While explant cultures are currently used to study HIV pathogenesis and transmission in tissues relevant to in vivo sites of infection, they are also being used to evaluate the antiviral activity of candidate topical microbicides prior to subsequent studies in animals and humans. Given the variety of protocols currently used in tissue explant studies (e.g., see references 1
, and 12
), the first aim of this study was to identify a uniform method for analyzing, comparing, and presenting virus growth curve data. To this end, the SOFT metric provides such a method. SOFT was developed to enable interlaboratory comparisons of HIV-1 replication across explant methods (with cervical, rectal, and tonsil tissues) by determining the endpoint to compare virus growth between control and microbicide-treated tissues. Comparison of virus growth between control and treated tissues at time points prior to the accelerated growth phase or after virus growth has tapered off could underestimate the level of virus inhibition exhibited by a microbicide candidate. Comparison of virus growth at SOFT provides an “ideal” endpoint comparison for each growth curve by excluding the variability in measurements at time points prior to SOFT, when growth in the virus control may be at low or nondetectable levels, or after SOFT, when the system may no longer be able to support viral growth. SOFT aims to provide a sensitive indicator of drug efficacy to be used early in a drug screening program, prior to additional confirmatory testing, when the risks of making a type II error (i.e., false-negative result when a potentially efficacious compound is excluded) outweigh the risks of a type I error (i.e., false-positive result when a potentially ineffective compound is included).
An evaluation of calculation time, ease of performance, availability of resources, and other factors that may be important to a laboratory should be considered prior to the adoption of any endpoint method. Although the AUC and slope calculations proved to be effective predictors of virus growth, the complexity of these calculations may reduce their utility for research teams without access to statistical software necessary to compute nonlinear regressions and trapezoidal functions. In contrast, SOFT can readily be calculated and thus allows direct comparison of growth curves by standard statistical tests (e.g., ANOVA, t test, Kruskal-Wallis test, and Mann-Whitney test). A multivariate comparison (LDA) of various endpoint approaches further supported the application of SOFT as an objective method for evaluating virus replication. In addition, SOFT neither over- nor underestimated virus growth compared to other endpoints and provided a measure of growth that reliably reflected the virus yield of the assay.
An added advantage to the identification of a robust assay endpoint is the ability to use it for identifying key parameters that, once standardized, can lead to more reproducible data between laboratories. As a second aim of this study, SOFT was used to compare conditions that could affect virus growth in explant assays. The variables found to affect virus growth in the explants studied here included (i) source/type of assay reagents, (ii) tissue type, (iii) how the tissue was cultured (i.e., tissue stimulation), (iv) HIV-1 isolate used, and (v) microbicide concentration. In contrast, laboratory measurement of p24 protein in culture supernatants was found to be generally consistent, regardless of assay-specific factors such as ELISA manufacturer, standard curve range, method of ELISA analysis, or the laboratory running the analysis.
Since HIV-1 replication can be monitored in tissue culture by a variety of methods (including measurement of p24 protein), different patterns of virus growth can be observed depending on whether virus is measured in the cell-free or cell-associated compartments of cultured cells and tissues (3
). In the microbicide field, researchers rely primarily on p24 protein levels in culture supernatants as a measure of virus replication over the course of the explant study. Using this approach for the current study, greater viral replication was observed in all explant models by use of the centrally provided VQAB stock and the laboratory's own IHM formulations. Use of a common virus stock provided by a central repository increased p24 production for all tissue types and methods, but medium formulations routinely used by each laboratory were found to be better suited to the laboratory's preferred explant method.
Perhaps the largest effect on p24 production was the tissue type and model used. For example, the least viral replication was observed using unstimulated cervical tissue. When cervical tissue was immune stimulated or cocultured with PBMC, viral replication was noticeably increased, by 10- or 1,000-fold, respectively. Use of either rectal or tonsil tissue also significantly increased viral replication, suggesting that viral replication can be significantly affected by endogenous stimulation or the concentration and distribution of target cells present in each tissue type or model system.
The absolute quantity of p24 protein produced by tissue explant cultures did not appear to correlate with the ability to detect a drug effect. The candidate microbicide PRO 2000 was found to inhibit virus growth for the majority (four of five methods) of explant methods used by the seven laboratories (i.e., cervical, tonsil, rectal, and cervical tissue-PBMC coculture explants), where p24 production at SOFT ranged from 14 to 670,300 pg/ml. Cultures releasing both high (tonsil) and low (cervical) p24 levels produced reliable drug effects, while another high-p24-release explant model (stimulated cervical tissue) failed to demonstrate any drug effect in this study.
