Excipients are often not inactive
Excipients are thought of as "inactive ingredients" but often without evidence that they are inactive. Although excipients can serve important functions ancillary to active pharmaceutical ingredients (APIs) in topical microbicide products, they bear their own potential toxicities. We studied a broad range of excipient classes, namely, preservative, antioxidant, chelating, humectant, solublizing, and emulsifying excipients commonly used in vaginal  and other topical products in an assay that determines their impact on susceptibility to infection with HSV-2. We tested excipients in vehicles (i.e., HEC placebo gel; PBS solution) known to be free of toxicities that increased susceptibility , and also undiluted (neat) excipients, thus in isolation from any API, thereby preventing antiviral activity by the API from masking excipient-induced increases in susceptibility. Although it could be argued that an effective API would make excipient-induced increases in susceptibility irrelevant, in practice imperfect microbicide adherence, the unpredictable timing of intercourse, and the prolonged duration of altered susceptibility  are such that it cannot be assumed that effective levels of an API will be present throughout the period of increased susceptibility. Moreover, the API may not protect against all pathogens to which it or its excipients increase susceptibility.
Preservatives might be expected to be of particular concern since these agents kill or inhibit a broad range of microorganisms and may also have similar cellular toxicity. However, neither parabens nor benzyl alcohol caused a detectable increase in susceptibility in our model, despite an apparent association of benzyl alcohol with irritative symptoms in a Phase I study of cellulose sulfate . Sorbic acid had previously been tested in our initial studies with this model  without detectable increase in susceptibility. In the present studies, we used the sorbic acid-preserved HEC placebo gel [16, 17] as a vehicle for tests of other excipients, and thereby again confirmed the lack of increased susceptibility with sorbic acid and the HEC placebo gel vehicle when compared repeatedly to a PBS control. These findings also match the benign clinical safety profile and lack of effect on HIV transmission of the HEC placebo in an HIV efficacy trial, where it was directly compared to a no-gel arm .
EDTA is a chelating agent used as a preservative synergist at 0.01-0.1% and as an antioxidant synergist at 0.005 to 0.1% [21, 25]. We studied disodium EDTA at two concentrations, 0.0186% and 0.1%. At 0.0186% (~ 0.5 mM) disodium EDTA showed no effect on susceptibility to HSV-2. However, at the higher end of the typically used concentration range (0.1%), there was a trend toward increased susceptibility to HSV-2 challenge (P = 0.095). Since EDTA has long been used to detach mucosal epithelia in vitro , the trend toward harm at the upper end of the typically used EDTA concentrations, though not statistically significant, raises concern about the safety of disodium EDTA at or beyond 0.1%.
Low molecular weight highly soluble organic compounds are commonly used as "humectants" in dermatological formulations and sexual lubricants to reduce evaporation-induced coldness and premature drying when aqueous formulations are spread over the skin, as well as for their lubricant character. The commonly used humectants glycerin, propylene glycol, and PEG-8 are also employed for their solvent and/or preservative-enhancing characters . The slug mucosal irritation assay reported toxicity (irritation and damage) with high osmolality preparations, including Astroglide® (containing glycerin and propylene glycol, osmolality ~5800 mOsm/kg), and an HEC gel with glycerin added at both 20 and 40% and an osmolality of 2200-4400 mOsm/kg) . Several clinical studies likewise reported that other strongly hyperosmotic formulations disrupt the columnar epithelium of the rectum [27–29]. In our studies, the moderately hyperosmotic formulations of 10% glycerin and 10% propylene glycol did not show a significant increase in susceptibility, though 10% glycerin showed a trend toward, and 30% glycerin caused a significant increase in susceptibility. Moreover, consistent with the reports of toxicity with the extremely hyperosmotic formulations cited, we found that an extremely hyperosmotic vehicle, KYWJ (osmolality > 10,000 mOsm/kg) caused a 7-fold increase in susceptibility to HSV-2 (Table 2). Moreover, each of the primary constituents of this formulation, the humectant/solvents propylene glycol and PEG-8, greatly increased susceptibility. Both are markedly hyperosmotic, in the range of the complete KYWJ formulation, and their osmolality may mediate the toxicity. However, both are not only humectants, but also solvents, and their toxicity may additionally or alternatively be due to solvent properties or other characteristics. Indeed, propylene glycol has been demonstrated to have in vitro cellular toxicity at relatively low concentrations, independent of osmotic effects .
