In vitro and ex vivo models for evaluating vaginal drug delivery systems.

The development of suitable models for evaluating vaginal drug delivery systems requires understanding of the lower female genital tract anatomy ( Fig. 1 ) and the physiology of the mucus and tissue barriers. Vaginally applied drugs will typically have direct access to the vaginal and ectocervical mucosae. The cervix can be divided into two regions: the ectocervix, which extends below the cervical os into the vaginal canal, and the endocervix, which connects to the uterus. The endocervix is responsible for the production of mucus, which is secreted into the vaginal lumen [ 8 , 12 ]. The vaginal mucosa of reproductive aged women is comprised of stratified squamous epithelium that sits on the lamina propria [ 16 ]. The vaginal epithelium has distinct strata: the basal layer which is mitotically active, and the super basal and superficial layers that are comprised of flattened cornified cells. The vaginal stratum corneum lacks intercellular junctions, so it does not usually keratinize or form a complete lipid envelope, thus remaining penetrable by microbes and cellular and molecular mediators of the immune system [ 16 ]. Extensively described in the microbicide literature, drug transporters in the vaginal epithelium impact drug distribution, retention, and metabolism [ 17 ]. While the vaginal mucosa does not contain mucus-producing cells, the CVM lining the epithelial surface contributes to the barrier function to pathogens and particulates [ 6 , 8 ]. The biosynthesis of mucus is a complex process, requiring significant post-translational modifications and packaging of mucins as they move through the secretory pathway [ 18 ]. There are two main types of mucins that are structurally and functionally distinct. Secreted mucins oligomerize and form hydrogels, while cell-associated mucins remain apically anchored to the plasma membrane [ 19 – 22 ]. Secreted and cell-associated mucins are translated in the endoplasmic reticulum and undergo folding, disulfide-bond formation, and glycosylation as they move through the secretory compartments [ 23 ]. Mucins typically consist of folded cystine-rich domains separated by long regions of unfolded serine-, threonine-, and proline-(PTS) rich repeats. The PTS domains serve as the site for extensive O-linked glycosylation ( Fig. 2A ) [ 21 , 24 ]. Oligosaccharide chains contain at least two sugars and are either neutral or negatively charged due to sialic acid or sulfate groups [ 19 , 25 ]. Cell-associated mucins are often covalently cleaved during synthesis into two sub-units, which remain non-covalently associated through and after shuttling to the cell surface [ 18 , 26 ]. Functionally, cell-associated mucins protect the cell and sense environmental cues, while secreted mucins form gel-like structures that lubricate and protect the epithelium from pathogen penetration [ 6 , 18 , 27 , 28 ]. In the ectocervix, only cell-associated mucins, including MUC1, MUC3A, MUC3B, and MUC12, are expressed [ 26 ]. Secreted mucins, including MUC2, MUC5AC, MUC5B, and MUC6, are expressed in the endocervix, with MUC5B ( Fig. 2A ) showing the highest expression [ 20 , 29 – 31 ]. Secreted mucin glycoproteins are approximately 200 kDa – 200 MDa in molecular weight and 0.5–5 μm in length [ 24 , 32 ]. As such, in the secretory pathway, secreted mucins must first oligomerize and condense through hydrophobic interactions and disulfide bonds between C-terminal cysteine-rich domains. As the pH decreases and the Ca 2+ concentration increases from the endoplasmic reticulum to the Golgi apparatus, pH-dependent Ca 2+ binding condenses the polymerized mucins into bundles [ 33 – 35 ]. When the secretory granules containing densely packed mucins are released from cells, the sudden increase in pH and decrease in Ca 2+ causes the mucins to begin unbundling into a porous net-like structure [ 36 – 38 ]. Thus, CVM is effectively a net-like structure with charged glycosylated regions and hydrophobic domains that can effectively trap foreign pathogens and particulates by both steric and adhesive interactions ( Fig. 2B ) [ 6 , 39 ]. Characterization of the barrier function of CVM in the context of viral infection and nanoparticle-based drug delivery is described in more detail in section 4.2.2 . Hormonal changes throughout life and with exogenous hormone use can dramatically impact the properties of the FRT mucosae ( Fig. 3 ). Prior to puberty, the cervical surface is mostly composed of columnar epithelium. At the time of puberty, increased estrogen levels promote the thickening of the stratified squamous epithelium of the vagina and squamous metaplasia at the cervix, in which the columnar cells of the ectocervix mature into stratified non-keratinized squamous epithelium and mucus production increases [ 40 ]. The high estrogen levels are maintained throughout the reproductive years, after which, the decline of estrogen during menopause results in vaginal atrophy and decreased mucus production that is similar to pre-puberty [ 41 ]. Hormonal fluctuations during the menstrual cycle in the reproductive years also has a dramatic effect on mucus properties. Increased estradiol levels during ovulation encourage production of copious amounts of more watery mucus to facilitate sperm penetration and transport into the uterus [ 42 , 43 ]. Ovulatory cervical mucus is significantly more viscoelastic, and exhibits “spinnbarkeit,” or the ability to be stretched between two fingers without the mucus strand breaking [ 44 ]. Functional studies have described that ovulatory cervical mucins are more neutral, with less sialylated oligosaccharides, and appears to contain more globular mucin aggregates with larger open pores than non-ovulatory mucus [ 19 , 45 ]. However, despite the increased water content and penetrability to sperm, midcycle cervical mucus was found to have adhesive barrier properties to some small viruses, such as herpes simplex virus type 1 [ 46 ]. Increased fluid content during sexual arousal [ 47 ] is also likely to play a role in the barrier function of mucus, which is particularly important to consider in the context of sexually transmitted infections. Post-ovulation, the return to high progesterone levels results in more dry, thick mucus [ 43 ]. High progestogen levels associated with hormonal contraceptive use also thickens cervical mucus, which has been described as a secondary contraceptive mechanism [ 48 ]. Similarly, it was observed that there was a general reduction in the sizes of pores in CVM collected from pregnant women, a state associated with increased circulating levels of both progesterone and estradiol, compared to non-pregnant women [ 49 ]. However, a study using cervicovaginal lavage fluid (CVL) suggested that the effects of progesterone in reproductive-aged women were reduced in women with dysbiotic vaginal microbiota [ 50 ]. The authors found that CVL from premenopausal women using hormonal contraceptives was more viscous than CVL from premenopausal women not