 2-and are therefore important to account for in translational studies that involve the ocular surface: • Basal tear turnover rate: Tear turnover rate is considered a global measure of the tear dynamics and integrity of the lacrimal functional unit. 119,120 The basal tear turnover rate is reportedly 13.1-17.5 %/min in humans, 121,122 12.2 %/min in dogs, 123 6.2-7.1 %/min in rabbits 124 and 5.2 %/min in mice 125 ; no information was available in rats. In other words, it takes approximately the same time for the tear film to replenish in dogs and humans (~6-8 min) but the duration is longer in rabbits (~14-16 min) and mice (~20 min). The slow tear turnover of rabbits and rodents has important repercussions in translational research, including a longer precorneal retention time of instilled eyedrops, or exaggeration of ocular surface disease due to delayed clearance of inflammatory mediators from the tear film. 126 • Tear volume: The volume of tears on the ocular surface is highest in dogs (65.3 µl), 123 followed by humans (7-12.4 µl), 122,127 rabbits (1.9-7.5 µl), 124,128 rats (4.6 µl), 129 and mice (0.06-0.2 µl). 125,130 Canine tear volume depends on the subject's body weight but not the dog's cephalic conformation. 123 Differences in study methodology notwithstanding, the canine tear volume is approximately five to ninefold larger than in humans. This discrepancy can be partly explained by the additional secretory tissue in dogs (third eyelid gland) and the larger corneal surface to lubricate in dogs (1.2-2.1 vs. 1.04-1.3 cm 2 ). 37,52,53 The canine tear film may also be thicker than in humans (15.1 vs. 2.3-11.5 µm), although measurements of tear thickness were only obtained in six dogs 123 and the calculation of tear thickness is reportedly highly variable within and between species. 131 • Spontaneous blink rate: The blink action distributes fresh tears on the ocular surface in a uniform layer, promotes secretion of tears from the accessory tear glands, and pumps excess tears (or instilled drop) into the nasolacrimal drainage system. Spontaneous blinking is triggered by higher centers in response to corneo-conjunctival nerve stimulation, presumably due to changes in ocular surface temperature that result from thinning and evaporation of F I G U R E 2 Diagram depicting the complexity of tear film dynamics and ocular surface physiology. Secretion of tear components, distribution of tears through blinking, and elimination through nasolacrimal drainage and evaporation must be precisely regulated to maintain homeostasis. Drug kinetics following topical eyedrop administration are impacted by key parameters highlighted in yellow, each being unique in different species. Adapted with permission from Tsubota et al 144  unstable tear film. 132 The spontaneous blink rate is very similar between dogs (14.2 blinks/min) 133  • Reflex blinking (or lack thereof): In dogs, a blink occurs immediately after eyedrop administration and is responsible for removal of any excess solution onto the periocular skin and nasolacrimal drainage system. 26 The same is true in humans, in whom an instilled eyedrop is partially lost (20%-30%) due to reflex blinking and spillage onto the eyelids and eyelashes. 177 Blinking in response to eyedrop instillation is also reported in mice 178 and rats. 179 In contrast, rabbits rarely blink following eyedrop administration, or do so infrequently. In one study, rabbits did not blink for 20-30 min after instillation of an eyedrop, and this alone could result in overestimating ocular drug exposure by threefold if findings were to be extrapolated to humans. 2 • Reflex tear turnover rate: Eyedrop administration abruptly increases the volume of fluid in the conjunctival sac and ocular surface. The sudden disruption in homeostasis promotes a faster nasolacrimal drainage until baseline conditions return. This physiologic response is prominent in dogs (50%/min) 123 and humans (31.5-100%/min), 122,127 but is minimal in rabbits (6.1-6.9%/min). 124 In fact, the tear turnover rate in rabbits is mostly unchanged whether a small (1-5 µl) or large volume (25-50 µl) of eyedrop is instilled on the ocular surface, 124 a finding likely related to the poor corneal sensitivity and inexistent/minimal reflex blinking in this species. 2,66 No available report in mice or rats can be found in the literature.  180 Of note, the volumetric capacity of the canine eye is positively correlated with the length of the palpebral fissure, 26 and may be larger in breeds larger than Beagles (e.g., German Shepherd dogs). 180 The exact volumetric capacity of the eye is not reported in laboratory species, but is presumably around 10-25 µl in rabbits (based on drug quantification in tears at various instilled volumes), 124,146 ≤5 µl in mice 147,148 and ≤20 µl in rats. 149

| Tear film composition
The tear film is a complex biological fluid containing thousands of compounds of diverse structures and functions, including proteins, lipids and mucins, as well as minor constituents such as electrolytes, vitamins, and growth SEBBAG AND MOCHEL | 2577 factors. 101,181 The integrated interactions of these constituents are responsible for the promotion of a stable tear film and, ultimately, the homeostasis of the ocular surface. Species differences in tear film components are summarized in Table 3.

