Structure-activity relationships of natural and semi-synthetic plecomacrolides suggest distinct pathways for HIV-1 immune evasion and vacuolar ATPase-dependent lysosomal acidification

Abstract

The HIV-encoded accessory protein Nef enhances pathogenicity by reducing MHC-I cell surface expression, protecting HIV infected cells from immune recognition. Nef-dependent downmodulation of MHC-I can be reversed by sub-nanomolar concentrations of concanamycin A (CMA, 1), a well-known inhibitor of vacuolar ATPase (V-ATPase), at concentrations below those that interfere with lysosomal acidification or degradation. We conducted a structure-activity relationship study that assessed 75 compounds for Nef inhibition, 24- and 72-h viability, and lysosomal neutralization in Nef-expressing primary T cells. This analysis demonstrated that the most potent compounds were natural concanamycins and their derivatives. Comparison against a set of new, semi-synthetic concanamycins revealed that substituents at C-8 and acylation of C-9 significantly affected Nef potency, target cell viability and lysosomal neutralization. These findings provide important progress toward understanding the mechanism of action of these compounds and identification of an advanced lead anti-HIV Nef inhibitory compound.

Graphical Abstract

Introduction

The development of combination antiretroviral therapy (cART) has improved the outcome for individuals with human immunodeficiency virus (HIV). However, HIV can persist even during cART by establishing a latent reservoir in a subset of infected cells¹. Eliminating these infected cells will require a combination of therapeutic agents with different mechanisms of action². Recently, researchers have proposed a two-pronged “shock and kill” strategy for virus eradication³. The “shock” step involves treatment with latency reversal agents (LRAs) to reactivate latent reservoirs of HIV infected cells. Reactivated infected cells could then be killed through the activity of anti-HIV cytotoxic T-cell lymphocytes (CTL) or via the cytopathic effect of the virus. However, CTL recognition and destruction of infected cells depends on expression of host major histocompatibility complex I encoded (MHC-I) proteins on the cell surface. The HIV-Nef protein promotes evasion from CTL recognition by downmodulating MHC-I HLA-A and -B⁴. Nef disrupts MHC-I cell surface expression by binding to the MHC-I cytoplasmic tail and stabilizing an interaction between MHC-I and clathrin adaptor protein 1 (AP-1)⁵. Formation of this complex results in the targeting of MHC-I to the lysosome for degradation⁵. A small molecule that inhibits Nef-mediated disruption of MHC-I antigen presentation could promote the “kill” step of the shock and kill strategy by restoring MHC-I surface expression and facilitating CTL-mediated elimination of cellular reservoirs⁶.

Designing an effective HIV-1 Nef inhibitor to achieve a potential cure has challenged researchers for decades. Nef is a multi-functional protein that disrupts signal transduction pathways and intracellular protein trafficking. In addition, Nef influences host cell proteins through distinct surface interactions rather than enzymatic or biochemical activity⁷. Despite these challenges, four classes of Nef inhibitors have been described^6,8,9. The first class is composed of 4-amino-diphenylfuranopyrimidine (DFP) analogs, which block Nef dependent activation of the Src-family kinase, Hck¹⁰. These DFP inhibitors potentially bind Hck in its active site, countering Nef induced Hck activity and selectively limiting Nef-dependent enhancement of HIV-1 replication¹¹. The second class of inhibitors are characterized by a hydroxypyrazole scaffold that binds Nef directly, and blocks Nef dependent Hck activity¹². Members of this class of inhibitors, including B9, were reported to increase MHC-I surface presentation 1.6–1.8-fold in HIV-1 infected CD4⁺ T cells at a concentration of 0.5–5 μM¹³. Lovastatin alone composes the third class of inhibitors, which is reported to bind Nef directly, disrupting its ability to bind AP-1⁸. Lovastatin treatment at supra-therapeutic single digit micromolar concentrations increases surface expression of SERINC5 in Nef overexpressing HEK293T cells and surface expression of MHC-1 and CD4 in HIV-1 infected CD4⁺ T cells⁸. The fourth class of inhibitors includes plecomacrolides, which selectively block Nef-dependent disruption of MHC-I antigen presentation, restoring CTL killing of HIV-infected T cells⁶. In our hands, nanomolar concentrations of concanamycin A (CMA, 1) are sufficient to near-fully restore MHC-I surface expression in HIV-1 infected CD4⁺ T cells compared to a control lacking Nef. In comparison, hydroxypyrazole B9, which was structurally confirmed by NMR, did not alter MHC-I HLA-A2 levels in HIV-infected primary T cells across a range of concentrations up to more than 10μM. Lovastatin partially restored MHC-I at 2,000-fold higher concentrations than CMA⁶. Based on this study that focused on MHC-I HLA-A2 restoration activity of Nef inhibitors in side-by-side assays, CMA is by far the most potent anti-Nef inhibitor identified so far.

CMA and other plecomacrolides are best known for their capacity to disrupt vacuolar-type H⁺-ATPase (V-ATPase) protein translocation¹⁴ which is needed for lysosomal acidification and protein degradation. Based on competitive inhibition experiments¹⁵, CMA is hypothesized to disrupt V-ATPase through a non-covalent interaction with the c subunit of its V₀ domain. Remarkably, however, CMA fully inhibits Nef-dependent MHC-I downmodulation activity at concentrations substantially below those needed to disrupt lysosomal acidification. This was best-shown by an experiment in which low dose CMA selectively blocked MHC-I downmodulation, but not lysosomal degradation of the HIV receptor, CD4⁶. Thus, CMA may bind a novel target or form of V-ATPase to selectively inhibit Nef-dependent MHC-I downmodulation. Previously, we reported that CMA inhibits Nef by disrupting the ability of Nef to interact with cellular factors (AP-1) needed to target MHC-I for lysosomal degradation⁶. The effects of CMA on Nef-dependent MHC-I downmodulation appear to be indirect as no binding between Nef and CMA was detected⁶.

While CMA inhibits Nef activity at sub-nanomolar (nM) concentrations⁶, its utility as a potential anti-HIV therapeutic requires identifying a derivative with lower toxicity. Because CMA has a much lower IC₅₀ for Nef inhibition than lysosomal neutralization⁶ (mean Nef IC₅₀ [0.18 nM] and lysosomal neutralization IC₅₀ is 1.7 mM of CMA [Figure 1]), we hypothesized the existence of separable CMA-dependent inhibitory targets. Therefore, our objective was to determine the extent to which Nef inhibitory activity (Figure 1B), lysosomal neutralization (Figure 1C) and toxicity (Figure 1A) were selectively enhanced, reduced, or even eliminated in relevant natural products and semi-synthetic derivatives. This was examined by comparing compounds based on their “therapeutic ratio” (ratio of TC₅₀ for 72 h viability / IC₅₀ for Nef inhibition at 24 h), and their “neutralization ratio” (ratio of IC₅₀ for lysosomal neutralization (Lysotracker) / IC₅₀ for Nef inhibition at 24 h) (Figure 1D, E).

Figure 1. SAR assay workflow.