using hormonal contraceptives, but the effect was mitigated when women had dysbiotic vaginal microbiota [ 50 ]. Just as hormonal changes during puberty, menopause, and pregnancy lead to changes in cellular structures and mucus production, there is also a striking change in the vaginal microbiome. In general, one of two communities of microbiota develop as estrogen levels increase: those dominated by a subset of Lactobacillus species, or polymicrobial communities that are relatively Lactobacillus-depleted. It is estimated that in around half of cases, this microbial dysbiosis is associated with symptoms that clinically classify as bacterial vaginosis (BV) [ 51 ]. Vaginal microbial dysbiosis is associated with increased risk of pelvic inflammatory disease (PID), urinary tract infections (UTI), and preterm birth (PTB), as well as increased susceptibility and transmission of sexually transmitted infections (STI) including HIV [ 52 – 55 ]. Dominance by Lactobacillus species leads to acidification of CVM to a pH in the range of 3.5–4.0 through the production of lactic acid, which proceeds from increased levels of host glycogen due to estrogen stimulation [ 56 ]. Vaginal lactic acid concentrations are capable of killing many other species of vaginal bacteria, leading to the dominance of a sparse population of lactobacilli [ 57 ]. Conversely, a dysbiotic community is densely populated by many species of bacteria, many of which produce enzymes that affect the CVM. Gardnerella vaginalis , Prevotella , and Bacteroides species all may contribute to the elevated sialidase activity in CVM during BV [ 58 , 59 ]. Sialidases cleave the sialic acid groups from the oligosaccharides which are catabolized by the bacteria, thus favoring their growth [ 60 , 61 ]. Similarly, lower levels of sialic acid and high-mannose glycans have been reported in the CVL of women with BV [ 62 ]. As described above, mucin degradation may have contributed to decreased viscosity in CVL samples from women with BV [ 50 ]. Further, we observed that CVM from women with BV had reduced adhesive barrier properties to fluorescently-labeled nanoparticles and HIV virions [ 11 ]. Hormonal and microbial differences between different individuals impact CVM barrier properties and are an important consideration when designing vaginal drug delivery systems.

As is often the case in many other research areas, model system development for the lower genital tract mucosae lags behind that of other organ systems, such as the gastrointestinal tract and the airways. Examples of technologies covered in this ADDR issue include organ-on-chip, purified mucins, and synthetic mucins, which could also be applied to studying vaginal drug delivery. Here, we further highlight a limited number of innovative and complex in vitro systems that may serve as inspiration for the next generation in model systems of the FRT. In the airways, goblet cells produce mucus that coats the columnar epithelium, which contains ciliated cells that beat in the upward direction to achieve rapid mucociliary clearance [ 158 – 160 ]. One unique feature of the respiratory mucosa is that there is an air–liquid interface, such that materials entering the airways are suspended in air prior to contacting the mucus layer. There are numerous examples of air–liquid interface (ALI) organotypic airway tissue models derived from primary tracheobronchial epithelial cells that have been developed [ 159 , 161 ]. ALI models typically incorporate primary normal human bronchial epithelial cells that are seeded onto a transwell system and differentiated into polarized, pseudostratified tissue with dispersed mucus-producing goblet cells, ciliated cells, and extracellular matrix deposited beneath. The use of human cells in ALI culture systems has facilitated the study of infection by human respiratory pathogens, toxicity responses, and etiology of chronic obstructive and fibrotic airway diseases [ 162 – 165 ]. ALI culture systems have also been used to study bacterial biofilms, mucosal penetration by drug- and nucleic acid-loaded nanoparticles, and inflammation caused by environmental nanoparticles [ 166 – 170 ]. Further, ALI cultures at various stages of differentiation, ALI culture medium kits, and nebulizer systems have become commercially available, along with various published protocols, increasing accessibility [ 158 , 159 , 171 , 172 ]. However, researchers continue to push the boundaries of what is possible with ALI culture systems, including miniaturization [ 158 ], increasing tissue complexity via 3D printing [ 173 ], and organ-on-a-chip flow systems [ 174 ]. As the field progresses, ALI model systems become increasingly complex in terms of the ability to co-culture a variety of relevant cell types and mimicry of in vivo physiology. Similarly, there are many examples of advanced in vitro models of the gut mucosa that have utility in characterizing drug delivery systems, including intestinal organoids and organ-on-chip models [ 175 – 177 ]. For example, Yin et al. developed 3D intestinal organoids derived from isolated crypts that maintain spatial organization of cells, cell-matrix interactions, and cell–cell interactions [ 178 , 179 ]. However, spherical organoids have some limitations in recapitulating certain structural, microbial, and mechanical properties of the human intestines. Organ-on-a-chip models of the gut are beginning to integrate intestinal tissue, neurons, key bacteria strains, and immune cells to understand the interaction between these systems when challenged with a pathogen or disease. Systems have been developed that incorporate flow to mimic peristalsis and physicomechanical intestinal cues, such as the model depicted in Fig. 8 [ 178 , 180 ]. Systems have also been described with the added complexity of modeling interactions between the gastrointestinal tract and other organs, such as the liver [ 181 , 182 ]. In many cases, the presence of mucus has been shown to affect nanoparticle uptake in vitro , highlighting the necessity of co-culture with mucus producing cells in intestinal models [ 71 , 183 – 185 ]. There are various efforts underway for developing organoids and organ-on-a-chip platforms to mimic various tissues in the FRT, though largely with a focus on fertility, infection, endometriosis, and cancer [ 186 – 191 ]. Thus, FRT organoid model systems have largely been developed as models of the ovary, fallopian tubes, endometrium, and cervix. In the context of vaginal drug delivery systems, expanding these model systems to include co-cultures of polarized vaginal epithelium with an adjacent mucus-producing cervical monolayer oriented to allow for bathing the vaginal epithelium with mucus would be an advancement. Further incorporating the vaginal microbiota and its impact on the mucus barrier function would be valuable. As the field advances toward an in vitro system that integrates the uterus, cervix, vagina, and immune cells, such a model would be highly valuable for testing drug delivery and formulation toxicity in the FRT.

