Sodium acetate, Kolliphor HS 15 (Solutol), isopropanol alcohol, and levonorgestrel (LNG) were purchased from Sigma-Aldrich (Saint Louis, MO). 0.1 % trifluoroacetic acid (TFA) in water and 0.1 % TFA in Acetonitrile were purchased from Fisher Scientific (Hampton, NH). Silicone-based polyurethane resin (SIL 30 resin) was generously supplied by Carbon, Inc. Dapivirine (DPV) was purchased from Selleckchem (Houston, TX), and Pritelivir (PTV) was purchased from MedChemExptress (Monmouth Junction, NJ).

CLIP 3D printing and the SIL 30 matrix provide the unique ability to incorporate APIs post IVR fabrication in a controlled fashion using an absorption process. 3D printed IVRs were loaded with APIs using previously described weight-based linear drug loading equations [44]. The macaque size IVR was allometrically scaled to the human size IVR to elicit equivalent specific surface area (SAA) and equivalent honeycomb unit cell strut thickness to allow for a direct translation of drug loading and release kinetics from the human size MPT IVR to a macaque size MPT IVR. DPV, PTV, and LNG were formulated in macaque size IVRs at equivalent weight percent (wt%) to the human size IVR. A direct comparison of the human and macaque size MPT IVRs is shown in Fig. 2. Dapivirine was loaded at 0.65 % w/w (30.0 mg/human-size IVR; 8.11 mg/macaque-size IVR); pritelivir was loaded at 0.43 % w/w (20.0 mg/human-size IVR; 5.36 mg/macaque-size IVR); and levonorgestrel was loaded at 0.043 % w/w (2.0 mg/human-size IVR; 0.536 mg/macaque-size IVR).

There is no established apparatus or standardized method reported in the literature for testing the release kinetics from IVRs in vitro, and several methods using different dynamic settings and release media have been reported in the literature [63]. Herein, we investigated in vitro release kinetics using both static and dynamic conditions. For the dynamic in vitro release studies, it was determined that shaking at 120 rpm in an orbital shaker elicited similar release kinetics to a USP Dissolution Apparatus (Vision 8 Elite Dissolution Tester, Hanson) set at 60 rpm. As such, in vitro release studies were carried out with human size IVRs in SVF (pH 4.2, 37 °C) under static and dynamic (120 rpm) conditions. As shown in Fig. 3, all three APIs elicited higher initial burst release in the first 24 h under dynamic conditions (4.7 % DPV, 7.4 % PTV, 5.6 % LNG) compared to static conditions (3.6 % DPV, 5.3 % PTV, 4.4 % LNG). However, the daily zero order release rate was not significantly different when comparing static versus dynamic settings. Following the burst release, DPV, PTV, and LNG daily release rates under dynamic setting were 369.9 μg/day, 358.4 μg/day, and 24.4 μg/day respectively. The daily release rate under static conditions was 441.5 μg/day (DPV), 353.3 μg/day (PTV), and 32.2 μg/day (LNG). The cumulative percent release over 35 days for DPV and PTV was ~40 % and 55 % respectively for both the static and dynamic release studies. Whereas LNG elicited higher cumulative percent release at day 35 (~55 %) under dynamic release compared to static release (~40 %).

In vitro release studies of human and macaque size IVRs were conducted at pH 4 and pH 8 SVF to mimic physiological vaginal pH of humans and macaques respectively. Release studies were conducted under static and dynamic conditions to determine the optimal conditions that better correlate in vitro release kinetics to in vivo pharmacokinetics. Results showed that IVRs exhibited similar release profiles regardless of pH or whether tested under static or dynamic conditions. Following the initial burst release, APIs elicited sustained zero order release kinetics from day 2–35. A linear regression was performed for each API under each testing condition.

In vitro release under dynamic conditions (pH 8) elicited higher initial burst release within the first 24 h for all APIs, with 9.96 % (808 μg) DPV, 10.3 % (551 μg) PTV, and 11.2 % (60 μg) LNG in comparison to lower burst release obtained under static (pH 8) conditions with 4.1 % (333 μg) DPV, 5.9 % (319 μg) PTV, and 5 % (27 μg) LNG (Fig. 4A–B, D–E). Following the initial burst, IVRs elicited similar daily release rates under static and dynamic conditions with (135.8–188.8) μg/day, (90.4–121.4) μg/day, and (10.1–12.7) μg/day for DPV, PTV, and LNG respectively (Fig. 4G). On the other hand, IVRs incubated at pH 8 elicited higher initial burst release and higher daily release rates for all APIs compared to pH 4 (Fig. 4B–C, 4E-F).