GML and the hyperosmotic K-Y Warming Gel vehicle
Since our previous work with this model demonstrated very large increases in susceptibility after exposure to a diverse range of surfactants , and to the non-surfactant but membrane-active chlorhexidine , we tested the excipient glycerol monolaurate (GML). GML has surfactant, emulsifier, membrane-active, and penetration-enhancing actions. Moreover, GML is currently being studied as a novel microbicide API , hypothesized to act by down-regulating the activation, recruitment, and accumulation of HIV target cells . GML is poorly soluble in water, and GML was formulated in the cited and present studies using the non-aqueous lubricant KYWJ as a vehicle, composed primarily of the low molecular weight humectant/solvents propylene glycol and PEG-8.
Using an inoculum of 1 ID50, we found that 5% GML formulated in KYWJ, caused a 10-fold increase in susceptibility to HSV-2, somewhat greater than the increase also observed after KYWJ vehicle alone. The rate of infection with 1 ID50 approached 100%, limiting an accurate estimation of the magnitude of susceptibility increase. Therefore we did additional experiments with a lower inoculum (0.1 ID50) and the results indicated the susceptibility increased 10-fold with GML in KYWJ and 7-fold with KYWJ alone. In light of the toxicity associated with KYWJ, we tested 5% GML without this vehicle, prepared as a colloidal suspension in PBS. Although GML formulated in PBS was nearly isotonic, it too significantly increased susceptibility (P < 0.006), indicating that both KYWJ and GML individually caused susceptibility-increasing toxicity.
The susceptibility-increasing toxicity of GML may be due to its surface-active and membrane-active properties. Indeed its effects on toxin production and signaling in bacteria and immune cells have been postulated to be due to intercalation of GML into cell membranes [31, 32]. In light of the increase in susceptibility caused by a wide diversity of surfactant types in our prior study , and the surfactant nonoxynol-9 in a similar model , the increased susceptibility observed after exposure to GML is perhaps not unexpected. Yet it is notable that neither colposcopic nor histological abnormalities were detected after 6 months of daily vaginal administration of 5% GML in KYWJ in rhesus macaques . However, in our prior studies of the surfactant nonoxynol-9, colposcopy was normal at the time of maximally increased susceptibility to HSV-2, twelve hours after exposure . Moreover, our studies evaluated surfactant contact with columnar-like epithelium of the medroxyprogesterone acetate-treated mouse vagina since microbicides will contact human columnar epithelium in the endocervix . Exposure to columnar epithelium will also occur on the face of the cervix when cervical ectopy is present. Examination of columnar epithelium (endocervix or ectopy) was not reported after chronic GML exposure in the GML toxicity study in macaques .
The present findings are examples of unexpected actions of excipients, and show that excipients are not necessarily inactive, nor non-toxic in mucosal contact. The previous demonstration of the effect of GML on signaling and toxin production in bacteria , its effect on cell immune cell proliferation , its activity as a penetration enhancer, the recent demonstration of its activity in inhibiting SIV , and the present demonstration of its action increasing susceptibility to genital herpes are examples of substantial activities of an excipient. Likewise the HSV-2 susceptibility-increasing effects of KYWJ, and the prior toxicity documented with another hypertonic lubricant, ID Glide® , also illustrate the potential toxicity of vehicles (lubricant gels) even in the absence of an API.
We tested an intermediate concentration (30%) of the humectant/solvent glycerin because glycerin was toxic in this concentration range in an in vitro model , and is present in the gel formulation used in CAPRISA 004, where 1% tenofovir vaginal gel was reported to provide protection against both HIV and HSV-2. Thirty percent glycerin caused a significant increase in HSV-2 susceptibility in our model.