| Proteins
The total protein content is generally similar in dogs (5.2-14.6 mg/ml) 209 and humans (6.0-11.0 mg/ml), 182 although qualitative and quantitative differences exist. Specifically, the three major constituents of the human tear proteome (lactoferrin, lysozyme, and lipocalin) 182 are only detected at low levels in dogs, 183,184,187,[189][190][191] although the relative abundance of other common proteins (e.g., lacritin, secretory IgA, and serum albumin) is generally similar between the two species. Importantly, homologous proteins have been described in canine tears and may play similar functions to their human counterparts-for instance, transferrin is an iron-binding protein with similarities to lactoferrin, while major canine allergen is an abundant protein in canine tears with similarities to lipocalin. [189][190][191] From a qualitative aspect, a recent in-depth proteomic study showed that 25 out of 125 proteins detected in canine tears were common to humans. 190 In rabbits, Wei et al. found that the total protein content was twofold higher in rabbits compared with humans (20.6 vs. 9.4 mg/ml), although the number of different proteins detected in tear samples was lower in rabbits. 208 Other differences in tear proteins among species are summarized in Table 3.

| Mucins
Ocular mucins are large glycoproteins expressed by conjunctival goblet cells, the corneal epithelium and the lacrimal gland(s), playing important roles on the ocular surface in lubrication, wettability and barrier function. 210 The main secretory mucin, MUC5AC, is described at large levels on the ocular surface of humans and animals. 11,44,210 The expression of membrane-associated mucins, however, differs among species. In a recent study by Leonard et al., dogs were found to have a very similar pattern of mucin expression to that of humans and rhesus macaques, with MUC16 being the most abundant mucin transcript. 203 In contrast, the rabbit had a unique mucin expression pattern with all mucin transcripts expressed at relatively similar levels; as such, the authors concluded that the predictive value of the rabbit as a model in ocular surface studies should be called into question. 203 In another study, the majority of ocular mucins detected in dogs and rabbits were neutral fucosylated glycans, while the ones in humans were mainly negatively charged sialylated glycans 205 ; however, the experiment lysed the ocular surface epithelium and could not discriminate between mucins of differing origin.

| Lipids
In a comprehensive lipidomic study comparing the meibum collected in several species, Butovich et al. found that the highest degree of biochemical similarity with humans was observed in mice, closely followed by the dog. 141 An earlier study by Butovich et al. also reported the close resemblance of the tear lipid composition between dogs and humans. 206 In these three species (humans, dogs, and mice), the major lipid classes included wax esters, cholesterol esters, and o-acyl-ω-hydroxy fatty acids (OAHFA). In contrast, the major lipid classes in rabbit tears were DiHL esters (24,25-dihydroc--lanosterol esters), diacylated diols, and OAHFA, with low to trace amounts of wax and cholesterol esters. 141 Such discrepancy between rabbits and humans was confirmed in a separate study by Wei et al., who noted significant differences in the tear film concentrations of triglycerides (higher in rabbits), free cholesterol (lower in rabbits), phosphatidylcholine (higher in rabbits), and phosphatidylethanolamine (higher in rabbits). 208 T A B L E 3 Comparative composition of the major components in tear film between humans, dogs and common laboratory species used in preclinical ophthalmic research Low to moderate 183,184,190,191 Low 185,189,192 Low 183,193  Homologous proteins are reported in dogs (major canine allergen), [189][190][191] rabbits (lipophilin) 208 and rats (VEGr1). 189 c Triacylglycerol, squalene, ceramides, phospholipids and sphingomyelins.
Taken together, the authors of these two studies argued that the rabbit is too different to serve as a valid animal model for humans, at least from a biochemical standpoint.

| TEAR COLLECTION FOR BIOANALYTICAL PURPOSES
The tear film, a complex body fluid uniquely exposed to both internal and external environments, contains numerous endogenous and exogenous molecules (e.g., proteins, lipids, mucins, and xenobiotic) that can be assayed for clinical or research purposes. Topically and systemically administered drugs can be quantified in tear fluid to determine the clinical efficacy and dosing frequency from fitting of kinetic data. [211][212][213][214][215] Multiple "omics" approaches can also be utilized for analysis of the tear fluid including proteomics, 173,190,191,[216][217][218][219][220]