We hypothesized that the difference in drug release between equivalent IVRs at pH 4 and pH 8 was due to the difference in swelling of the SIL 30 IVR matrix as a function of pH. To test this hypothesis, a swelling study was conducted using solid (no design) human size placebo IVRs. IVRs were placed in conditions that mimic the aforementioned static in vitro release study, where n = 3 IVRs were incubated individually in SVF (pH 4 or pH 8) at 37 °C. At t = 0, 1, 7, 14, 21, and 28 days, IVRs were removed from the media and the mass (mg), outer diameter (OD, mm), and cross section (CS, mm) were recorded to capture the swelling of the SIL 30 matrix in the release media. Results showed that IVRs exhibited higher swelling at pH 4 than pH 8. At t = 0 the average mass, OD, and CS of the IVRs tested at both pH conditions were equivalent. Over the 28-day study, IVRs incubated at pH 4 exhibited a 73 % increase in mass, 20 % increase in OD, and 21 % increase in CS. IVRs incubated in SVF at pH 8 demonstrated about 10× less swelling as demonstrated by the significantly lower changes in IVR metrics (5 % increase in mass, 1 % increase in OD, and 3 % increase in CS) relative to pH 4 (Fig. 5).

To assess the long-term stability of the MPT IVRs, we completed 1, 3, and 6 months stability studies under two storage condition: 4 °C and accelerated stability conditions (40 °C/75 % RH). At each time point, n = 3 IVRs were used to investigate post-storage in vitro release, and n = 1 IVR underwent full extraction in acetone to quantify drug concentration and assess drug stability by HPLC analysis. Results showed that DPV remained stable over 6 months with no change in drug concentration and absence of degradation peaks by HPLC analysis. On the other hand, results showed that the PTV concentration decreased by 50 % after 3 months of storage with no further decrease recorded after 6 months of storage. Similarly, LNG concentration decreased by 50 % after 3 months of storage, with an additional 50 % decrease in concentration recorded after 6 months of storage [64]. Unknown degradation peaks were present in the HPLC chromatogram at 7.7 and 8.7 min (Supplemental Fig. 1). To identify the degradation peaks detected post storage, a series of studies were set up to evaluate whether the degradation was caused by the SIL 30 IVR matrix, individual APIs, or a combination of APIs in the SIL 30 matrix.

First, placebo HC IVRs were stored in a humidity chamber under accelerated stability conditions (40 °C /75 % RH) for 6 months, and subsequently assessed for in vitro release (SVF, pH 4, 37 °C). Sample aliquots (1 mL) were collected on day 1, 7, and 14 based on when the unknown degradation peaks were detected during the in vitro release study post 6-months storage of the MPT IVR. Unlike the MPT IVRs, no unknown degradation peaks were detected with the placebo IVRs in the release samples as determined by HPLC analysis (Supplemental Fig. 1). To assess the APIs’ stability under storage, neat APIs were stored in glass vials individually (n = 3) or in combination (n = 3) and analyzed after 3 and 6-months of storage under accelerated stability conditions (40 °C /75 % RH). At each time point, APIs were dissolved in acetone to mimic the API extraction from an IVR, and samples were analyzed by HPLC. HPLC analysis showed no presence of degradation peaks in these samples (Supplemental Fig. 1). Based on the above results, it was determined that the degradation peaks were a result of API’s interaction with the SIL 30 matrix under accelerated storage conditions.

Previous studies showed that the DPV/PTV/LNG combination in the SIL 30 matrix resulted in presence of degradation peaks after storage under accelerated stability conditions (40 °C/75 % RH) [64]. To investigate the nature of degradation peaks, a comprehensive stability study was carried out using IVR segments. Human size HC IVRs were cut into segments (~400 mg each) as representative units of the whole IVR. IVR segments were batch loaded with each API individually (DPV, PTV, or LNG) or in dual combination (DPV/PTV, DPV/LNG, PTV/LNG). IVR segments (n = 3) were stored at 40 °C/75 % RH or at 4 °C and evaluated after 1, 3, and 6 months of storage (Supplementary Fig. 2, 3). At each time point, IVR segments were removed from their respective storage conditions and underwent a 24 h extraction in acetone. Samples of the extract were analyzed by HPLC to quantify each API and investigate presence of degradation peaks. The weight percent loading of each API in the segments was kept constant by incubating all segments in the same drug solution. To compare drug loading across segments, the weight percent (wt%) of API loaded was determined using the following equation: APIextracted(mg)IVRsegment(mg)×100%=wt%APIpersegment. The wt% of each API loaded individually and in dual combination was compared across the 1, 3, and 6-month time points for both storage conditions (Fig. 6, Table 1, and Table 2).