Given the lack of a standard endpoint for virus growth in explant studies, few laboratories to date have reported microbicide activity in terms of an IC50
. In fact, only one laboratory has published an IC50
for PRO 2000 in cervical tissue, at ~80 μg/ml (9
). In the current study, the IC50
for PRO 2000 varied widely among the tissue types evaluated (9.7 to 98.0 μg/ml) but remained fairly consistent within an explant method, as demonstrated by the three laboratories working with cervical tissue (24.5 to 29.3 μg/ml). These data indicate that inhibitory concentrations of PRO 2000 in explant tissues are higher than those in cell lines and primary T cells or macrophages (IC50
s ranging from 0.7 to 12.8 μg/ml) (20
), suggesting that more drug is required to penetrate tissue than to penetrate monolayers or cell suspensions. Comparing the same tissue type used in different model systems, such as cervical culture and cervical tissue-PBMC coculture, PRO 2000 was also consistently active (24.5 to 29.3 μg/ml compared to 28.1 μg/ml, respectively), but not compared with the stimulated cervical (IC50
could not be determined) or rectal (9.7 μg/ml) model. This suggests that drug IC50
s could be compared between laboratories using the same tissue type under the same model conditions, but not necessarily between laboratories using either different explant tissues or different assay methods.
Although the HIV-1 isolate used had a significant impact on virus growth (clade C strains resulted in less virus growth and clades A and A/G showed comparable growth to that of HIV-1Ba-L
), it is important to note that the explant models examined in this study supported growth of viruses that represent only a subsample of isolates available to microbicide researchers (https://www.aidsreagent.org/Index.cfm
). These results support the inclusion of non-HIV-1Ba-L
isolates in future explant studies by determining microbicide activity against more clinically relevant, non-laboratory-adapted strains of HIV-1.
While an intra-assay variability of <20% is generally considered acceptable for in vitro assays (24
), this study found a wide range for explant intra-assay %CVs (0.8 to 163.4%), where rectal tissue measurements were the most variable. High variability in rectal explant tissues may be due to such factors as susceptibility of different tissue areas to infection (immune cell pockets, areas with large M-cell populations, etc.) or friability of tissue that is of a less structured type (1
). However, even though the high intra-assay variability reduced the statistical power of treatment comparisons, consistent drug and isolate effects were found. Additionally, statistical power was not consistent across explant methods; cervical explants used by three laboratories provided greater statistical power for drug and isolate effects than did other explant methods tested by only a single laboratory (i.e., stimulated cervical, cervical tissue-PBMC coculture, tonsil, and rectal methods).
In summary, in order to improve the quality of data from laboratories using tissue explants to evaluate topical microbicides, this study addressed the following two aims: (i) identification of a uniform method for analyzing, comparing, and presenting virus growth curve data; and (ii) identification of those key parameters that, once standardized, will lead to more reproducible data between laboratories. A comparison of various endpoint approaches demonstrated that the SOFT can provide a measure of virus growth that is easily calculated, reliably indicates the virus yield of the assay, and allows direct comparison of growth curves by standard statistical tests. In terms of key assay parameters, use of a common virus stock was shown to be most important for improving virus growth among the tested explant models. In addition, as the data with primary strains indicated, identification of clinically relevant HIV-1 clades that are able to replicate well in tissue explants will be important for future microbicide studies. Although the tissue type and model used had the greatest impact on virus growth, this study also demonstrated that a reliable drug effect can be observed in tissues with low or high p24 protein levels and that IC50s can be compared between laboratories using the same tissue type. Finally, wide intra-assay variability may be inherent to tissue explant studies, where the extent of measurement variability found here depended upon the tissue type used. Further studies using a variety of tissue types are recommended to determine the number of donors and replicates necessary for adequate statistical power in comparisons of HIV inhibition between compounds.
Reproducible tissue explant research has the potential to provide an early indicator of drug efficacy, thus accelerating the best microbicide candidates into subsequent animal studies and, ultimately, human clinical trials. It is hoped that the identification of a reliable endpoint and other key methodological parameters will increase future standardization and comparability of preclinical explant assays.