A commonly used approach for evaluating vaginal drug delivery systems is the use of freshly excised tissue ex vivo . Typically, animal tissues are more readily accessible than human tissues, although there are important considerations when comparing the human FRT to traditionally used animal models. Differences to consider include the anatomy, the vaginal microbiota and local microenvironment, effects of the hormone cycle on cellular and tissue structures and mucus production, and the duration of the hormonal cycle [ 96 ]. It has been shown using fluorescently labeled nanoparticles that the barrier properties of mouse CVM in the estrus phase of the estrous cycle is similar to that of human CVM [ 97 ]. However, the mouse estrous cycle is rapid and characterized by significant changes in vaginal tissue structure and thickness during each of the four stages lasting a total of only 4–5 days [ 98 , 99 ]. Further, hormonal changes were demonstrated to significantly impact the structure and barrier properties of mouse CVM [ 97 , 100 ]. Ensign et al. vaginally administered fluorescently labeled nanoparticles to mice that had been treated with estradiol or progestin (Depo Provera) to artificially induce morphological changes similar to what occurs during the high estrogen and high progesterone stages of the estrous cycle, respectively [ 97 , 100 ]. The tissue was then excised and sliced open to visualize particle movement in the CVM coating the vaginal tissue. Using MPT, they observed that the mobility of mucoinert nanoparticles as small in size as 100 nm was significantly hindered in the CVM of mice that were pretreated with Depo Provera. In contrast, mucoinert particles were freely diffusive in the CVM covering the vaginal tissue of mice that were either treated with estradiol or naturally cycling in the estrus phase. Described in more detail in Section 5 , these differences in particle mobility in mouse CVM ex vivo also correlated with vaginal tissue coverage in vivo [ 97 , 100 ]. Porcine and bovine vaginal tissues, often obtained from slaughterhouses, are more commonly used as an ex vivo tissue model. There are a variety of experimental set-ups and geometries that have been used to evaluate mucoadhesion of various formulations to porcine and bovine vaginal tissues ex vivo . For example, Jalil et al. examined the mucoadhesion of microbicide-loaded gellan gum-based vaginal films on porcine vaginal tissue ex vivo [ 101 ]. The mucosa was rinsed with SVF (pH 4.2) for 5 min, followed by contact with different polymeric films for 10 min to allow for surface adsorption. The tissue was then rinsed with SVF at a rate of 1 mL/min using a peristaltic pump, and SVF fractions were collected and analyzed to quantify mucoadhesion. Films containing gellan gum modified with thiol residues showed increased mucoadhesion compared to unmodified gellan gum [ 101 ]. Another approach described in the evaluation of liposome-based foam aerosols and nystatin-loaded microparticles involved placing the formulation between two pieces of porcine vaginal tissue each fixed to a plank [ 102 , 103 ]. One plank was fixed to a stainless steel base, and the other was connected to a thread strung over a pulley with a plastic beaker attached to the end. Water was then added dropwise to the beaker until the top piece of porcine tissue was pulled away ( Fig. 5A ). The mass of the water added was used to calculate the mucoadhesive force [ 102 , 103 ]. Gerton et al. utilized bovine vaginal tissue to determine the mucoadhesion of a hyaluronic acid-based film loaded with metronidazole [ 104 ]. Tissue was mounted with the mucosal surface facing outward to a half-cylinder of PVC pipe and propped at an approximate 60° angle to simulate gravitational forces. Films were placed on the mucosa, and SVF (pH 4.2) was pumped over the film at 1 mL/min for 3 min to simulate a rush of vaginal fluid, or at 1 μL/min for 24 h to simulate normal vaginal fluid secretion. The starting position of the film was marked to assess whether the film migrated over time. The authors noted that their films were able to stay in place on the bovine vaginal mucosa over time under both SVF flow conditions [ 104 ]. Excised vaginal tissues from rabbits, guinea pigs, cows, sheep, pigs, and non-human primates have also been used to study drug and particle permeability ex vivo [ 105 , 106 ]. Van Eyk et al. aimed to determine whether pig vaginal tissue was structurally similar enough to be predictive of drug permeability across human vaginal tissue [ 107 ]. Due to the limited availability of human vaginal tissue, they utilized excess vaginal tissue removed during hysterectomy procedures in postmenopausal women. Both the human tissue and pig vaginal tissue were mounted in a flow-through diffusion cell with PBS containing tritium-labeled drugs flowing over the tissue at 1.5 mL/h for 24 h. Good concordance was observed for flux of tritiated water and vasopressin, whereas statistically significant differences were observed for flux of 17β-estradiol, r-arecoline, and oxytocin. For r-arecoline, the flux was lower through porcine mucosa, whereas the flux was higher through porcine mucosa for 17β-estradiol and oxytocin [ 107 ]. It is possible that the difference in structure of vaginal tissue post-menopause compared to pre-menopause could play a role, and it is worth noting that the tissues were frozen and stored at −85°C prior to experimentation. For some drugs, the loss of the soluble mucus layer during storage in transport buffer could also be a factor in the apparent flux. The removal of the soluble mucus layer was also mentioned in a study investigating the mucosal permeability of dapivirine-loaded nanoparticles [ 108 ]. Discussed in more detail in section 5 , das Neves et al. unexpectedly observed that despite being adhesively immobilized in SVF containing mucins, positively charged nanoparticles showed increased permeability into porcine vaginal tissue mounted in a Franz cell ( Fig. 5B ). However, in addition to the lack of soluble mucus layer on the washed porcine tissue, the increased permeability to positively charged nanoparticles was likely due to toxicity [ 108 ]. Ex vivo vaginal tissue has also been used to demonstrate a lack of penetration through the vaginal tissue, and rather retention of a therapeutic in the vaginal lumen to exert a local effect. Bouchemal et al. used macaque vaginal tissue to demonstrate that a peptide microbicide (M48U1) formulated in a thermosensitive gel did not pass through the tissue, supporting local retention [ 109 ]. Excised macaque vaginal mucosa was rinsed with Ringer solution, mounted