Results showed that DPV remained stable with absence of degradation peaks when formulated alone or in combination with either PTV or LNG under both storage conditions. However, when formulated in combination with PTV, the weight percent of DPV quantified after 6-months storage under accelerated conditions decreased significantly (0.89 ± 0.03 wt%) compared to 1-month post storage (1.10 ± 0.07 wt %).

PTV did not elicit statistically significant changes in weight percent when stored at 4 °C (Table 1, Supplementary Fig. 3). On the other hand, after 3 months of storage under accelerated stability conditions (40 °C/75 %RH), PTV individually and in dual combination with DPV and LNG elicited a significant decrease in weight percent when compared to the 1-month storage (Fig. 6, Table 2). After 6 months of storage under accelerated conditions, PTV alone and PTV in dual combination with DPV elicited significantly lower weight percent values compared to both the 1 and 3-month storage.

LNG remained relatively stable when stored at 4 °C when formulated alone or in combination with DPV or PTV (Table 2).

Further analysis of the individual and dual-loaded IVR segments with DPV, PTV, and LNG demonstrated that degradation peaks were detected in the DPV/PTV/LNG combination and were primarily generated from PTV degradation post storage at 40 °C/75 % RH. In addition to the observed changes in API wt%, degradation peaks were also observed when PTV was formulated alone or in combination with DPV or LNG (Supplemental Fig. 2). No degradation peaks were present for DPV or LNG when formulated alone. Furthermore, samples of the PTV solutions were analyzed using mass spectrometry analysis to identify a proposed structure for the degradation peaks. Results from the mass spectrometry analysis showed that the degradation peaks were by-products of PTV, most likely from an amid hydrolysis reaction (Supplemental Fig. 4).

Results from the triple-drug combination IVR segments stored at 4 °C for 1, 3, and 6 months are shown in Fig. 7. All three APIs remained stable over 3 months of storage at 4 °C but elicited a significant decrease in drug wt% after 6 months of storage. No degradation peaks were observed by HPLC analysis (Supplemental Fig. 3).

Macaque size IVRs were loaded with DPV, PTV, and LNG and investigated in vivo in pigtailed macaques (n = 3) for safety and pharmacokinetics (PK) over 30 days. Plasma samples, vaginal tissues (proximal and distal to the cervix), and vaginal fluid samples (proximal and distal to the cervix) were collected longitudinally. An illustration of the sampling schema for the PK samples is shown in Fig. 8. At t = 0 plasma and vaginal fluid samples were collected, and DPV/PTV/LNG MPT IVRs were subsequently administered vaginally to each macaque. The IVR remained in place for 28 days and then removed, and final samples were collected on day 30, two days post-IVR removal. At t = 3, 7, 14, 21, 28, and 30 days, plasma and vaginal fluid samples were collected. At t = 7, 21, and 30 vaginal tissue samples were collected.

Results of the PK analysis for each biological sample are summarized in Fig. 9, and the exact weight and API dosing per macaque is shown in Supplemental Table 3. Results showed that the plasma drug concentrations remained constant over the 28 days of IVR use for each API. Median (range) plasma concentrations of API while the rings were present from day 3-day 28 for the group (n = 3) were 138 (110–192) pg/mL for DPV, 18,700 (7550–49,000) pg/mL for PTV and 335 (125–1720) pg/mL for LNG (Fig. 9A–C, Supplemental Table 4). At day 0 prior to IVR insertion and at day 30, two days post-IVR removal, the concentrations of DPV, PTV, and LNG each reached concentrations below the limit of quantification (BLOQ). Drug concentrations in both proximal (Fig. 9D–F) and distal (Fig. 9G–H) vaginal fluids likewise remained relatively constant across the 28-day study. Median (range) API concentrations proximal to the cervix for the group (n = 3) while the rings were present from day 3-day 28 were 375,520.7 (6521–1,103,053.2) pg/mL for DPV, 632,327.9 (12,152.7–2,918,559) pg/mL for PTV, and 6642.2 (500–60,648.1) pg/mL for LNG. Median (range) API concentrations distal to the cervix for the group (n = 3) while the rings were present from day 3-day 28 were 76,174.6 (38,081.6–330,178.8) pg/mL for DPV, 171,289.3 (22,365.4–800,431.4) pg/mL for PTV, and 2469.9 (500–15,668.1) pg/mL for LNG. The concentration of DPV, PTV, and LNG was about 5, 4, 3-fold higher respectively in the proximal vaginal fluid compared to distal vaginal fluid. DPV had the highest vaginal tissue concentration with a median (range) for the group (n = 3) while the ring was present at 808,163 (151,170–1,630,721) pg/g, and all three APIs had steady concentrations from 7 days to 21 days, with the vaginal tissue concentrations falling below the limit of quantification post-IVR removal. Data presented as median with 95 % confidence interval of n = 3 macaques/timepoint per dose. Individual replicates shown in Supplemental Fig. 5.