Choice of challenge dose and estimation of fold-increase in susceptibility
In our previous publication  the surfactants tested caused very large increases in susceptibility, and we therefore used a viral dose of 0.1 ID50 to obtain a more accurate evaluation of the magnitudes of increased susceptibility. With agents of unknown, and perhaps only modest effects on susceptibility, an inoculum size of 1 ID50 is optimal for providing maximum statistical power to detect significant alterations in susceptibility. It is notable that a large number of animals was required to detect even fairly substantial increases in HSV-2 susceptibility (see Sample size calculations in Results). It is also worth emphasizing, as illustrated in Figure 1, it is inappropriate to calculate the magnitude of the increase in susceptibility simply by stating the ratio of the fractions of animals infected in the test and control groups. For example, with a challenge dose of 1 ID50, even an agent that increases susceptibility by 1000-fold could only double the fraction of animals infected. In contrast, human transmission of HSV-2 infections typically occurs with low probability of infection per coital event, implying 'low-dose' viral challenges. In such cases, increased susceptibility as defined here implies the probability of transmission per coital event will likely increase by the magnitude that the agent increases susceptibility; a 10-fold increase in susceptibility would lead to a 10-fold increase in human transmission rate.
Limitations and strengths of this model
Our model has several limitations. Observations were made in mice rather than humans or non-human primates. Medroxyprogesterone acetate-treated mice have a thinned epithelium with living cells on the surface, mimicking columnar epithelium and probably with greater sensitivity to chemical damage or irritation than the multilayer squamous epithelium of the human vaginal epithelium. However, vaginal microbicides contact the columnar epithelium in the endocervix  via the well-documented mechanism of uterine peristaltic uptake [34–36] and also contact columnar epithelium on the ectocervix when cervical ectopy is present. It is appropriate for a screening test to be highly sensitive, and to mimic the exposures that are potentially most damaging. Moreover, vaginal microbicides will very likely be used to attempt rectal protection, where they will also contact columnar epithelium. For both these reasons, a model exposing a thin epithelium with living surface cells is highly relevant.
The challenge virus in this model is HSV-2, whereas HIV is the primary focus of most microbicide development efforts. However, it is plausible that toxicities that result in heightened susceptibility to HSV-2 may also increase susceptibility to HIV. Indeed results from our model with nonoxynol-9 and C31G, and a similar model  with nonoxynol-9, and cellulose sulfate, have correlated well with significantly harmful (nonoxynol-9, cellulose sulfate) or borderline harmful (C31G) clinical trial results. Moreover, HSV-2 is an important pathogen in its own right, and as a cofactor that substantially increases the risk of HIV acquisition. We therefore believe that a microbicide whose API, component excipients, or complete formulation significantly increase susceptibility to HSV-2 in mice cautions against advancing to clinical trials in humans.
In the present study, we investigated only a single time interval between application of the test agent and viral challenge. We elected to use the interval found to be associated with the maximum increase in sensitivity after exposure to the surfactant nonoxynol-9. However, we acknowledge that the timing of maximum susceptibility may vary depending on the agent tested.
Our model employs only a single exposure rather than repeated exposures over time. We chose this exposure protocol because in pilot experiments, multiple exposures with the surfactant microbicide nonoxynol-9 did not result in greater susceptibility than a single exposure (data not shown). In addition, the single-exposure model allows resources to be directed toward increasing the number of animals in groups, possibly providing greater overall sensitivity of the model. However, some agents may only show toxicities after multiple exposures, for example agents that are sensitizers, where inflammatory/immune-mediated toxicities may only be observed after sub-chronic or chronic dosing schedules.
Finally, neither this nor any model can be certain to detect all possible toxicities, and hence lack of toxicity in this model does not guarantee safety in human use. Conversely, when this or similar models detect increases in susceptibility to important STI pathogens, we believe such results should be considered important red flags cautioning against the use of an API, excipient, or completed formulation in the vagina or rectum.
Our model has the following strengths: It directly detects toxicities that result in heightened susceptibility to HSV-2 acquisition, rather than surrogate endpoints that are only postulated to cause increased susceptibility. It thus directly assesses what may be reasonably considered one of the most harmful potential toxicities of a candidate microbicide, a paradoxical increase in susceptibility to a serious sexually transmitted viral pathogen. Importantly, the results in this and similar models  correlate well with trends toward or statistically significant adverse effects of microbicides on HIV endpoints in clinical studies [1–4]. The model is efficient and sensitive, employing genital tract tissue with highly susceptible living surface cells and an optimized inoculum strength. It can use sufficient animals to provide statistical power to detect relatively small changes in susceptibility, increases in susceptibility increases as low as 3-5 fold over controls. We believe it can provide useful guidance on the suitability of APIs and excipients for use in vaginal microbicide formulations.