in an Ussing chamber, and the thermosensitive gel was loaded onto the mucosal surface in the liquid state ( Fig. 5C ). The chamber was then heated to 37 °C to facilitate gelation, and at specified time points up to 2 h, 0.1 mL of sample was taken from the donor and recipient compartments and replaced with fresh media. The tissue was also dissected to evaluate the presence of M48U1, and it was found that the peptide did not pass through the vaginal mucosa [ 109 ]. Similar to in vitro cell models described previously, ex vivo tissue models are also often used to assess the potential toxicity of vaginal formulations. Machado et al. sought to establish a reproducible model for the evaluation of vaginal semisolids [ 110 ]. They compared cervical, vaginal, and uterine cells lines to ex vivo porcine vaginal tissue. An array of approved cream and gel products containing estrogens and antimicrobials were compared to well-characterized reference gel products. Overall, they concluded that the cell lines gave more reproducible results but were much more sensitive to product exposure. Thus, they recommend that an ex vivo tissue model should be used prior to further preclinical testing [ 110 ]. Dezzutti et al. described the use of polarized human ectocervical biopsies to assess the toxicity of different vaginal lubricants [ 111 ]. The apical side of samples were exposed to different lubricants for 18 h, and then the biopsies were washed and assessed for cell viability and histology. In general, increasing osmolality correlated with increasing cell toxicity [ 111 ]. While it was reported that CVM had an average osmolality of 370 ± 40 mOsm/kg, lubricants used in toxicity studies have osmolalities as high as 8,600 mOsm/kg [ 112 ]. Such ex vivo tissue models have been used to demonstrate that a variety of prescription and over-the-counter vaginal products are not optimally designed for biocompatibility in the cervicovaginal tract. In particular, high osmolality and the use of some polymers and excipients have been highlighted as toxicity concerns [ 111 – 115 ]. There are a variety of approaches used to collect mucus secretions and mucosa from the FRT. Because mucus is secreted from the endocervix and plays a major role in fertility, many early studies characterizing the properties of cervical mucus involved direct aspiration from the cervical os [ 44 , 46 , 116 , 117 ]. During pregnancy, a cervical mucus plug (CMP) forms to prevent bacteria and other pathogens in the vagina from ascending into the uterus [ 8 ]. To collect an intact plug and not compromise the pregnancy, the CMP is often collected during labor when it is spontaneously shed [ 118 , 119 ]. Alternatively, an approach using a sterile, thin catheter to collect both distal and proximal CMP during pregnancy has been described [ 120 ]. Another approach that is often used to collect material for evaluating soluble markers in the lower genital tract is to perform a cervicovaginal lavage (CVL). A CVL involves flushing of the vaginal cavity with 2.5 – 10 mL of sterile fluid to collect mucus, cells and cellular debris, small metabolites, and bacteria from the vagina [ 121 ]. One limitation of the CVL collection approach is that there is inconsistent dilution between participants, and there is a not a clear marker that is universally used for normalizing analyte concentrations [ 122 ]. The dilution also negatively impacts the ability to characterize the structural and barrier properties of the mucus, as well as potential interactions between the mucus components and drug delivery formulations. To collect cervicovaginal mucus (CVM) from the vagina in the native and undiluted state, a technique was described that involves the insertion of a disposable menstrual cup [ 10 ]. The study participant is able to self-collect the sample by quickly inserting, twisting, and removing the menstrual cup, which is then centrifuged to collect the undiluted mucus. Samples are typically stored in the refrigerator and analyzed within 24 h to ensure integrity of the sample [ 10 , 11 ]. Menstrual cups have also been used to collect samples for cytokine and antibody analysis where the participant is instructed to leave the cup inserted for 1–2 h [ 11 , 123 ]. While this approach may lead to increased sample volumes, leaving the cup in for longer times was described to lead to increased water content and pH, which can affect structural and barrier properties to particulates [ 11 ]. Mucus is non-Newtonian and shear-thinning, behaving more solid-like at low shear rates and more liquid-like at high shear rates. Using traditional macroscopic rheological characterization techniques, non-ovulatory CVM is 10 4 –10 6 times more viscous than water at low shear rates, and only 10 2 -10 3 times more viscous than water at high shear rates that are similar to what would be expected during copulation [ 13 , 24 , 124 ]. Further, mucus is a hydrogel composed of a network of mucin proteins with fluid-filled pores, where the viscosity of the interstitial fluid is similar to that of water. Thus, when considering the rheological properties on the microscopic scale experienced by small molecules or viruses, the local viscosity is reflective of a fluid and not a viscoelastic gel [ 46 ]. However, larger particulates that are not small enough to pass through the pores unobstructed can become sterically trapped, experiencing the viscoelastic properties of the mucin network [ 13 , 125 ]. Early work toward the characterization of the pore structure of cervical mucus involved methods such as scanning electron microscopy (SEM) and transmission electron microscopy (TEM). However, the measured pore size was inconsistent between the different approaches and likely impacted by the drying steps involved in the specimen preparation [ 126 , 127 ]. A functional estimate of the mucus pore size within the intact hydrogel was proposed based on the rate of diffusion of different sized fluorescently-labeled proteins, viruses and virus-like particles in ovulatory cervical mucus [ 46 , 128 ]. Two methods were applied to calculate the rate at which fluorescently-labeled probes diffuse in cervical mucus and saline. The first, Fluorescence Recovery after Photobleaching (FRAP) method involves photobleaching a defined region and calculating the rate at which unbleached probe returns to the photobleached area [ 129 ]. FRAP is a useful technique for probes that are too small to track individually based on the limitations of the resolution of the microscope. The second technique, multiple particle tracking (MPT), calculates the effective rate of diffusion (D) based