(A) Measurement of cell viability and determination of TC₅₀. (B) Assessment of Nef functional inhibition and determination of Nef IC₅₀. (C) Determination of Lysotracker/V-ATPase inhibition IC₅₀. The equations used to determine therapeutic and neutralization ratios are shown as indicated in (D) and (E). See Supplemental Figure 1 for details of flow cytometric assay and gating strategy.

For the current structure activity relationship (SAR) analysis, we screened a panel of macrolactones, macrolides, and plecomacrolides against primary T cells transduced with a replication-defective HIV reporter virus expressing Nef to assess which structural variations led to differences in relative Nef and lysosomal inhibitory activity (neutralization ratio, Figure 1E). Our results identified key modifications to 18-membered ring plecomacrolides that retain single-digit nM potency for Nef inhibitory activity and provide substantive reduction in lysosomal neutralization (e. g. V-ATPase inhibition), pointing the way toward final lead optimization. Longer-term, we aim to reach a neutralization ratio of 40–50, and a therapeutic ratio of 15–20 in an anti-HIV Nef therapeutic candidate. Because in vitro results may not translate to in vivo conditions, future studies in a primate model(s) will be crucial for guiding target therapeutic ratios and a threshold for toxicity with CMA/CMB semi-synthetic analogs prior to human trials. The current effort demonstrates our design of bafilomycin and concanamycin semi-synthetic analogs, and a detailed report toward achieving our longer-term objectives.

Results

Screening for Nef activity and lysosomal neutralization reveals the importance of macrolactone ring size.

To investigate structure activity relationships (SAR) for Nef inhibitory activity across a variety of related macrocyclic compounds, we tested a panel of 79 molecules with varying ring sizes and additional select structural modifications (see SI Tables, S1–S4, for all compounds). We also tested two known V-ATPase inhibitors, archazolid A (3, Figure 2C, E) and diphyllin (4, Figure 3A), that are structurally unrelated to plecomacrolides¹⁶.

Figure 2. Concanamycin has significantly greater differences between Nef inhibitory activity and lysosomal neutralization than other plecomacrolides.

(A–C) Chemical structures of concanamycin A, bafilomycin A₁ and archazolid A, respectively. (D) Summary graph of Nef activity (24 h), cell viability (24 and 72 h), and lysosomal neutralization (Lysotracker, 24 h) in primary cells transduced with HIV DGPE⁶ and treated with the indicated concentrations of CMA. (E) Summary table of key values. Neutralization (Lysotracker IC₅₀/Nef IC₅₀) and therapeutic (72h Viability/Nef IC₅₀) ratios were calculated within experiments; Mean ± standard deviation is shown; n, number of donors; * p<0.05, ** p<0.01, *** p<0.001 indicates significance of difference compared to 1 by 1 way ANOVA with Brown-Forsythe and Welch’s correction for unequal standard deviations and Dunnett’s multiple comparisons test. Significant differences in neutralization and therapeutic ratios are highlighted red for smaller ratios relative to 1.

Figure 3. Various natural products that were tested for inhibition were found to have little to no activity.

A. Structures of macrocycles. B. Table of Nef inhibitory activity, 72-h viability, and lysosomal neutralization (Lysotracker) in HIV infected primary cells after treatment with non-plecomacrolide substrates 4–7. Nef activity and lysosomal neutralization were measured at 24 h. >1000 indicates a sigmoidal curve was not generated at the highest concentration tested. n=1 donor for 6, n=3 donors for all other compounds.

Inhibition of HIV-Nef, V-ATPase and cell viability were assessed flow cytometrically as previously described⁶ (Figure 1 and Supplemental Figure 1). We utilized primary CD4⁺ T cells transduced with a replication-defective HIV containing a GFP reporter (ΔGPE) and assessed: (1) inhibition of Nef-dependent MHC-I HLA-A2 surface downmodulation via antibody staining of HLA-A2, (2) inhibition of V-ATPase-dependent lysosomal acidification by assessing fluorescence of a pH sensitive dye (Lysotracker) that accumulates in the lysosome and, (3) cell viability using a fluorescent vital dye at 24 h and 72 h after addition of drugs. In this SAR study, the upper limit of the dose-response assay was 1000 nM, therefore compounds in which Nef IC₅₀s were not below 1000 nM were classified as inactive.

All compounds tested with relatively small macrocyclic rings (12- and 14-membered macrolactones) failed to demonstrate potent Nef or lysosomal neutralization activity at concentrations below 1000 nM, except for 10-DML (Figure 3, Tables S1 & S2). In addition, none of the non-macrolide compounds tested, including the V-ATPase inhibitor diphyllin (4), had a Nef IC₅₀ below 1000 nM (Figure 3, Tables S1 & S2). A small number of compounds, such as borrelidin (S1), were toxic at 24-hrs making any determination of Nef inhibitory activity unreliable (Table S1). Of the two 24-membered macrolactones that were tested, only archazolid A (3), a V-ATPase inhibitor that binds to the same subunit as the plecomacrolide bafilomycin A₁ (Baf A₁, 2),¹⁷ had a Nef IC₅₀ less than 1000 nM (Figure 2E). Indeed, 3 was more potent than 2, achieving the sub nM potency observed with CMA (1) (Figure 2E). In comparison to 1, archazolid A (3) had a smaller therapeutic ratio (6.6 vs 1.6-fold, respectively) as well as a smaller neutralization ratio between Nef and lysosomal neutralization IC₅₀s (9.4 vs 3.6-fold, respectively) (Figure 2E). Nevertheless, this 24-membered macrocycle (3) was the only structurally distinct compound from the plecomacrolides tested that displayed sub-nM potency.

A detailed comparison between CMA (1) and Baf A₁ (2) confirmed our prior finding that 1 shows a more potent Nef inhibitory activity relative to 2 (approximately 67-fold, Figure 2D–E)⁶. The concentration that the 18-membered plecomacrolide CMA inhibited Nef by 50% (IC₅₀) was about 100x lower than Baf A₁ [(2), Figure 2B,E]⁶. We extended our prior analysis to a comparative assessment of lysosomal neutralization between these two molecules and found that CMA (1) had a three-fold greater neutralization ratio than Baf A₁ (2) (Figure 2D–E). A greater magnitude was also observed in the therapeutic ratio of CMA (1) versus Baf A₁ (2) (6.6 vs 1.4-fold, respectively; Figure 2E). While the explanation for these variant inhibitory activities within the plecomacrolides tested are unclear, three major structural differences exist between CMA and bafilomycins A–D: (1) macrolactone ring size (18-membered macrocycles (CMA)) displayed reduced effects on the lysosome and lower toxicity than 16-membered macrocycles); (2) pyran (hemiketal) moiety (bafilomycin D (Baf D) lacks this moiety and is the least potent Nef inhibitor⁶); and (3) β-D-rhamnose sugar (CMA differed from more toxic/less active bafilomycins by the presence of this moiety). In sum, these results support the hypothesis that chemically related molecules can vary in the extent to which they inhibit Nef versus V-ATPase-dependent lysosomal acidification and that chemical modifications may reduce toxicity by limiting off-target effects on lysosomal acidification.