| Direct tear sampling
A microcapillary glass tube (1-10 µl) placed in contact with the inferior lacrimal lake is the most commonly reported technique to collect tear fluid. This method directly samples tear fluid by capillary action and is extensively described in humans, 182 243 and does not take into account possible losses (e.g., transfer and storage) or the need to repeat certain assays in duplicates.
Several strategies can be used to overcome current obstacles with the tear volume; however, each come with its own set of drawbacks (listed in parentheses): (i) Sedate or anesthetize the animal to extend collection duration and obtain a larger volume (altered lacrimal functional unit and ocular surface homeostasis) 183,188,202,238,[245][246][247] ; (ii) Pool tear samples from several subjects (reduced statistical power and loss of information regarding interindividual variability) 216,241,245 ; (iii) Induce reflex tearing with a stimulant-either physical (e.g., irritation to nasal mucosa or cornea), chemical (e.g., parenteral pilocarpine or ammonium fumes) or physiological (e.g., yawn or sneeze reflex)-thereby accelerating tear flow and shortening collection time (diluted tear sample, unable to control flow rates) 202,235,246,249 ; (iv) Instill fluid (e.g., saline) on the ocular surface immediately before tear collection, a process called "flush" or "washout" that yields a larger tear sample in a shorter amount of time (diluted tear sample, nonstandardized instilled volume, nonhomogenous mixing of fluid with tears). 193,229,235,[245][246][247]249 In particular, the diluting effect of reflex tearing or flush methods may drop the concentration of low-abundant compounds below the analytical limit of quantification, and potentially mask differences between groups due to reduced variance in tear composition. 229 Taken together, although direct tear collection remains the preferred method of some investigators given the "undisturbed" tear sample retrieved, 250 the serious drawbacks listed above have prompted a growing number of clinicians and researchers to consider indirect tear sampling in humans as suitable alternatives. 190,231 It is the authors' opinion that indirect tear sampling is also preferred in dogs, cats and laboratory animals. Ultimately, the patient's safety and comfort during tear collection is paramount and, as suggested by Berta, "it is better to use wellcontrolled methods than to try to cause as little irritation as possible." 251

| Indirect tear sampling
Indirect techniques involve tear fluid absorption with either Schirmer tear strips or absorbent sponges, followed by extraction of tear compounds by centrifugation and/or solvent elution.
Schirmer tear strips are routinely used to measure tear volume for clinical assessment of dry eye disease in humans and veterinary species. 160 236 A material with hydrophilic and hydrophobic properties (e.g., polyvinyl acetal and polyurethane) 236,264 is generally preferred to optimize the amount of fluid absorbed and the amount of fluid retrieved from the sponge. For tear fluid collection, the sponge is held against the lacrimal lake by the operator (to minimize reflex tearing), 234,239,242,244,266 or placed beneath the lower eyelid for a given period of time. 214,236,264 Tear fluid recovered from absorbent sponges can be assayed for selected tear compounds, similar to Schirmer strips.
Indirect tear collection is superior to direct capillary sampling in many aspects, namely: (i) Improved tolerance and acceptability by patients 182,231,266,267 ; (ii) Ease of use and operator safety, especially for Schirmer strips, allowing non-specialists to perform the procedure with minimal training 231,236,266 ; and (iii) Larger volume of tears collected in a shorter duration. 234,236,239,244 Absorbent materials collect tears but can also pick up cellular and extracellular "debris," an attribute considered beneficial by some as the sample obtained is more representative of the dynamic microenvironment at the ocular surface, 268 but also perceived as a limitation by others as the fluid retrieved is not "pure" tears. 250 On this note, the main limitation of indirect sampling is the "invasiveness" of the technique, at risk of promoting reflex tearing and altering the composition of the tear fluid; indeed, several studies showed variable tear composition between directly and indirectly collected samples, with notable differences in the qualitative and quantitative profiles obtained for tear lipids 221 and tear proteins. 216,220,226,269 Another important drawback is related to the adsorptive properties of Schirmer strips or absorbent sponges, that is, incomplete release of tear compounds following extraction 209,214,257,258,265 ; however, the authors believe this limitation can be minimized/controlled with adequate precautions (see Section 3.3).

| Schirmer strips versus absorbent sponges
Sponges can rapidly absorb up to 106 µl of tear fluid in dogs, 264 while the maximum absorptive capacity of Schirmer strips is~31 µl (i.e., 35 mm wetness). 270 With sponges, however, the operator can only control the duration of tear collection and not the volume of tears soaked up in each individual. The resulting variability in tear volume absorbed often translates into large intra-and inter-subject variability in the concentration of the compound(s) of interest, as shown for protein content 209 and various drugs such as doxycycline, 214 minocycline, 271 voriconazole, 272 and ofloxacin. 244 On the other hand, the ability to control the volume of tears absorbed with Schirmer strips (i.e., same mm mark) generally improves the reproducibility of the results. 209,214,215,273 As such, the authors prefer (i) absorbent sponges for collecting large volumes of tears in canine subjects-that is, for further use as blank tears in bioanalytical assays, for example-and (ii) Schirmer strips for collecting known amounts of tears in any scenario where reproducibility of the data is important (i.e., for group comparisons, or follow-up of the same individual over time).