on the mean squared displacement (Δr 2 ) of individual fluorescent particles over time (Δt) using the equation D = Δr 2 /Δt [ 13 , 130 , 131 ]. Using both approaches, the authors observed that the diffusion of globular proteins (15–650 kDa) and two capsid virus-like particles (38 and 55 nm diameter) was similar in cervical mucus and saline, suggesting that the particles were small enough to diffuse unhindered in the interstitial fluid within the pores [ 46 ]. In contrast, the diffusion of herpes simplex virus type 1 (HSV) (180 nm diameter) was reduced 100–1000 fold in cervical mucus compared to saline, which was thought to be because of steric trapping of the virus. Using this information, an obstruction scaling model was applied to calculate an effective pore size of ~ 100 nm, which was consistent with predictions based on mucin fiber concentration using a cubic-lattice model [ 46 ]. However, later work went on to show that HSV was adhesively trapped in mucus, necessitating the use of non-adhesive particles to obtain more accurate estimates of effective pore size. Lai et al. utilized the MPT technique with a range of different sizes of fluorescently-labeled nanoparticles to characterize the structural and barrier properties of undiluted, non-ovulatory CVM [ 13 ]. Commercially available carboxylate-modified polystyrene nanoparticles were either densely coated with low molecular weight polyethylene glycol (PEG) or used without the PEG coating. The diffusional rate of uncoated nanoparticles of all sizes was much lower in CVM than in water, reflecting adhesive interactions between the mucus components and the hydrophobic nanoparticles. In contrast, the PEG-coated nanoparticles as large as 500 nm in diameter diffused in CVM at rates approaching that of what would be expected for free diffusion in water [ 13 ]. Using these so-called “mucus-penetrating particles” (MPPs) in the size range of 100 nm – 1 μm, the diffusional mobility assessed by MPT was used to estimate the effective mucus pore size in undiluted, non-ovulatory CVM [ 125 ]. It was found that the average diameter of the pores was much larger than previous estimates, 340 ± 70 nm, with a range of 50 – 1800 nm. To compare to the prior studies in ovulatory cervical mucus, it was confirmed that the mobility of HSV in CVM was highly restricted, whereas similarly sized 200 nm PEG-coated nanoparticles showed highly diffusive mobility. This confirmed that the immobilization of HSV was due to adhesive interactions with the mucus components, and not because of the virus size and steric hindrance [ 125 ]. Similarly, other studies using undiluted, non-ovulatory CVM described adhesive trapping of fluorescently-labeled HIV virions [ 132 ] and C. trachomatis bacteria [ 133 ]. However, neutralization of the sample pH to mimic mixing with seminal plasma reduced the adhesivity of the pathogens to the mucus mesh [ 132 – 134 ]. Further, reduced adhesive barrier properties of CVM was described in samples collected from women with bacterial vaginosis (BV) and some strains of L. iners, potentially due to enzymatic degradation of mucins by bacterial products [ 11 , 135 ]. These observations are not only important fundamental characterizations of mucus as a function of pH and the vaginal microbiota composition but give key formulation criteria for how to design nanoparticles that can effectively penetrate CVM and deliver drugs or nucleic acids to the underlying epithelia. Alterations in the vaginal microbiota and changes in pH are not the only factors that can affect the structural and barrier function of mucus. The introduction of polymer materials and other excipients can also impact mucus barrier function, which is important to consider when designing vaginal drug delivery systems. The Saltzman group analyzed the mechanical properties and structure of ovulatory cervical mucus after adding polymers such as PEG, polyacrylic acid (PAA), and polyvinyl pyridine (PVP) [ 117 , 136 ]. They found that while the polyanionic PAA had little effect on mucus structure as assessed by SEM, nonionic PEG and cationic PVP caused structural changes [ 136 ]. They then went on to show that the alterations in mucus structure caused by the addition of PEG and PVP increased the mobility of migrating monocytes [ 117 ]. Cell migration in mucus may have an impact on susceptibility to some types of infection, such as HIV. Ensign et al. used MPT with non-ovulatory CVM to investigate the effect of mucus pretreatment with Pluronic F127, a tri-block copolymer consisting of a poly(propylene oxide) flanked on either side by a poly(ethylene oxide) chain, on the mobility of nanoparticles [ 134 ]. They observed that CVM pretreatment with Pluronic F127 generally increased the mobility of uncoated 200 nm carboxylate-modified polystyrene nanoparticles but did not appear to affect the adhesive trapping of HSV or affect the CVM pore structure as inferred by the unaltered mobility of 500 nm PEG-coated particles. Thus, incorporation of Pluronic F127 in CVM may improve nanoparticle penetration without reducing other aspects of the barrier function of CVM [ 134 ]. Alternatively, it has been described that the barrier function of CVM to pathogens and particles can be increased by the use of antibodies. Early studies demonstrated that high levels of antisperm antibodies prevent sperm from moving through cervical mucus, an effect mediated by the Fc region acting as a dock to the mucin proteins [ 137 ]. Antibodies have slightly lower rates of diffusion in ovulatory cervical mucus than saline as determined by MPT, potentially due to weak affinities between the Fc region and the mucin network [ 46 ]. Antibodies can be exploited to enhance the barrier properties of CVM as a protective mechanism. For example, the addition of anti-HSV IgG antibodies to neutralized CVM restored the adhesive trapping, leading to prevention of infection in a mouse model of HSV-2 infection [ 9 ]. In the context of drug delivery, it was also demonstrated that the addition of anti-PEG antibodies to CVM could reduce the mobility of PEG-coated nanoparticles [ 138 ]. Whether the generation of anti-PEG antibodies at mucosal surfaces could result from repeated mucosal exposure to PEG-coated nanoformulations and impact the lasting effectiveness has yet to be determined. During pregnancy, the cervical mucus condenses and forms a cervical mucus plug (CMP) that obstructs the cervix during until just before labor when it’s shed [ 139 , 140 ]. This dense CMP structure forms a physically and immunologically protective barrier between the womb and the vaginal canal [ 119 , 141 ]. SEM