Semi-synthetic modification of 16-membered plecomacrolides identifies critical moieties required for Nef inhibition.

The role of the pyran ring at C-19 in Baf A₁

To better understand which structural variations within the plecomacrolide family determine cell toxicity, Nef inhibition and lysosomal neutralization, functional group modifications were tested first with Baf A₁ (2), a more readily accessible plecomacrolide¹⁸, leading to a total of 17 semi-synthetic derivatives. Bacterial fermentation of a Streptomyces lohii strain was conducted to generate sufficient quantities of Baf A₁ for derivatization (detailed procedure can be found in Methods)⁶. Because Baf D, which lacks a pyran moiety, is a poor inhibitor of both Nef⁶, and V-ATPase^19–21, we started by investigating the impact of pyran ring modifications. We found that C-19 methylation (8), deoxygenation (9), and/or C-21 methylation (10) and C-20 dehydration (11) all decreased potency (increased Nef IC₅₀) (Figure 4)²². To directly confirm the requirement of the hemiketal functionality in a semi-synthetic derivative, we used Mitsunobu conditions²³ to yield “ring-opened” Baf A₁ (12; Figure 4A). Consistent with results obtained with Baf D, this alteration led to a 26-fold decrease in potency compared to 2 (Nef IC₅₀ = 310 nM; p≤ 0.01; Figure 4B). Dramatic reductions in V-ATPase inhibition and toxicity were also observed. These results clearly demonstrate the critical requirement for the pyran moiety for potent plecomacrolide activity.

Figure 4. Functional group modifications in 16-membered ring plecomacrolides that impact Nef inhibition.

(A) Structures of bafilomycins 2, 8–10, & 12–21. (B) Summary table of Nef inhibitory activity, neutralization (Lysotracker IC₅₀/Nef IC₅₀) and therapeutic (72-h Viability/Nef IC₅₀) ratio in HIV-GPE- transduced primary cells after treatment with 16-membered macrocyclic natural products and semi-synthetic derivatives. Nef activity and lysosomal neutralization were measured at 24 h. Mean ± standard deviation is shown; n, number of donors; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001 indicates statistically significance of difference compared to 2. Neutralization ratios (Lysotracker IC₅₀/Nef IC₅₀) and therapeutic ratios (72h Viability/Nef IC₅₀) were calculated within experiments. Significant differences in neutralization ratio and therapeutic ratio are highlighted in green for larger, and in red for smaller ratios relative to 2.

Enhancement of Nef inhibition and lysosomal neutralization through modifications at C-21 in 16-membered ring plecomacrolides

Based on our observation that the pyran ring was amenable to synthetic derivatization^14,24,25, efforts focused on modifying C-21-OH by employing Steglich esterification conditions²⁴. To determine if chain extensions at this position [increasing the overall size of Baf A₁ (2)], would mimic the larger concanamycin analogs and consequently improve potency, we modified C-21-OH with an acetyl group²⁴ (13).This change led to a small but significant decrease in the Nef IC₅₀ (4.5 nM vs 12 nM) (Figure 4). Introduction of esters with longer acyl chains or an unsaturated chain at C-21-OH (14, 15 & 16) resulted in decreased potency and worsening of the neutralization ratio (Figure 4). Lastly, incorporation of a bulky aromatic substituent (17) led to larger decreases in potencies (Nef IC₅₀ = 92 nM and Lysotracker IC₅₀=128 nM). The correlation between size and disruptive impact suggested the possibility that some substitutions may reduce activity through steric hindrance.

In addition to the larger macrocyclic ring, CMA (1) contains a β-D-rhamnose at C-23, instead of a hydroxyl at the analogous C-21 position in Baf A₁ (2) (Figure 2A, B). To determine the extent to which this sugar moiety is required for Nef and lysosomal neutralization, we examined the activity of two 16-membered ring plecomacrolides that contain modified β-D-rhamnose moieties at the analogous C-21 position: leucanicidin (18) and PC-766B (19) (Figure 4A). Interestingly, 18 and 19 were the first 16-membered ring analogs we tested to have single digit nM potency (Nef IC₅₀ = 1.2 nM for both). However, they were also significantly more toxic (TC₅₀ = 1.4 nM and 1.3 nM, Figure 4B), and thus the therapeutic ratio was not significantly different from Baf A₁ (1.3 and 1.1, respectively).

The Baf A₁ C-7 hydroxyl group is essential for both Nef inhibition and lysosomal neutralization

A recently reported cryo-EM structure of Baf A₁ (2) bound to the V₀ domain of V-ATPase highlights the importance of the hydrogen bond interaction between the C-7-OH on the macrocyclic ring and the corresponding tyrosine residue at position 144 of the c subunit of the V₀ domain²⁰. To investigate whether this interaction also plays a role in Nef functional inhibition and to confirm its requirement for V-ATPase inhibition, we utilized a Corey-Suggs oxidation²⁵ to generate a new Baf A₁ derivative bearing a keto group at C-7 as well as the corresponding C-7 and C-21 diketo derivative (20 & 21, respectively; Figure 4). Consistent with a role for a critical hydrogen bond interaction at this position, oxidation at C-7 was sufficient to entirely abolish both Nef and V-ATPase inhibition with IC₅₀s above the activity threshold of our study (Figure 4).

Critical moieties in 18-membered ring plecomacrolides required for Nef inhibition.

The role of the Concanamycin macrolactone C-8 position

In addition to CMA (1), there are other naturally occurring concanamycin analogs that can be obtained by fermentation²⁶, including concanamycin B (CMB, 22) and concanamycin C (CMC, 23). CMB (22) differs from CMA (1) in that CMB (22) has a methyl instead of an ethyl group at position C-8. CMC (23) differs from CMA (1) in that it lacks the 4’-carbamoyl group on the β-D-rhamnose (Figure 5A). The ethyl versus methyl difference between 1 and 22 was associated with an almost 10-fold decrease in Nef inhibitory potency (Nef IC₅₀ = 1.2 nM vs. 0.18 nM; Figure 5B). Remarkably, however, CMB (22) exhibited an even greater decrease in V-ATPase inhibitory potency as assessed by Lysotracker IC₅₀ (Figure S4), resulting in a significantly improved neutralization ratio compared to CMA (1) (24-fold vs. 9.4-fold; Figure 5B). Somewhat disappointingly, the therapeutic ratio for CMB (22) was not correspondingly improved compared with CMA (1). Nevertheless these results are encouraging because they suggest that further modifications in the macrocyclic scaffold of 22 via semi-synthesis could yield a less toxic compound. By contrast, CMC (23) displayed both a decrease in potency compared to CMA (1) (Nef IC₅₀ = 0.71 nM) and a decrease in the therapeutic ratio, although the neutralization ratio was not significantly affected (Figure 5B).

Figure 5. Functional group modifications in 18-membered ring plecomacrolides that affect Nef inhibition.

