Dry eye disease (DED) is a chronic inflammatory pathology of the ocular surface which affects 5–50% of the general population, the incidence of which is constantly increasing due to our lifestyles and environment [1]. Many etiologies can lead to entry into the vicious circle of DED - from auto-immune disease, Meibomian gland dysfunction, allergy, and surgical procedures, to iatrogenic and xenobiotic exposures [2,3]. Once the vicious circle has commenced, eliminating the cause is not enough to restore tissue homeostasis, and at present, no curative treatment is available. This explains the important need to anticipate and prevent the triggering of the vicious circle.

The cornea is one of the main ocular surface barriers, permanently exposed to an adverse environment including occupational and household cleaning products, cosmetics and eye drops. The corneal surface is composed mainly of epithelial cells and nerve endings emerging from sensory neurons located in the trigeminal ganglion (TG) [4], both impacted in Toxicity-Induced Dry Eye (TIDE) [[5], [6], [7], [8], [9]], as has been described with benzalkonium chloride (BAC), a quaternary ammonium compound, known to induce dry eye when instilled over the long term [10]. Indeed, BAC is the most commonly used preservative in eye drops at concentrations ranging from 0.004% to 0.020% [[11], [12], [13]], the toxic effects of which on ocular surface cells have been widely studied [14] and lead to oxidative stress, elaboration of inflammatory mediators and cellular and neuronal damage [5,6,8,15,16].

TIDE was recently proposed to be the first step in ocular toxicity, before the onset of irritation and clinical manifestations, after exposure to xenobiotics, even at low concentrations [9]. To better anticipate TIDE, it is essential to develop new, sensitive, in vitro models [9] that evaluates not only the corneal epithelium but also its innervation provided by trigeminal neurons [5,15,16]. Indeed, the neuronal component of the cornea is implicated in ocular surface homeostasis and tear production [17,18].

Importantly, only the nerve endings of the trigeminal neurons are embedded in the corneal epithelium and thus exposed to BAC compound. This complex corneal unit represents a challenge for anyone who wishes to create a model which would take all of these characteristics into account. To date, only ocular irritation detection is considered [[19], [20], [21]], analyzing the parameters of the epithelium alone, but these do not evaluate the neural component of the cornea, mainly due to the lack of a validated model of the corneal nerve network [22].

Several studies have shown BAC neurotoxicity in vitro on monolayer culture or compartmentalized trigeminal neurons. Indeed, Launay et al. showed a 70% decrease of axonal length on a TG monolayer culture after a 6h BAC 5.10−4% exposure [5], and Sarkar et al. observed a 90% decrease of in neurite length in a compartmentalized culture of TG cells exposed for 4h to BAC 10−3% [23]. However, these studies are far from cornea physiology, not taking into account the interaction between trigeminal nerve endings and the corneal epithelial cells that may act as surrogate Schwann cells, responsible for corneal nerve homeostasis [24].

Of significant interest over the past twenty years, organs-on-a-chip aim to miniaturize an organ, facilitate assembly of the various cell types, and better recreate the in vivo dynamics of the organ [[25], [26], [27], [28], [29]]. Moreover, compartmentalized microfluidic chip has become the method of choice because it allows a selective treatment of nerve endings, independently of the soma.

In this study, we design and validate a microfluidic co-culture between mouse trigeminal sensory neurons and primary mouse corneal epithelial cells (MCECs). Our compartmentalized chip allows us to more closely emulate corneal physiology and to separate the evaluation of the trigeminal cells (proximal compartment, PC) and MCECs and sensory nerve endings (distal compartment, DC) with the acquisition of high-content fluorescence images and automated multiparameter analysis of both compartments. Using this new innovative and sophisticated tool, we aim at better deciphering and anticipating the toxicity mechanisms occurring in a TIDE context, through the analysis of indirectly exposed neuronal cells and directly exposed nerve endings/epithelial components. High-content imaging and multiparameter analysis are combined to evaluate spatiotemporal protein expression in neurons and epithelial cells following xenobiotic exposure. To detect these fine mechanisms of toxicity, we coupled the investigation of phenotypic alterations (CK3, ZO1, GFAP, βtub) of each cell types, to stress responses and activation factors (ATF3, ATF6, phospho-p44/42).

To validate our corneal neuroepithelial compartmentalized microfluidic chip, we confirmed that topical (single and repeated) applications of BAC to the neuroepithelial compartment induces a direct toxicity, even at a low concentration of 5.10−4%, on sensory nerves endings and MCECs, as well as an indirect toxicity on trigeminal soma not directly exposed to the toxic. The concentration of 5.10−4%, that is to say, 8 times lower than the lowest concentration of 0.004% used in eyedrops, was chosen to attest model sensitivity to detect low adverse molecular events. In addition, we demonstrate the potential role of the tripartite combination of “compartmentalized co-culture/high-content imaging/multiparameter analysis” for the systematic and automated testing of toxicants for public health purposes and opens the way to the evaluation of regenerative or neuroprotective compounds.