studies demonstrated a tightening of the mucin meshwork in the CMP as pregnancy progressed [ 142 ]. However, the Ribbeck group has described differences in structural and barrier function of the CMP in women at high risk of preterm birth [ 143 , 144 ]. Using extensional and shear rheology, they demonstrated that cervical mucus collected from women in preterm labor (high-risk for preterm birth) formed a weaker gel than cervical mucus from pregnant women at low risk of preterm birth ( Fig. 6 ) [ 143 ]. SEM images reflected thinner and more collapsed mucin filaments in cervical mucus collected from women at high-risk of preterm birth, which correlated with increased permeability to biotinylated 200 nm polystyrene particles as assessed in a bead translocation assay [ 143 ]. Similarly, reduced diffusional mobility of 1 μm carboxylate-modified polystyrene particles was observed by single particle tracking in cervical mucus obtained during pregnancy compared to ovulatory cervical mucus samples, suggesting smaller pore sizes within the mucus plug [ 144 ]. Further, the penetration distance of fluorescently-labeled peptide probes in a microfluidic channel was significantly reduced in cervical mucus collected from women at low-risk for preterm birth, whereas peptide penetration in cervical mucus from women at high-risk for preterm birth was similarly high compared to that of ovulatory cervical mucus [ 144 ]. Together, these data highlight that cervical mucus from women at high-risk for preterm birth shows increased permeability, and thus may be less of a barrier to pathogen and microbe translocation to the uterus. Such information is not only important when considering methods for characterizing interactions between mucus and vaginal drug delivery formulations, but also the potential effect that components of the formulations on mucus structure and barrier function.

A wide range of diseases and conditions affecting the FRT are best treated using vaginal products. Local administration can lead to increased delivery to target cells and tissues with decreased systemic side effects. Efforts toward increasing the effectiveness and biocompatibility of vaginal drug delivery systems requires the use of model systems that demonstrate the potential for clinical utility. The development and selection of useful model systems requires consideration of the anatomical structure of the FRT, the mucus and tissue barriers, and the interplay with hormones and the vaginal microbiota. In vitro and ex vivo approaches outlined here include simulated vaginal fluids, cell lines, animal tissues, multilayered cell culture systems, organoids, and isolated mucus secretions. In some cases, in vitro and ex vivo characterizations have been reliable at predicting in vivo behavior. Thus, when selected appropriately, in vitro and ex vivo models provide a way to screen and compare larger numbers of formulations to reduce the need for preclinical animal testing. In other cases, disparate results have been apparent between model systems, highlighting the need for care in model selection and interpretation. Generally, using several complementary methods and observing consistency in outcomes can increase confidence in interpretation. Further, as three-dimensional, layered cell co-culture systems become more advanced, incorporating features of the mucus barrier, the microbiota, and the immune system, these models will further increase the predictive power of in vitro assessments.

The vaginal route of administration is often preferred for locally treating diseases and conditions affecting the female reproductive tract (FRT). Local administration has the potential for increasing drug absorption and delivery to target tissues while minimizing off-target side effects [ 1 – 3 ]. In addition to direct access to the vaginal and cervical tissues, vaginal drug administration can also lead to increased drug delivery to the uterus via direct transport through the local vasculature, a phenomenon referred to as the uterine first pass effect [ 4 , 5 ]. Thus, vaginal drug delivery systems can be optimal for treating and preventing vaginal infections, treating cervical intraepithelial neoplasia, and delivering hormones for hormone replacement therapy or to support assisted fertility procedures. However, vaginal drug administration requires overcoming numerous obstacles to achieve effective drug absorption and retention. The vaginal epithelium is highly folded and collapsed due to intraabdominal pressure, which can preclude uniform drug distribution. Further, the continuously secreted mucus and dynamic fluid regulation systems that clean, protect, and lubricate the reproductive tract epithelia help remove pathogens and foreign substances [ 6 – 9 ]. In the FRT, mucus is secreted by the endocervix and enters the vaginal canal where it mixes with vaginal fluids, shed vaginal epithelial cells, and microbiota. Thus, the material collected from the vagina is often referred to as cervicovaginal mucus (CVM), reflecting its origin from the cervix [ 10 – 12 ]. As the first line of defense, CVM can also act as a steric and adhesive barrier to effective drug delivery, and thus, drug delivery systems must be optimized to bypass the mucus barrier [ 3 , 13 ]. Here, we focus on in vitro and ex vivo models that have been developed to evaluate vaginal drug delivery systems, as well as opportunities for new research and method development. Vaginal products may come in the form of rings, films, gels, ovules, creams, or liquid formulations. While largely studied pre-clinically, nanoparticles have also made their debut in a line of dendrimer-based products for treating bacterial vaginosis (BV) and anti-viral applications [ 14 , 15 ]. When developing formulations for vaginal drug delivery, interactions with CVM and the underlying epithelia must be considered. For particulate systems, these interactions are influenced by particle size, surface characteristics, stability, and formulation composition. In vivo validation in animal models is uniquely challenging with vaginal products, as there are numerous differences in FRT anatomy, microbiota composition, and hormonal cycles between animal species and humans. Thus, there is high value in developing in vitro and ex vivo models for formulation characterization, including in vitro human cell culture models and ex vivo human mucus models. Herein, we discuss model systems for evaluating vaginal drug delivery systems, as well as highlight methodologies that have been developed for other mucosal surfaces that can serve as inspiration for development of model systems of the FRT.