| Schirmer strips for protein quantification
For consistency purposes, the authors recommend the use of dye-free Schirmer tear strips, being consistent with the manufacturer and lot number (given the reported variability in absorptive and adsorptive properties among Schirmer strips), 274,275 as well as the time of collection (e.g., morning), 231 because of known diurnal variability in lacrimal protein composition in humans 226,249 and dogs. 273 The distal end of the Schirmer strips should remain in position (ventrolateral conjunctival fornix) until 20, 25, or 30 mm wetness is reached. 215,240,259,270,273 Strip wetness <20 mm is discouraged given the potential "concentrating effect" of the absorbent filter with low tear volumes, 209  irritation ensued by the prolonged test duration. Importantly, investigators should be consistent with the selected mm-mark (strip wetness) within and between patients to standardize the volume of tears collected among subjects.
This strategy provides a lower coefficient of variability in tear protein content compared to ophthalmic sponges 209 and capillary glass tubes 273 in dogs, thereby improving the reliability and reproducibility of the data. Tear extraction and protein analysis can be done directly after tear collection, or can be postponed to a future date as long as Schirmer strips are stored immediately at −80°C and the stability of the compound(s) of interest is verified. 249 Following tear extraction with centrifugation, 209,231,270,273 elution in solvent, 173,182,190,216,240 or a combination of both, 258,276 total protein content (TPC) should be quantified to standardize the amount of sample used for subsequent analyses. 173,190 The authors' preferred method is infrared spectroscopy with Direct Detect TM (EMD Millipore, Danvers, MA) as the technique utilizes merely 2 µl of tear sample, without any of the drawbacks of colorimetric protein assays (e.g., Bradford, Lowry), including variability with specific protein composition and potential contamination from the absorbent material. 234,277

| Schirmer strips for drug quantification
The following steps should be considered to optimize drug quantification in pharmacological studies: • Study design: In studies that assess tear film pharmacokinetics following topical drug administration, one must consider a potential limitation associated with Schirmer strips which remove most of the tear fluid in early collection times, thereby negatively impacting the "true" tear concentration at later time points. 249 For this reason, the authors recommend to conduct pharmacokinetic studies in tears over several days (e.g., 10 days for although this method should account for differences between subjects such as greater tear volumes in dogs of larger body weight. 123 Another aspect to consider in the study design is the assessment of drug kinetics in diseased eyes, rendering the study results more clinically applicable (see Section 4.2); in fact, tear film concentrations and ocular bioavailability are likely to differ in healthy versus diseased eyes (e.g., excessive lacrimation, increased absorption into congested conjunctival vessels, albumin binding), 215,278 yet the majority of ocular studies to date are conducted in healthy individuals which is a clear limitation for translation of research findings from bench to bedside.
• Tear collection with Schirmer strips: It is important to homogenize the volume collected within and between subjects by standardizing the extent of strip wetness (≥20 mm mark)-this approach limits the variability in tear concentrations related to the collection method itself. 214,215 The amount of wetness is then converted to a volume (µl) to calculate actual tear film concentrations 270 ; data reporting is otherwise limited to µg/g of strip. 211,244 In dogs, the median volume absorbed by Schirmer strips is 18 µl ( In parallel, investigators should record the duration of tear collection (e.g., 50 s to reach 20 mm) to calculate a flow rate (µl/min) for each sample obtained. 214,235 In one of our experiments with doxycycline in dogs, flow rate did not influence tear concentrations, 214 but this finding might not be generalizable to other drugs and/or other species of interest.
• Extraction protocol optimization: A drug can be extracted from Schirmer strips via centrifugation, solvent elution, or a combination of both methods. However, a single extraction protocol cannot be generalized to all pharmacological studies as the specific physicochemical properties of each drug (e.g., molecular weight, lipophilicity) can affect the extraction efficiency from the filter papers. 257 As such, investigators should consider conducting a preliminary experiment to determine the optimal extraction protocol for the drug studied, and report specific recovery rates (mean ± SD, range). For instance, the recovery of prednisone and prednisolone was maximized with a combination of centrifugation and elution in methyl tert-butyl ether, a solvent chosen over methanol and acetonitrile based on superior drug extraction from Schirmer strips ( Figure 4). 215 Of note, a comprehensive review of all reported protocols is beyond the scope of the present work, and further research is warranted to assess the potential benefits (or lack thereof) of extraction steps reported in the literature, such as cutting Schirmer strips into small pieces 173,219,262 or using ultrasonic agitation. 215,219,249 Ultimately, an optimized extraction protocol is important as it enhances the reliability of the data at hand, improving the sensitivity of the bioassay, and providing drug concentrations closer to "true" biological levels in the tear film.
• Bioanalytical method optimization: An internal standard represents a structural analog or a stable labeled isotope with nearly identical physicochemical properties to the analyte of interest (e.g., molecular weight, elution time).
The selected internal standard should be applied directly onto the dry portion of the Schirmer strip ( Figure 4), 215 that is, before tear extraction instead of postelution with solvent as routinely described 211,262 ; this step allows for drug quantification to account for potential variability in extraction efficiency between samples.
Further, the standard calibration curve solutions should be constructed by spiking known drug concentrations and internal standard onto Schirmer strips, followed by the same extraction protocol as for biological samples; this step is equivalent to "spike and recover" experiments recommended for other analytes such as cytokines 265 and proteins. 257 Last, actual tear fluid should be used whenever possible as the selected matrix for standard calibration curve and quality control solutions, 214,215,272 as the reported surrogates (e.g., artificial tear solution) 279 do not account for chemical interferences and matrix effects that typically occur with a complex biological fluid (e.g., ionization suppression). 272 Blank tears can be collected with absorbent materials before study initiation, retrieving up to 84 µl in 1 min with ophthalmic sponges in dogs 264