(A) Structures of concanamycins 1 & 22–39. (B) Summary table of Nef inhibitory activity, neutralization ratio (Lysotracker IC₅₀/Nef IC₅₀) and therapeutic ratio (72-h Viability/Nef IC₅₀) in HIV-GPE- transduced primary cells after treatment with compounds in (A). Nef activity and lysosomal neutralization were measured at 24 h. Mean ± standard deviation is shown; n, number of donors; * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001 indicates significance of difference compared to 1 or parent compound. Compounds 22–23, 26, 32, and 33 were compared to 1. Compounds 25 and 27–31 were compared to 22. Neutralization ratios (Lysotracker IC₅₀/Nef IC₅₀) and therapeutic ratios (72h Viability/Nef IC₅₀) were calculated within experiments. Significant differences in neutralization ratio and therapeutic ratio are highlighted in green for larger, and red for smaller ratios relative to 1 or parent compound.

Role of the 4’-carbamoyl-β-D-rhamnose at C-23 of CMA

To further explore the role of the 4’-carbamoyl-β-D-rhamnose in CMA (1), we synthesized and tested its aglycone derivative, concanamycin F (CMF, 24; Figure 5A) that was accessible through acid-catalyzed hydrolysis²⁷. As expected, lack of the 4’-carbamoyl-β-D-rhamnose led to a decrease in potency compared to CMA (1) (Nef IC₅₀ = 1.6 nM vs 0.18 nM, Figure 5B). Further highlighting the importance that the 4’-carbamoyl-β-D-rhamnose plays in Nef inhibition, the synthesized aglycone of CMB (22), CMX (25), also displayed a reduction in potency compared to its parent (Nef IC₅₀ = 9.1 nM vs 1.2 nM; Figure 5). Both 24 and 25 also showed significant decreases in neutralization and therapeutic ratios compared to their parent compounds (Figure 5B).

Enhancement of Nef inhibition and lysosomal neutralization by modifications at C-3’ in 18-membered ring plecomacrolides

Applying what was learned from testing semi-synthetic derivatives of the 16-membered ring plecomacrolides to the 18-membered ring system, a series of acyl chains, incorporated by Steglich esterifications²⁴, were selected to derivatize CMA (1) and CMB (22) to assess changes in inhibitory activities. Addition of longer, flexible chains to C-3’, such as a pentanoate group to each parent (26 and 27), resulted in a significant increase in the Nef IC₅₀ (0.18 to 0.75 nM and 1.2 to 3.4 nM, respectively). A similar increase in the neutralization ratio of 26 was also observed (9.4-fold to 15-fold). Addition of a second pentanoate group to 27 at C-9-OH (28) and extension of the chain in the form of a 3’-nonanoate CMB derivative (29) resulted in small but statistically significant increases in Nef IC₅₀ (6.3 and 9.1 nM respectively; Figure 5). Importantly, double acetylation at C-3’-OH and C-9-OH (31) maintained Nef potency (Nef IC₅₀= 0.99nM) while widening the therapeutic ratios (8.8-fold vs 5.5-fold), without significantly changing the neutralization ratio (Figure 5). Because 31 had the largest therapeutic ratio, we synthesized corresponding single and double acetylations of CMA (1), CMC (23), and CMF (24). We found that acetylation of C-3’-OH (32) or of both C-3’-OH and C-9-OH positions (33) improved the parent compound CMA (1) by increasing the neutralization ratios (19 and 21-fold respectively vs 9.4-fold, p<0.01, Figure 5B) while maintaining a sub-single digit nanomolar Nef IC₅₀ (0.20 and 0.39 nM, respectively). Similar modifications of CMC (23), and CMF (24) (yielding 34/35 and 36/37, respectively) had no significant impact on neutralization or therapeutic ratios (Figure 5).

Restoration of MHC-I expression by CMA or its derivatives was further evaluated by comparing the degree of MHC-I restoration to control cells lacking Nef. As shown in Figure 6, and as expected, MHC-I levels were reduced on the surface of infected (GFP+) primary T cells transduced with an HIV GFP reporter construct (ΔGPE) as compared to uninfected (GFP−) cells in the same sample at all concentrations of solvent tested (Figure 6D). Treatment with CMA (1) or 3’,9-diacetyl CMA (33) restored MHC-I levels similar to those of uninfected (GFP−) cells in a dose-dependent manner (Figure 6E, F). Also as expected, MHC-I levels on GFP+ cells harboring a mutant reporter HIV lacking Nef (ΔGPEN) were essentially normal as compared to uninfected (GFP−) cells in the same culture at all concentrations of solvent tested (Figure 6A). Moreover, CMA (1) and 3’,9-diacetyl CMA (33) did not substantially alter MHC-I surface levels in cells lacking Nef (Figure 6B, C). These data provide strong evidence that these compounds selectively target Nef-dependent MHC-I downmodulation, restoring MHC-I HLA-A2 to near-normal levels previously shown to be sufficient for anti-HIV CTL recognition⁶ and do not substantially increase MHC-I expression on cells lacking Nef.

Figure 6. CMA and 3’,9 diacetyl CMA reverse Nef-dependent HLA-A2 downmodulation.

Flow cytometric analysis of MHC-I HLA-A2 on infected primary CD4⁺ T cells transduced with ΔGPE or an HIV construct that does not express Nef (ΔGPEN) for 48 h then treated with the indicated concentration of CMA (1) or 3’,9 diacetyl CMA (33) for 24 h.

Requirement for the 16-methoxy in 18 membered ring plecomacrolides.

Noting that relatively small structural variations in the macrolactone core altered potency, we tested analogs with modifications on the right side of the molecule. Previously reported studies suggest that the concanamycin 16-OMe position is crucial for maintaining the active macrolactone conformation and retaining potent V-ATPase inhibitory properties²¹. This hypothesis was examined by converting the C-16-OMe to the C-16-OH derivative from the previously synthesized 21-deoxy CMF (38) (deoxygenation at the C-21 showed no change in potency (Figure 5)), to form 21-deoxy-16-hydroxy CMF (39; Figure 5). As expected, based on prior studies, the change at C-16 led to a dramatic loss of Nef inhibitory activity (over 200-fold reduction) compared to CMF (24) (Figure 5).

Discussion

The identification of a potent Nef inhibitor that enhances CTL-mediated clearance of active and latent reservoirs of HIV is an important goal toward developing a cure. Through the curation and testing of a SAR panel of 75 compounds, including a series of novel semi-synthetic plecomacrolides, we determined that the most potent Nef inhibitors identified thus far are microbially derived 16-, 18-, or 24-membered macrocycles (Figure 7). Prior to our discovery of the plecomacrolides as potent Nef inhibitors, CMA (1), Baf A₁ (2), and archazolid (3) were primarily known for their highly specific and potent V-ATPase inhibitory activity in eukaryotic cells^6,19,23. Recently, cryo-EM structures revealed that Baf A₁ (2) and archazolid (3) have distinct binding sites on the c subunit of the V₀ domain of V-ATPase. Despite the difference in binding sites, the C-7-OH of both 2 and 3 forms a critical hydrogen bond with a tyrosine residue that mediates binding to the c-subunit¹⁷. In our study, a semi-synthetic derivative of Baf A₁ (2) that has been oxidized at C-7-OH to a keto group (20) had reduced activity at all concentrations tested, highlighting the role of the hydrogen bond donating abilities of the C-7-OH to achieve Nef and V-ATPase inhibition. Currently, there is no structure available detailing how CMA (1) induces V-ATPase inhibition, only a hypothesis, supported by competitive inhibitor studies¹⁵, that it binds the c-subunit similarly to its 16-membered ring family member, Baf A₁ (2)^19,28.