The most useful model systems provide predictive or correlative value for in vivo performance of drug delivery systems. In many cases, a combination of in vitro and/or ex vivo models are used to motivate and complement in vivo studies. In the case of vaginal drug delivery systems, interactions with mucins in vitro and/or mucus secretions or tissues ex vivo are characterized in addition to in vivo animal studies. For example, Cu et al. characterized the diffusional mobility of PEG-modified PLGA nanoparticles (170 ± 57 nm) in ovulatory human cervical mucus [ 145 ]. PLGA particles were coated with avidin, which facilitated coating with various molecular weight (2, 5, 10 kDa) biotinylated PEG. Using ovulatory cervical mucus in a capillary tube, they demonstrated that the theoretical degree of PEG coating (10 % vs 100 % of available binding sites) and the PEG molecular weight affected the particle diffusion [ 145 ]. They went on to show that PLGA nanoparticles with a 2 kDa PEG coating (PEG-NP) (150–170 nm) that showed increased mobility in human cervical mucus also showed greater vaginal retention at 6 h after vaginal administration in mice compared to PLGA nanoparticles and avidin-coated PLGA nanoparticles without the PEG coating [ 146 ]. However, the mice were pretreated with progestin to synchronize the estrous cycle, and the vaginal tract was extensively washed to remove the luminal mucus content [ 146 ], making comparisons to the nanoparticle characterization in human cervical mucus less clear. Das Neves et al. published a series of articles describing the in vitro , ex vivo , and in vivo assessment of dapivirine-loaded nanoparticles (180–200 nm) for vaginal and rectal HIV-prevention. As described previously in section 3 , it was found that neutral (PEO-PCL) and negatively-charged (SLS-PCL) polymeric nanoparticles showed improved stability and mobility in SVF, whereas the positively-charged CTAB-PCL particles were largely immobilized by mucins [ 68 ]. However, permeation of dapivirine across porcine vaginal tissue was increased with the CTAB-PCL nanoparticles, which was attributed to increased cellular toxicity [ 108 ]. The potential removal of mucus from the porcine tissue during processing may also be a factor, as adhesion of the CTAB-PCL particles to mucus would be expected to reduce the drug permeability. Further, the SLS-PCL nanoparticles showed evidence of toxicity when incubated with porcine vaginal tissue. In contrast, the PEO-PCL nanoparticles showed no signs of toxicity in cell culture or during incubation with porcine tissue [ 108 ]. Thus, follow-on studies aimed to characterize the biodistribution and pharmacokinetics of dapivirine after vaginal administration of PEO-PCL nanoparticles in Depo-Provera treated mice [ 147 ]. In comparison to a dapivirine suspension, the PEO-PCL nanoparticles provided increased delivery to the vaginal and lower uterine tissues and reduced systemic exposure [ 147 ]. These studies highlight the importance of selecting the appropriate in vitro and ex vivo model systems and the need for careful interpretation when experimental outcomes are disparate across model systems. Ensign et al. characterized the mobility of non-biodegradable and biodegradable nanoparticles with and without PEG coatings in ex vivo human and mouse CVM [ 97 ]. It was found that mice in the estradiol-dominated estrus stage of the cycle, whether naturally cycling or induced with exogenous estradiol injection, had CVM properties more similar to non-ovulatory human CVM. Increased mobility of densely PEG-coated particles (mucus-penetrating particles or MPP, 112 ± 3 nm) in ex vivo mucus was correlated with more uniform and longer-lasting retention after vaginal administration in mice. This was found using an ex vivo approach, where particles were administered vaginally, the vagina was excised, and MPT was performed on particles in excised tissue ex vivo . About 60 % of vaginally administered MPPs remained in the vaginal tract after 6 h compared to 10 % of the mucoadhesive, uncoated conventional nanoparticles (CPs, 87 ± 4 nm). This understanding was used as a precursor to drug loaded MPPs in murine disease models. Acyclovir-loaded MPPs (65 ± 10 nm) provided protection against vaginal HSV-2 infection in 53 % of mice, whereas only 16 % of mice remained uninfected after free drug treatment at the same concentration [ 97 ]. Similarly, paclitaxel-loaded PEG-coated PLGA nanoparticles (PTX/MPP, 239 ± 5 nm) showed increased mobility compared to conventional paclitaxel-loaded nanoparticles (PTX/CP, 234 ± 4 nm) in human CVM ex vivo [ 148 ]. The non-adhesive PTX/MPP showed more uniform vaginal distribution and proximity to GFP-expressing cervical tumor cells in the cervicovaginal epithelium in progestin-treated mice, leading to enhanced cervicovaginal retention of paclitaxel compared to locally or systemically administered Taxol. Further, vaginal PTX/MPP more effectively suppressed tumor growth and prolonged survival in mice inoculated with TC-1 tumors compared to PTX/CP and Taxol [ 148 ]. Lechanteur et al. described PEGylated lipoplex formulations to deliver siRNA as a treatment for HPV-induced cervical lesions [ 149 ]. The authors address the “PEG dilemma”, which refers to the fact that PEG-coated particles often show decreased cell uptake in vitro , but show improved stability and transport through biological matrices, such as blood and mucus. They used a lipid film hydration approach to produce lipoplexes with and without 20 % PEG-containing lipids, and observed that the PEG-lipoplexes (199.7 ± 14 nm) were more stable than the lipoplexes in SVF containing 1.5 % porcine gastric mucin. Further, despite the fact that decreased cell uptake is often observed with PEG-coated particles in vitro , they observed increased penetration and siRNA delivery in 3D cervical lesion organoids by the PEG-lipoplexes compared to the lipoplexes, reflecting the importance of the tissue barrier. Lastly, they demonstrated that the PEG-lipoplexes provided uniform siRNA signal without signs of toxicity in sections of mouse vaginal tissue without hormonal treatment to arrest the estrous cycle or pretreatment to remove mucus [ 149 ]. In related work, Xu et al. developed a tunable biodegradable nanoparticle system using PLGA-PEG block copolymers to evaluate the effect of PEG density on vaginal drug delivery [ 150 ]. Blending of PLGA-PEG and PLGA in the formulation process resulted in nanoparticles with varying degrees of PEG surface density (0–25 % target PEG