| Spontaneous ocular surface diseases in dogs with translational applications to humans
Spontaneous ocular surface disorders are common in dogs and represent one of the major causes for referral visits to veterinary practitioners. 280 In contrast, naturally acquired ocular surface pathology is much less common in rabbits 281,282 and is rare in mice and rats. 283-285

| Dry eye disease
Dry eye disease (DED) represents one of the most common ocular conditions in humans with an estimated prevalence ranging from 5% to 50% in different regions worldwide. 286 The disease is also very common in dogs (prevalence 1.5-35%), 280 although not a single report of spontaneous dry eye case exists in laboratory animals such as rabbits, mice, and rats.
The pathogenesis of DED is very complex, involving diverse physio-anatomical factors such as lacrimal gland integrity, meibomian glands function, hormonal balance and neuronal input. 286 Numerous models of dry eye have been established in animals over the years, [287][288][289] helping to elucidate complex pathological mechanisms involved in DED and develop novel therapeutics for humans. However, the major drawbacks of most animal models are the acute nature of the induced pathology (vs. chronic disease in humans) and the focus on a single component of the lacrimal functional unit, such as surgically removing the lacrimal gland in mice to reduce tear secretion, 97 cauterizing the lid margin in rats to induce meibomian gland dysfunction, 34 or instilling topical 1% atropine in rabbits to disrupt the efferent neural input. 290 These experimental models can be generally improved by increasing the number of interventions in the study animals, for instance combining lacrimal glands removal with chemical destruction of the conjunctiva in rabbits, 291 or combining scopolamine administration with desiccating environmental stress in mice 165 ; yet, these complex models remain suboptimal at best given the acute nature and the inability to fully encompass the complexity of DED pathophysiology. Dogs, on the other hand, develop DED in a spontaneous manner and do not require invasive procedures to disrupt the lacrimal functional unit. 160,253 The canine condition, termed keratoconjunctivitis sicca (KCS), is clinically and immunopathologically similar to humans with aqueous deficient dry eye disease (ADDE) and possesses several attributes that are beneficial for translational research: • Canine KCS is typically bilateral, develops in middle-aged animals, is more common in female dogs and in certain breeds (e.g., American Cocker spaniel, English Bulldog), mimicking the diversity of dry eye in humans related to sex and race. 253,286 • Immune-mediated dacryoadenitis is the most common etiology of KCS in dogs-similar to human patients with ADDE secondary to Sjögren's syndrome-in which progressive lymphocytic infiltration of the lacrimal gland(s) damages the secretory tissues and reduces aqueous tear production. 253 • Spontaneous symptoms of ocular irritation, conjunctival hyperemia, and corneal scarring correlate directly with aqueous tear production, a parameter that is easily measured/quantified using a standard Schirmer tear test strip.
• Multiple diagnostic tools used in humans can easily be applied in dogs (but not rodents) given the large size of the canine globe, 160,288 including tear osmometry, vital staining, strip meniscometry test, infrared meibography and corneo-conjunctival impression cytology.
• Dogs and humans display a similar response to common therapeutics for dry eye disease; in fact, the two FDA approved anti-inflammatory drugs for dry eye disease in humans (cyclosporine and lifitegrast) were first developed in canine patients with spontaneous KCS. 253,292 The main limitation to consider in dogs is the tendency for clinical signs to be more pronounced in that species vs. humans with ADDE (e.g., tenacious mucoid discharge, corneal melanosis, neovascularization), in part because canine KCS is often diagnosed at a later stage when owners fail to recognize more subtle clinical signs early on.