Figure 7. Concanamycins A–C were found to be the most potent inhibitors of Nef activity.

A heat map of the Nef IC₅₀ of the natural products tested. A log scale was used. Additional data, including statistical significance, number of replicates, and exact Nef IC₅₀s can be found in the Tables S1–S4.

In addition, our SAR study highlights how CMA (1), and its semi-synthetic derivatives, act differently than Baf A₁ (2), and archazolid (3) by inhibiting Nef and thereby restoring anti-HIV immune recognition at low concentrations where there is minimal effect on lysosomal function⁶. This ability to separate Nef functional inhibition from lysosomal neutralization supports the hypothesis that distinct pathways exist for downmodulation of MHC-I and V-ATPase dependent lysosomal acidification, and that these two pathways might be separately targeted by semi-synthetic CMA/CMB derivatives.

One of the initial factors investigated was the impact of macrolactone ring size on Nef function inhibition. Comparing 16- versus 18-membered-ring plecomacrolides [Baf A₁ (2), vs. CMA (1)] revealed that the latter had a 100-fold greater Nef inhibitory activity. In addition, the 24-membered macrolactone archazolid (3) had potent Nef inhibitory activity, however others such as rifampicin (5, Figures 3 & 7), did not, indicating that sub-nM Nef inhibitory activity is not a general feature of large macrocyclic compounds. Smaller rings with 10-, 12-, and 14-membered macrolactones failed to display inhibitory activity at all concentrations tested (Figure 7). For Baf A₁ (2), CMA (1), and archazolid (3) the trend observed for Nef IC₅₀ corresponded to that observed for lysosomal neutralization (1 < 3 < 2,). Despite the similar trends, chemical differences unique to each compound differentially affected these two readouts, resulting in differences in the neutralization ratios (9.4, 3.6, & 2.9-fold for 1, 3, and 2, respectively). This conclusion supports the hypothesis that distinct targets with differential sensitivity to these compounds are inhibited to disrupt Nef-dependent MHC-I downmodulation versus lysosomal acidification.

Differences in the level of potency for lysosomal neutralization that we observed for Baf A₁ (2), CMA (1), and archazolid (3) may relate to differential binding to V-ATPase based on ring size and other structural features. Indeed, previous studies have noted that the size of the macrocyclic core impacts binding orientation to the c-subunit of V-ATPase^19,20. Identification of the direct target of binding and further structural studies are required to explain the relative potencies of 1–3 for Nef inhibition.

In addition to differences in ring size, CMA (1) contains a 4’-carbamoyl-β-D-rhamnose moiety at the pyranyl C-23 position that is absent from the analogous C-21 position of Baf A₁ (2). Two microbially derived 16-membered ring glycosylated plecomacrolides, leucanicidin (18) and PC-766B (19), were tested that contained modified β-D-rhamnose moieties at the analogous C-21 position. We found that when C-21-OH is glycosylated the 16-membered scaffold becomes a significantly more potent Nef inhibitor with single-digit nM IC₅₀s. However, neutralization and therapeutic ratios were relatively poor. The necessity of a sugar moiety for high potency was also observed with 18-membered ring plecomacrolides when comparing the Nef IC₅₀s of CMF (24) and CMX (25) (1 vs 24 and 22 vs 25, respectively). In both instances, addition of the sugar led to improvements in potency. It is also important to note that both aglycones displayed significantly smaller neutralization ratios compared to their parent compounds, especially for 25. Moreover, the presence versus absence of 4’-carbamoyl on the β-D-rhamnose moiety [CMA (1) versus CMC (23)] increased potency and reduced toxicity. These results confirm, using the 18-membered ring plecomacrolide scaffold, that the β-D-rhamnose and 4’-carbamoyl moieties are integral motifs for nM inhibition of HIV-Nef and neutralization of the lysosome (Figures 8A).

Figure 8: Summary of SAR panel of plecomacrolides.

(A) Structural components of 1 that impact activity. Specific values can be found in Figure 3 & Table S4. (B) Structure of promising semi-synthetic compounds 31 and 33.

For both CMA (1) and Baf A₁ (2), we found that the pyran ring is vital for Nef inhibition and lysosomal neutralization, as conversion to a linear system (12) drastically reduced both inhibitory activities for 2. More limited changes, including modification at C-19 of 2 reduced Nef inhibitory potency while other changes, such as esterification of C-21-OH of 2 yielded results that depended on chain length [small acyl chains (13) improved potency for both inhibitory activities whereas the addition of longer acyl chains (14, 15) had the opposite effect]. We similarly observed both positive and negative effects on potency for changes at C-21 of 19-deoxy Baf A₁ (9); conversion of C-21-OH to an allyl-ester (S2, Table S3) decreased potency whereas addition of a longer chain with a terminal alkyne improved potency (S3, Table S3). This trend was reinforced by the C-21-penta-3-enoate (S4, Table S3) and C-21-penta-2,4-dienoate (16) derivatives of 2 where addition of acyl chains bearing variant degrees of unsaturation improved potency compared to 14 or 15. Based on these results, we hypothesize that hydrophobic interactions mediated by these linear substituents may be important for target binding especially because they parallel the effects of hydrophobic substituents previously reported to play a role in V-ATPase binding²⁰.

A key result of our study is that CMB (22), which bears a C-8 methyl compared to C-8 ethyl in CMA (1), represents a second lead compound for pre-clinical development. Although CMB (22) has a 10-fold lower potency for Nef inhibition compared to CMA (1), it has an even lower potency for V-ATPase inhibition as assessed by Lysotracker fluorescence, resulting in an expanded neutralization ratio. While we were initially disappointed that CMB (22) did not have a correspondingly improved therapeutic ratio, further semi-synthetic modification of 22 to 3’,9-diacetyl CMB (31) succeeded in reducing toxicity compared to the parent compound CMB (22). These results indicate that continued modification along these lines is a promising avenue for drug development.

We also made progress towards improving the characteristics of these compounds by esterifying (individually or in tandem) the C-3’-OH and C-9-OH positions, which motivated synthesis of acetyl and diacetyl derivatives 32/33, 34/35, and 36/37 (derived from CMA (1), CMC (23), and CMF (24), respectively). The most notable compound was 3’,9-diacetyl CMA (33, Figure 8B), as this molecule displayed neutralization ratios that had only been observed with CMB (22) and its derivatives, while maintaining both potent Nef inhibition and a therapeutic ratio similar to CMA (1). While a significant improvement to the therapeutic ratio was not observed in our best compound (33; slightly less than 2-fold relative to CMA), the excellent potency, and potential for low dose treatment with 33 could enable development of a viable therapeutic. Concanamycin derivatization efforts are being planned to further improve the therapeutic ratio.