After completion of the in vivo study, the MPT IVRs were extracted to quantify residual drugs remaining in the IVR. IVRs (n = 3) were placed in acetone to extract residual drug, and extraction samples were quantified using HPLC (Table 3). The average residual DPV was of 8.82 ± 0.11 mg (94.3 %), corresponding to an average of 0.53 ± 0.11 mg DPV released over 28 days (~19 μg/day DPV). Residual PTV was quantified at 2.15 ± 0.49 mg (36.7 %) corresponding to an average of 3.71 ± 0.49 mg released over 28 days of use (~133 μg/day PTV). For LNG, the quantified residual drug was 0.08 ± 0.02 mg (15.4 %) corresponding to an average of 0.44 ± 0.02 mg LNG released over 28 days (~16 μg/day).

A comparison between the in vitro and in vivo release rates was done for all three APIs (Table 4). Results showed that the in vitro and in vivo release rates for PTV and LNG were closely comparable, with correlative values of 1.1 and 1.2 respectively. LNG exceeded the target in vivo macaque daily release rate of 5.4 μg/day and human in vitro daily release rate of 20 μg/day. Similarly, PTV elicited release rates above the set target of 200 μg/day in vitro and 54 μg/day release in vivo in macaques. Unexpectedly, DPV elicited daily release rate, based on residual drug, of about 10× lower in vivo (~19 μg/day) compared to the in vitro release rate (~189 μg/day) resulting in an in vitro-in vivo ratio of 0.1.

In vitro-in vivo comparisons are difficult to do for IVRs due the complexities of the formulations, lack of standardized in vitro release testing methods, and inter-subject variability in pharmacokinetic data [63,65]. The comparison reported here is a level B correlation that compares the cumulative mean API released over 28 days in the translational in vitro release study and the in vivo study in macaques. The in vitro release conditions that most closely matched the in vivo data were the macaque size IVRs tested under pH 8 dynamic conditions. It was found that PTV and LNG had similar cumulative release when tested under the translational in vitro conditions and in vivo. While the in vitro results cannot explain the drug distribution that is observed in vivo, the translational in vitro release conditions can provide an estimate of how much PTV and LNG will be released from the MPT IVR.

The measured plasma progesterone correlated with the LNG detected following insertion of the IVR with an average six-fold decrease in the progesterone concentrations on day 3 (Supplemental Fig. 6). A 47-, 45-, and 39-fold decrease in the average progesterone concentrations on days 7,14, and 21 respectively in comparison to day 0 was detected. Progesterone concentrations on days 28 and 30 were similar to that measured on day 0. This corresponded with the LNG being below the detection limit at those time points. Moreover, the plasma measurement at day 30 was obtained 2 days following ring removal. It could be likely that the low levels detected on days 0–21 in macaque B corresponds to it having a different cycling pattern in comparison to the other two macaques. We did not collect samples in these macaques at time points prior to day 0. Thus, we have provided a representative progesterone pattern in a normally cycling pigtailed macaque with no contraceptive modality (not a part of this study) to exhibit the effects of LNG released in the 3D IVR macaques (Supplemental Fig. 6).

The in vivo safety of the application and the retention of the MPT IVRs were evaluated by colposcopic observations, changes in vaginal pH, mucosal cytokine and chemokine concentrations and systemic T cell subset population frequencies. Routine visual examination during regular mucosal sample collection and colposcopic observations on days 0 and 14 revealed a normal finding with no tissue abrasion or inflammation. The vaginal pH measurements revealed no abnormal variations at any of the time points tested following ring insertion and removal (Fig. 10). The pH range observed during the study is similar to that normally detected in pigtailed macaques [48,66,67].

Mixed-effects model analysis with Tukey test correction for multiple comparisons revealed no statistically significant changes in any of the 22 cytokines and chemokines tested including the pro- (IL-1β, IL-6, IL-8, IFN-γ, TNF-α, IL-17, IL-18) and anti-inflammatory (IL-10, IL-12) cytokines while the 3D IVR was in place or upon its removal (Fig. 11). To determine the influence, if any, of the three drugs, and specifically the contraceptive LNG on the immune cell population, we measured the frequency of systemic CD4+ and CD8+ T cells during the entire study. We observed no variation in cell frequencies in all samples collected following ring insertion and post removal on day 30 (Supplemental Fig. 7).