| Allergic conjunctivitis
Allergic conjunctivitis is a common disorder in humans with an approximate prevalence of 40% in the North American population. 293 The disease is characterized by an immunopathological reaction of the ocular surface to the external environment, resulting in clinical symptoms that range from mild conjunctivitis (seasonal or perennial) to the more severe, vision-threatening vernal keratoconjunctivitis and atopic keratoconjunctivitis. 293 Over the past few decades, extensive research on small laboratory species (mice, rats, guinea pigs) has helped elucidate some of the complex molecular and cellular processes involved in the pathogenesis of ocular allergies. 294,295 However, these experiments primarily relied on a relatively small selection of allergens (e.g., ovalbumin, compound 48/80, ragweed pollen), using an experimental design that merely mimics acute forms of the disease-not chronic allergen exposure over months to years-therefore limiting the long-term clinical significance of these findings. On the other hand, dogs possess notable benefits for the comparative study of allergic conjunctivitis, especially when considering companion animals rather than laboratory Beagles: (i) these animals share the same environment (and related allergens) as their human owners, unlike commonly used species who are housed in a laboratory setting; (ii) companion dogs are outbred, providing a genetic diversity background that better reflects the human population than inbred laboratory species; and (iii) dogs develop a spontaneous form of allergic conjunctivitis. Spontaneous allergic conjunctivitis is relatively common in dogs, often associated with other allergic disorders such as canine atopic dermatitis. 296 Similar to humans, the clinical signs of allergic conjunctivitis involve conjunctival hyperemia, chemosis, pruritus, and ocular discharge, the disease can be diagnosed with high sensitivity and specificity using the conjunctival provocation test, 296 and similar therapeutics are used in both species including topical antihistamines, mast-cell stabilizers, NSAIDs, and immunomodulators. 293

| Microbial keratitis
It is well recognized that a natural host is best suited for studying infection, as several species-specific factors (e.g., anatomical, physiological, genetic, immune) closely influence the host-pathogen interactions and subsequent clinical response. 297 These factors likely explain why rabbits and rodents-traditionally used to model ocular surface infections in humans-cannot fully recapitulate the disease presentation and progression that occur in human patients. 81,298 As such, there is an emerging appreciation for the translational advantage of studying spontaneous (and not experimental) ocular infections in dogs: • scientists to investigate the ocular surface in health and disease states. 200 The model specifically focused on conjunctivitis as this condition is frequently encountered in humans and dogs, 157,303 developing either as a primary condition (e.g., bacterial, viral), or as a bystander to common ophthalmic diseases such as blepharitis, keratitis, uveitis, and glaucoma. This model is particularly appealing given the low cost, noninvasiveness, self-resolving nature, ability to adjust the duration and severity of the disease, and shared features with naturally occurring diseases in human and veterinary medicine.
The main highlights of the translational "large animal" model are as follows: • The selected compound (histamine) is inexpensive and triggers local inflammation in a nonspecific manner.
• Disease duration is dose-dependent, self-resolving within an average of 115 min (1 mg/ml), 190 min (10 mg/ml), or 390 min (375 mg/ml). The duration of conjunctivitis can be lengthened by repeating topical histamine administration at set intervals. 215 • Topical histamine is safe and generally well-tolerated, although selected eyes receiving the highest dose of histamine (375 mg/ml) can develop mild ocular irritation (lasting <1 min), blepharitis or miosis.
• Tear film composition changes in eyes with experimentally induced conjunctivitis (e.g., higher levels of serum albumin and inflammatory cytokines), mimicking clinical patients with ocular surface inflammation.
• A transient increase in tear quantity and decrease in tear quality occur, although tear film homeostasis is rapidly restored in ≤5 min. 304 Levels of serum albumin are increased in tear film of canine eyes with experimentally induced or naturally acquired conjunctivitis, 200,273 a physiological variation caused by the breakdown of the blood-tear barrier ( Figure 6). Disruption of the blood-tear barrier is also described in human patients with spontaneous ocular surface disorders (e.g., dry eye, allergic conjunctivitis) 226,305-307 and other animal species (rabbits, horses, and guinea pigs). 164 (ii) proinflammatory mediators other than histamine are also responsible for triggering conjunctival inflammation in clinical patients (e.g., leukotrienes, cytokines); (iii) conjunctival inflammation is relatively short-lived (115-390 min) and cannot mimic the physiological changes noted in patients with chronic conjunctivitis (e.g., reduced goblet cell density).