Limitations and future research.

While our results are promising with respect to the identification and development of potent Nef inhibitors that restore MHC-I on infected primary cells from multiple unique donors, our conclusions are limited by the fact that our assays were performed in vitro. Thus, it is important to acknowledge that there are likely many challenges to translating the results presented here to the in vivo setting. For example, in vivo it is possible that CMA could be deglycosylated, leading to decreased efficacy. However, adventitious deglycosylation has not limited the effectiveness of other “macrolactone/glycosides” including FDA approved antibiotics erythromycin, clarithromycin, and azithromycin. In addition, while the compounds we tested appear to be efficacious in an overnight incubation in culture media containing 10% fetal bovine serum, it is possible that in vivo exposure to esterases in human plasma could result in hydrolysis of the acetylated derivatives. If these issues arise, more stable ester, amide or carbamate derivatives will be pursued based on well-documented examples for FDA approved antiretroviral therapeutics³⁰ and nucleoside anticancer drugs such as capecitabine³¹ (and numerous other pro-drugs) to enhance stability. Finally, future studies using suitable animal models are needed to determine the toxicity, accessibility, and efficacy of the concanamycin analogs reported in this work. In sum, future research is needed to address these and other related questions to determine the feasibility of utilizing these compounds as well as our SAR insights in guiding future drug development.

In this SAR study over 70 compounds were tested, including 29 macrolides, 21 16-membered plecomacrolides, and 23 18-membered plecomacrolides. By comparing structural variations of the natural and semi-synthetic compounds, we determined that 18-membered plecomacrolides with an intact pyran ring system, a C-8 methyl or ethyl group, a 4’-carbamoyl-β-D-rhamnose, and with additions of short acyl ester chains to both C-9-OH and C-3’-OH represent the current best opportunities for identifying improved lead compounds (Figure 8). Modifications to the macrocycle, or the pyran ring impact potencies in both 16- and 18-membered ring compounds. Sugar diversification might also facilitate a deeper understanding of the role this key moiety plays in achieving nM potency and the mechanism of action (Figure 8A). Differences in the way these compounds inhibit Nef-dependent MHC-I downmodulation versus lysosomal neutralization suggest there are structural parameters in the respective molecule(s) for each inhibitory pathway that remain to be identified at the molecular level. Previous work has confirmed that CMA (1) does not bind directly to the Nef:MHC-I:AP-1 complex, therefore, further studies are required to understand the relationship between Nef functional inhibition and V-ATPase binding/inhibition during plecomacrolide treatment⁶. Toward this end, semi-synthetic probes of lead plecomacrolide molecules are being developed for on-going proteomic analysis and target binding assays to determine the distinct pathways involved in HIV-Nef inhibition versus lysosomal neutralization.

Methods

Materials and General Information.

Unless otherwise noted, chemical reagents and solvents were purchased from EMD Millipore, Sigma- Aldrich, Oakwood chemical, Combi blocks, Chem Impex, Thermo-Fisher Scientific, AABlocks, Advanced Chem Blocks, TCI, or Arctom. Analytical samples purchased from Cayman chemical, with item numbers, include: bafilomycin B₁ (Baf B₁; 14005), bafilomycin C₁ (Baf C₁; 19625), elaiophylin (15583), PC-766B (20587), leucanicidin (25479), virustomycin (27736), and borrelidin (14436). A sample of archazolid was obtained from the Muller lab. Compounds 5–7, S1, and S6-30 were obtained from the Sherman lab.²⁹

All reactions, unless noted otherwise, were performed in flame dried glassware under dry nitrogen. Reactions run at elevated temperatures were controlled by IKA RET control visc (model RS 232 C). RT reactions were conducted between 21–23°C. NMR spectra were recorded on a Bruker 600 NMR System (600 MHz) either with a broadband or cryoprobe system. Chemical shifts are reported in parts per million (ppm) using the solvent resonance as an internal standard (CDCl₃ 7.26 and (CD₃)₂CO 2.05 ppm for ¹H NMR; CDCl₃ 77.16 ppm and (CD₃)₂CO 29.84 ppm for ¹³C NMR). Data are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, h = hextet), coupling constant (Hz), and integration. Both NMR systems were from the University of Michigan’s BioNMR core. High resolution mass spectra as well as analytical reaction analysis was performed on an Agilent Technologies 6250 TOF LC/MS equipped with an Agilent 1290 infinity II HPLC (monitored at 235 nm) at the Life Sciences Institute at the University of Michigan.

Analytical thin layer chromatography and preparative thin layer chromatography (pTLC) was performed on EMD Millipore 0.25 mm silica gel F254 plates and visualized by a UV lamp. Purifications were performed by flash column chromatography using EMD Millipore Silica Gel 60 (40 – 63 um) or a Biotage Isolera one flash purification system. Columns used with the Biotage Isolera system were SiliaSep and SiliaSepHP cartridges.

Ethics Statement.

Anonymized leukocytes isolated by apheresis were obtained from New York Blood Center and determined to be exempt from human studies requirements by the University of Michigan Institutional Review Board.

Preparation of Primary CD4 T Lymphocytes.

Peripheral blood mononuclear cells (PBMCs) were prepared and infected as previously described⁶. Briefly, they were prepared from anonymized buffy coats purchased from New York Blood Center using Ficoll-Paque Plus (17144002; GE Healthcare) centrifugation and SepMate tubes (85450; StemCell Technologies) according to the manufacturer’s protocol. CD8⁺ lymphocytes were depleted with Dynabeads according to the manufacturer’s protocol (11147D; Invitrogen) and the remaining cells were incubated at a density of 1×10⁶ cells per mL in 20 mL of R10 medium⁶ and stimulated with 10 ug/mL phytohemagglutinin-L (PHA-L) (431784; Millipore Sigma) in T25 cell culture flasks. Then, 20–24 h post-PHA activation, 15 mL of R10/PHA medium was removed and replaced with 5 mL of R10 + 100 U/mL recombinant Human IL-2 (202IL010; R&D Systems). 48 h after IL-2 stimulation, cells were infected via spinoculation with previously described HIV ΔGPE construct⁶. Infection ranged from about 3% to about 10% based on flow cytometric quantification of GFP expression by the reporter virus. The proportion of cells that were CD4⁺ at the time of harvest was >90% (Supplemental Figure 2).

Viral Construct & Infections.

Infectious supernatants for the HIV ΔGPE construct were prepared as described previously⁶. HIV ΔGPEN was constructed through a frameshift mutation created in the ΔGPE nef open reading frame by digesting the ΔGPE DNA construct with XhoI, filling in with Klenow, and re-ligating. Viral supernatants were produced by HEK-293T cells transfected using polyethylenimine (PEI) with the ΔGPE or ΔGPEN construct, the HIV packaging plasmid pCMV-HIV and a plasmid coding the broad tropism envelope glycoprotein G of Vesicular Stomatitis Virus (VSV-G) at a mass ratio of 1:1:1. Infections were performed by spinoculation as previously described⁶. Briefly, for spinoculation, cells are plated at a density of 1×10⁶ cells/well in a 24 well plate and spun for 2 h at 1050 × g in 1 mL of viral supernatant + 4 μg/mL hexadimethrine bromide (polybrene, H9268; Sigma-Aldrich). After spinoculation, cells are resuspended in 2 mL R10+50 u/mL IL-2 per well.