content, ~100–150 nm in size). NMR methods were used to determine both the total PEG content and the surface PEG content to allow for calculating PEG surface density. It was shown that PLGA-PEG nanoparticles with 5 % or higher target PEG content showed reduced binding to purified bovine submaxillary gland mucins in vitro , which correlated with increased particle mobility in human CVM ex vivo . Similarly, PLGA-PEG nanoparticles with 10 % and 25 % PEG content showed more uniform cross-sectional vaginal distribution in mice compared to particles with 0 % and 3 % PEG content, with the PLGA-PEG with 25 % target PEG content showing the highest surface coverage on flattened mouse vaginal tissue at 82 %, compared to only 28 % coverage by the PLGA particles without PEG ( Fig. 7 ) [ 150 ]. Yu et al. investigated the impact of PEG content on vaginally delivered liposomes containing MRI contrast agents [ 151 ]. Liposomes were formulated containing 1,2-disteoroyl- sn - glycero -3-phosphatidylcholine (DSPC), cholesterol, and 0–12 M percent of 1,2-distearoyl- sn -glycerophosphoethanolamine poly(ethylene glycol)2000 (DSPE-PEG2k). The liposomes with no DSPE-PEG2k (129 ± 18 nm) showed high polydispersity index, whereas the liposomes containing DSPE-PEG2k (~130–150 nm) were relatively uniform in size. Using multiple particle tracking in freshly obtained human CVM, they observed that the trajectories of liposomes without DSPE-PEG2k were largely adhesively immobilized or hindered. While the PEG-containing liposomes showed increased mobility in CVM, liposomes containing [notdef] 5 mol%-PEG showed decreased mobility over incubation time, suggesting instability. When administered intravaginally, fluorescently labeled 7 mol%-PEG liposomes provided increased vaginal tissue surface coverage compared to 0 mol%-PEG liposomes and 3 mol%-PEG, potentially due to the predicted brush-like PEG conformation on the surface of the 7 mol%-PEG liposomes. However, the 12 mol%-PEG liposomes, which may have been expected to perform similar or better to the 7 mol% liposomes in vivo , actually showed decreased vaginal tissue coverage compared to the 7 mol%-PEG liposomes. The authors speculate that the higher PEG content of the 12 mol%-PEG liposomes may have led to instability and disassembly via micellization in vivo . They last confirmed that the 7 mol%-PEG liposomes loaded with MRI contrast agent showed increased and longer-lasting MRI signal in the mouse vagina compared to unencapsulated contrast agent [ 151 ]. These works highlight the importance of PEG density on nanoparticle mobility in CVM, as well as the potential for correlation between in vitro , ex vivo , and in vivo model systems. With a focus on the development of vaginal drug delivery systems for use in pregnancy, Hoang et al. characterized particle movement in CVM from pregnant women across the three trimesters [ 49 ]. They observed that densely PEG-coated polystyrene nanoparticles as large as 200 and 500 nm in size were rapidly mobile in CVM, whereas uncoated polystyrene particles were adhesively immobilized. Further, larger microparticles (2 μm poly-styrene) or particles coated in polycarbophil, mimicking the characteristics of the micronized progesterone in the clinically tested Crinone ® vaginal progesterone cream, were adhesively immobilized in CVM from pregnant women. Thus, they went on to develop a mucoinert progesterone nanosuspension (NS) for improved vaginal progesterone delivery in a mouse model of preterm birth (PTB) induced by of progesterone-withdrawal. The final mucoinert progesterone NS was ~ 260 nm in size and coated with Pluronic F127 and provided increased delivery of progesterone throughout the reproductive tract tissues after vaginal administration compared to Crinone. In an RU486 mouse model of PTB, vaginal administration of the progesterone NS resulted in 55 % of mice going to full term, compared to 32 % of mice treated with Crinone and 15 % of untreated control mice [ 49 ]. In later work focused on inflammation-induced PTB, Zierden et al. found that progesterone alone was insufficient as a preventative treatment [ 152 ]. However, formulation of a mucoinert trichostatin-A (TSA) nanosuspension (215.9 ± 12.9 nm) and vaginal dosing in combination with a progesterone NS (305.9 ± 5.8 nm) in an adapted intrauterine inflammation-induced mouse model of PTB led to a significant increase in mice that went on to deliver live pups at term [ 152 , 153 ]. These works highlighted the potential for new vaginal delivery systems in the prevention of PTB [ 154 ]. Vaginal drug-loaded films are thin sheets that have potential advantages in their ease of application and low water content for stability during storage [ 155 ]. As the drug delivery properties are dependent on the rehydration and retention in the vagina, films are typically characterized through a series of in vitro rehydration tests and ex vivo tissue tissue adhesion tests that culminate in in vivo assessments in rabbits, mice, or macaques [ 155 ]. Li et al. developed bioadhesive vaginal films comprised of thiolated chitosan for simultaneous delivery of dapivirine and levonorgestrel [ 156 ]. Dapivirine release from the thiolated chitosan film was more sustained compared to a quick dissolving film in vitro . The ex vivo adhesional force to pig intestinal tissue was 9.7-fold higher for the thiolated chitosan film compared to the quick dissolving film. Dye-loaded films were applied vaginally in macaques, demonstrating that the bioadhesive thiolated chitosan film was visualizable for up to 7 days, whereas the quick dissolving film was no longer evident by day 3. The bioadhesive film was shown to deliver dapivirine and levonorgestrel individually or simultaneously to macaque reproductive tract tissues for at least 7 days [ 156 ]. Tyo et al. developed composite electrospun fibers containing Griffithsin-loaded nanoparticles to provide sustained release [ 157 ]. They first determined the anti-HIV activity and confirmed that the fibers did not cause cell toxicity in vitro . The antiviral activity was similar for the nanoparticles whether or not they were embedded in electrospun fibers, suggesting that the processing did not negatively impact the drug-release properties. They then tested the composite nanofibers in a murine model of lethal HSV-2 infection and demonstrated 70–80 % survival in mice treated vaginally with the composite nanofibers, compared to 5 % in the control group. Further, there was no apparent tissue toxicity in the mice [ 157 ].