| Clinical relevance of serum albumin leakage in tear film
Elevated serum albumin levels in the tear film represents a biomarker for ocular insult or inflammation in humans, dogs, and other species. 200,226,305,309 In brief, plasma-derived albumin leaks onto the ocular surface from congested conjunctival vessels and mixes with the tear film; as such, albumin concentration in tears is generally low in healthy state but increases substantially in diseased eyes. 226 For instance, a recent study showed that canine eyes with diverse ocular diseases (e.g., corneal ulcer, uveitis, glaucoma) had lacrimal albumin levels that were up to 14.9-fold greater than contralateral healthy eyes. 200 Albumin is a relatively large protein that has a remarkable capacity for binding ligands. 171 At the level of the eye, protein binding represents an important restriction to drug absorption as only the unbound fraction of the drug diffuses across the ocular tissue barriers. 164 Combined with the rapid drainage of tears following eyedrop administration (in humans/dogs, not true in rabbits), any portion of drug that binds to albumin in tear film can be considered as "lost" from a pharmacological standpoint. Broader implications of the blood-tear barrier breakdown on ocular drug pharmacokinetics are listed below: • Reduced bioavailability for intraocular targets: The inability of bound therapeutic drugs to penetrate the cornea lowers the amount of drug available inside the eye to exert its pharmacological action. The physiological effects of increased albumin levels in tears was recently demonstrated with tropicamide (and to a lesser extent latanoprost) in dogs, 278 as well as pilocarpine in rabbits. 164 Of note, the impact of lacrimal albumin on the pharmacological activity of a given drug is likely modulated by various factors including the concentration of the formulation, the mechanism of action and the potency of the drug for its biological target. 278 • Reduced bioavailability for ocular surface targets: Drug-albumin interactions in the tear film could also be detrimental for management of ocular surface disorders, for instance reducing the efficacy of therapeutics for bacterial keratitis as only the unbound portion of an antibiotic is microbiologically active. 172 Preliminary experiments conducted by the authors showed that the presence of albumin results in higher minimal inhibitory concentrations (i.e., reduced susceptibility) for various antibiotics against common bacterial isolates in dogs. 335 • Tear film concentrations of systemically administered drugs: Drug in the plasma compartment can access the tear film by active secretion from the lacrimal gland, or passive diffusion through the conjunctival vessels. The latter is theoretically enhanced when the blood-tear barrier is disrupted. In humans, this physiological feature could explain why the concentration-time profiles of cetirizine were similar in serum and tears in patients with allergic conjunctivitis. 212 In dogs, tear film corticosteroid levels were generally higher in conjunctivitis versus control eyes following oral prednisone administration (up to +64%), although differences were not statistically significant. 215 The degree of conjunctival permeation is likely to vary among therapeutic drugs given differences in their physico-chemical properties. 174,312 These findings highlight the importance of conducting pharmacological studies in clinically relevant preclinical species that are able to recapitulate leaky conjunctival vessels and elevated albumin levels in the tear film of clinical patients with ocular diseases. For topical drug administration, the authors recommend using an experimental model of blood-tear barrier breakdown (e.g., histamine-induced conjunctivitis or alkali burn models) 200,313 so that albumin and other relevant proteins are already present on the ocular surface at the time the drug mixes with the tear film. 278 For systemic drug administration, the authors suggest conducting a preliminary experiment to assess whether conjunctival inflammation affects tear film concentrations to a significant extent. If not, pharmacological assessment in healthy eyes should be sufficient. Incidentally, the physicochemical properties of some drugs (e.g., size, lipophilicity, polar surface area) 174,312 may allow for high conjunctival permeation under normal conditions, thereby rendering differences between healthy versus diseased eyes insignificant.