Concanamycin A–C Purification.

Twelve g of a raw mixture of concanamycin A, B, and C, (yellow powder) was purchased from Adipogen Life Sciences. Flash column chromatography (normal phase) was used to remove residual anti-foam and oils from the extract with a 12% isocratic IPA run for solvent A and 75% chloroform/25% hexanes solvent B. Fractions containing concanamycin analogs were separated and solvent removed in vacuo. The dried material (200 mg) was weighed and dissolved in 1 mL of ACS grade acetone. Sonication was used to ensure complete resuspension and after centrifugation sample was immediately injected onto a preparative HPLC (Shimadzu HPLC with Frac10 Collector) to maximize recovery of concanamycin analogs. This step was performed quickly, as over time concanamycin analogs degrade in acetone. The reverse phase separation utilized a Phenomenex Luna 5 μm C18(2) 100 A (250 mm × 30 mm) column and a time program of 52% isocratic over 70 mins (A: Water, no acid; B: ACN, no acid). Concanamycin analogs eluted as follows: CMB at 34 min, CMC at 46 min, and CMA at 56 min. Fractions with concanamycin analogs were separated and dried overnight using air or N₂ at a low flow rate to avoid flaking of dried crystals and loss of material recovery.

*See SI section concanamycins A–C and their derivatives for details on spectral data processing and caveats.

NMR data of CMA (1).

¹H NMR (599 MHz, 278 K, CDCl₃) δ 6.60 – 6.53 (m, 1H), 6.40 (s, 1H), 5.87 – 5.84 (m, 1H), 5.77 (d, J = 10.5 Hz, 1H), 5.68 (d, J = 9.8 Hz, 1H), 5.56 (dt, J = 15.0, 6.6 Hz, 1H), 5.29 (dd, J = 15.4, 7.9 Hz, 1H), 5.22 (dd, J = 15.1, 9.0 Hz, 1H), 5.01 (d, J = 9.2 Hz, 1H), 4.69 (d, J = 4.2 Hz, 1H), 4.55 (dd, J = 9.7, 1.8 Hz, 1H), 4.28 (t, J = 9.2 Hz, 1H), 4.04 (p, J = 6.2 Hz, 1H), 4.01 (s, 1H), 3.97 (t, J = 9.1 Hz, 1H), 3.86 (t, J = 9.1 Hz, 1H), 3.85 – 3.80 (m, 1H), 3.80 – 3.70 (m, 2H), 3.56 (s, 3H), 3.40 – 3.32 (m, 1H), 3.26 (s, 3H), 3.22 (d, J = 10.3 Hz, 1H), 2.73 (s, 1H), 2.32 (dd, J = 12.1, 4.8 Hz, 2H), 2.23 – 2.15 (m, 2H), 1.97 (s, 3H), 1.87 (s, 3H), 1.76 (s, 1H), 1.68 (dd, J = 12.2, 9.8 Hz, 1H), 1.61 – 1.56 (m, 3H), 1.54 – 1.48 (m, 1H), 1.28 (dd, J = 10.5, 6.9 Hz, 1H), 1.25 (d, J = 6.1 Hz, 3H), 1.19 – 1.10 (m, 3H), 1.09 (d, J = 6.6 Hz, 3H), 1.05 (s, 3H), 1.04 (s, 3H), 0.89 (d, J = 6.5 Hz, 3H), 0.85 (t, J = 7.6 Hz, 3H), 0.82 (d, J = 6.8 Hz, 3H). ¹³C NMR (151 MHz, 278 K, CDCl₃) δ 166.7, 157.6, 142.2, 141.8, 139.7, 133.4, 132.2, 130.8, 130.7, 127.8, 127.0, 123.0, 99.6, 96.4, 81.3, 79.9, 79.7, 76.1, 75.6, 75.5, 74.3, 70.4, 70.1, 69.5, 64.5, 59.1, 55.8, 43.3, 41.3, 41.2, 40.0, 39.8, 36.9, 36.3, 34.6, 22.8, 21.7, 17.8, 17.7, 16.8, 16.4, 14.2, 13.4, 11.7, 9.3, 7.1. LC-MS (IT): calcd for C₄₆H₇₅NO₁₄ [M+Fa-H]⁻ 910.5; found, 910.5.

NMR Data of CMB* (22).

¹H NMR (600 MHz, 278 K, CDCl₃) δ 6.56 (dt, J = 14.9, 9.2 Hz, 1H), 6.36 (d, J = 72.7 Hz, 1H), 5.92 – 5.73 (m, 2H), 5.67 (t, J = 10.2 Hz, 1H), 5.54 (dq, J = 18.6, 6.4 Hz, 1H), 5.32 – 5.24 (m, 1H), 5.20 (dd, J = 15.0, 9.4 Hz, 1H), 5.08 (d, J = 9.5 Hz, 1H), 5.01 (d, J = 9.5 Hz, 1H), 4.96 – 4.90 (m, 1H), 4.55 (d, J = 9.7 Hz, 1H), 4.45 (d, J = 4.1 Hz, 1H), 4.28 (t, J = 9.2 Hz, 1H), 4.08 – 3.94 (m, 2H), 3.86 (t, J = 9.4 Hz, 1H), 3.80 – 3.64 (m, 3H), 3.57 (d, J = 4.4 Hz, 3H), 3.47 – 3.42 (m, 1H), 3.36 (dd, J = 9.4, 6.1 Hz, 1H), 3.26 (d, J = 5.5 Hz, 3H), 2.32 (dd, J = 12.1, 4.6 Hz, 1H), 2.19 (dd, J = 12.3, 5.6 Hz, 2H), 2.16 – 2.04 (m, 1H), 2.00 (d, J = 36.5 Hz, 4H), 1.91 – 1.71 (m, 7H), 1.71 – 1.62 (m, 1H), 1.58 (d, J = 6.3 Hz, 3H), 1.30 – 1.25 (m, 1H), 1.24 (d, J = 6.3 Hz, 3H), 1.16 (d, J = 12.1 Hz, 1H), 1.12 (d, J = 7.2 Hz, 2H), 1.05 (dd, J = 21.6, 7.0 Hz, 6H), 0.94 (dd, J = 16.6, 6.9 Hz, 2H), 0.84 (ddd, J = 34.3, 18.7, 6.6 Hz, 9H). ¹³C NMR (151 MHz, CDCl₃) δ 166.1, 157.6, 141.0 (d, J = 432.2 Hz), 140.1 (d, J = 120.8 Hz), 133.2 (d, J = 95.6 Hz), 133.0 (d, J = 234.2 Hz), 130.8, 130.4 (d, J = 237.1 Hz), 128.0, 127.0, 124.4 (d, J = 367.3 Hz), 99.5, 96.5, 84.5, 80.9 (d, J = 95.4 Hz), 79.8, 76.0, 75.6 (d, J = 12.1 Hz), 75.4, 75.2 (d, J = 14.3 Hz), 70.3, 70.1, 69.5, 59.3 (d, J = 52.6 Hz), 55.8, 45.1 (d, J = 118.0 Hz), 41.3, 41.2, 40.0, 39.8, 37.1, 36.5 (d, J = 41.8 Hz), 34.9, 31.4, 22.1, 17.8, 17.6, 17.5, 16.0 (d, J = 61.3 Hz), 14.4 (d, J = 90.1 Hz), 13.4, 12.7, 9.3, 7.2. LC-MS (IT): calcd for C₄₅H₇₅NO₁₄ [M+Fa-H]⁻ 896.5; found, 896.5.