| Corneal injury in dogs: In vivo and ex vivo models
Corneal injury is common in human and veterinary patients-whether due to trauma, surgery, or other causes-and the resulting corneal scar remains one of the leading causes of blindness in animals and people worldwide. 314 Although small laboratory animal species are commonly used in corneal scarring research, 315 results derived from these models have several limitations. The corneal thickness is much smaller in rabbits and rodents compared to humans (Table 1). In addition to thin corneas, mice and rats have corneas that are much smaller in diameter compared to people; consequently, it is often difficult to isolate the central cornea when performing the experimental procedure (e.g., chemical burn) and the damage caused to surrounding limbal stem cells negatively impacts the wound healing process. Using the dog as an animal model is therefore more appropriate, not only due to closer resemblances in ocular surface anatomy and physiology with humans, but also the relatively high prevalence of naturally acquired corneal pathology in the canine species.
In that regard, Gronkiewicz et al. recently developed a novel in vivo corneal fibrosis model in canines 313 ; the authors induced corneal scarring with an alkali burn and investigated the ability of suberanilohydroxamic acid (SAHA) to inhibit fibrosis using this large animal model. The availability of such a model presents a clear opportunity for translational research (i.e., intact innervation, tear film, blood supply), although experimentally induced corneal wounding (at risk of secondary infection) and subsequent corneal scar in dogs represent potential ethical challenges. As an alternative, other authors have established ex vivo canine corneal cultures that can be used to model wound healing and assess antifibrotic compounds, [316][317][318] or better understand the pathophysiology of herpesvirus in a virus-natural-host environment 319 ; in that study, the authors established an air-liquid canine corneal organ culture model to study acute herpetic keratitis, showing important similarities in the response to CHV-1 to what has been described for HSV-1. 319

| Limitations of using dogs for translational research
Canine studies can improve the predictive power of preclinical research when translated to human clinical trials, as highlighted in this manuscript for ocular pharmacology, although the use of dogs in research settings presents a few challenges. Dogs are expensive to purchase and maintain in specific housing, the molecular tool kit specific to dogs is more limited compared to traditional laboratory animals (mice, rats, and rabbits), and there are important ethical and public-perception considerations associated with the use of dogs as experimental animal models. 6,81,287,288,320 Further, most canine subjects have a mixed genetic background (outbred) that can affect the variability in disease characteristics (e.g., severity, response to therapy); in contrast, rodents and rabbits have readily available inbred strains with the same genetic makeup, as well as knockout and transgenic strains that allow investigators to explore details of pathways and factors involved in disease pathogenesis. 81,201 Last, clinical trials involving dogs can be limited by rigorous scientific and ethical review, unpredictable case enrollment, variability in disease phenotype among patients, economic challenges (e.g., owner ability to provide care), and owner compliance, to name a few. 6,321 SEBBAG AND MOCHEL | 2591 5 | CONCLUSIONS "Considerable reservations may be felt about comparing results from rabbits with those from humans because of the differences between the physiology of tear flow and mixing and general anatomy. Nevertheless, the rabbit is the principal experimental animal in ophthalmology, so comparisons are needed." 322 Sadly, this quote published over 45 years ago is still representative of today's state of ophthalmic research. Rabbits and small laboratory rodents continue to be used primarily (if not exclusively) in most areas of ophthalmic research, 1 a concerning fact given the vast anatomical and physiological differences that exist with humans. Of note, such differences should not be regarded as merely "weaknesses" for translational research, but rather evolutionary adaptations optimally suited to the environment and behavior of each species; for instance, rabbits likely developed a very stable tear film to limit intermittent blindness that occurs with each blink, 323 thereby reducing the risk of predation. Noteworthily, recent innovations have helped mitigate some of the drawbacks of traditional laboratory species-for instance providing manual blinking and supplementary tear flow in anesthetized rabbits, 324 or reverse-engineering the ocular surface using human cells in vitro 325 -however, the authors believe the complexity of the ocular surface and integrated lacrimal functional unit cannot be fully recreated without in vivo conditions in awake subjects.
The comparative work presented throughout this review provides evidence that dogs are best suited for translational research in ophthalmology. Unlike small laboratory animals, dogs share similar anatomical and physiological features to humans, similar environmental stressors and genetic variation, and a range of naturally occurring ophthalmic diseases that closely resemble clinical phenotypes in human patients. The resemblance between dogs and humans is particularly relevant in the field of ocular pharmacology, with notable similarities in blink rate, tear turnover rate (basal, reflex), volumetric capacity of the palpebral fissure, and other factors pertinent to drug diffusion (e.g., globe volume, corneal thickness); nonetheless, a few differences should be accounted for in comparative studies, such as the presence of a nictitating membrane, greater tear volume, larger corneal size and lower corneal elastic modulus in dogs.
Similar to other fields of medicine, preclinical studies in ophthalmology could involve canine patients with spontaneous ocular diseases-many of which share striking resemblances with their human counterparts-integrating the expertise of veterinarians, physicians and basic science researchers under the umbrella of the One Health Initiative. 6,326 Alternatively, or complementarily, preclinical animal work could be performed in laboratory dogs in whom ocular disease is experimentally induced, making sure to account for the blood-tear barrier breakdown (noted in clinical patients with "red eyes"). In all cases, tear fluid can be easily collected from canine eyes for various bioanalytical purposes, favoring Schirmer tear strips over other collection methods given the excellent safety profile and enhanced reliability in analyte quantification (e.g., proteins and drugs) provided by this method.