NMR Data of CMC (23)

¹H NMR (599 MHz, 278 K, CDCl₃) δ 6.56 (dd, J = 15.1, 10.7 Hz, 1H), 6.39 (s, 1H), 5.84 (s, 1H), 5.78 (d, J = 10.0 Hz, 1H), 5.67 (d, J = 9.8 Hz, 1H), 5.55 (dq, J = 13.2, 6.3 Hz, 1H), 5.29 (dd, J = 15.2, 8.0 Hz, 1H), 5.22 (dd, J = 15.1, 9.0 Hz, 1H), 5.01 (d, J = 9.2 Hz, 1H), 4.70 (d, J = 4.2 Hz, 1H), 4.60 – 4.55 (m, 1H), 4.05 – 4.00 (m, 1H), 3.97 (t, J = 9.1 Hz, 1H), 3.86 (t, J = 9.1 Hz, 1H), 3.82 (d, J = 10.0 Hz, 1H), 3.77 (td, J = 10.7, 4.7 Hz, 1H), 3.62 – 3.58 (m, 1H), 3.56 (s, 3H), 3.26 (s, 3H), 3.25 – 3.19 (m, 2H), 3.09 (t, J = 8.9 Hz, 1H), 2.75 – 2.69 (m, 1H), 2.33 (ddt, J = 19.8, 12.5, 7.6 Hz, 2H), 2.18 (dt, J = 10.7, 6.8 Hz, 1H), 2.12 (dd, J = 12.6, 5.0 Hz, 1H), 2.00 – 1.91 (m, 5H), 1.87 (s, 2H), 1.76 (d, J = 8.4 Hz, 1H), 1.66 – 1.61 (m, 1H), 1.58 (d, J = 6.4 Hz, 3H), 1.51 (dt, J = 11.8, 5.9 Hz, 1H), 1.30 (d, J = 6.1 Hz, 3H), 1.25 (d, J = 4.0 Hz, 1H), 1.16 (dt, J = 16.6, 9.7 Hz, 3H), 1.08 (d, J = 6.7 Hz, 1H), 1.05 (s, 2H), 1.04 (s, 0H), 0.89 (d, J = 6.5 Hz, 3H), 0.85 (t, J = 7.8 Hz, 2H), 0.82 (d, J = 7.1 Hz, 3H).¹³C NMR (151 MHz, 278 K, CDCl₃) δ 166.6, 142.1, 141.8, 139.6, 133.3, 132.2, 130.9, 130.7, 127.8, 127.1, 123.0, 99.6, 96.6, 81.3, 79.7, 77.6, 76.0, 75.5, 74.3, 71.8, 71.3, 70.1, 59.1, 55.8, 44.7, 43.3, 41.3, 41.2, 39.9, 39.5, 36.9, 36.3, 34.6, 22.7, 21.7, 17.8, 17.6, 16.8, 16.4, 14.2, 13.5, 11.7, 9.3, 7.2. LC-MS (IT): calcd for C₄₅H₇₄O₁₃ [M+Fa-H]⁻ 867.5; found, 867.5.

Baf A₁ Production and Purification.

Streptomyces lohii¹⁸ was cultured on MS plates (2.0% agar, 2.0% mannitol, 2.0% soya flour) for 3–5 days at 28°C until sporulation occurred. A single colony was inoculated in 10 mL of 2xYT media and incubated for 3 days at 28°C. The entirety of the 10 mL culture was used to inoculate bafilomycin production media (see Li et al for composition¹⁸). Growths were performed on 6L scales in baffled 2.8 L Fernbach flasks containing 1 L of bafilomycin production media per flask, where 1 pre-inoculum culture was needed for each. Fermentation was performed with agitation (160 rpm) at 22°C for 7 days. Centrifugation (4°C at 6200 rpm for 40 mins) was utilized to isolate the cells. The pellets were combined and broken by coating the cells with generous volumes of acetone and shaking for 4 h before allowing it to incubate overnight. Bafilomycin was further extracted and purified using the protocol reported by Painter et al⁶.

Compound Preparation.

All compounds were transferred to pre-weighed vials (weighed in triplicate) for dilution calculations prior to being dissolved in sterile DMSO (Sigma-Aldrich 276885). 5 mM or 10 mM solutions were made depending on the overall mass of the material and serial dilutions (in DMSO) were completed to achieve a 100 μM solution. Serial dilutions were vortexed after each step. 50 μL of the 100 μM solution was transferred to a glass vial insert (will add brand later) and stored at −20°C, if necessary, before testing. Thermo Scientific^™ E1-ClipTip Bluetooth electronic single channel pipettes were used throughout the entire sample preparation to avoid carry over of errors due to miscalibration or pipette tips variation.

Compound Treatments of Primary CD4⁺ T Cells.

48 h after spinoculation, cells are pooled and plated in a flat bottom 96-well plate. Mock cells are plated at 1×10⁵ cells/well in 100 μL R10+100 U/mL IL-2 while HIV ΔGPE infected cells are plated at 1×10⁵ cells/well in 50 μL R10+100 U/mL IL-2. Compounds are solubilized in sterile DMSO (276885; Sigma-Aldrich) to a concentration of 100 μM and frozen at −20°C for storage. Compounds are thawed at RT and diluted in R10 medium to 2x starting concentration. Compounds are then serially diluted in a 96-well round bottom plate in R10. Dilutions of compounds are then added to respective ΔGPE and Mock plates to bring final concentration of compounds to 1x and IL-2 concentration to 50 U/mL.

Flow Cytometry Staining.

Cells were analyzed as described previously⁶. Briefly, ΔGPE infected cells were collected 24 h after treatment. Cells were suspended in Lysotracker Red DND-99 (L7528; Invitrogen, 1:5000) and anti-HLA-A2 (BB7.2 from HB-82 hybridoma 0.5μg/mL) in phosphate buffered saline (PBS) for 1 h at 37°C³. The cells were washed once with FACS buffer (2% fetal bovine serum (26140079; Gibco), 1% human antibody serum (BP2525; Fisher), 2 mM HEPES (151630080; Gibco) 0.025% sodium azide (S8032; Sigma-Aldrich) in PBS lacking calcium and magnesium). After washing, the cells were resuspended in 4 ng/mL DAPI (4’,6-diamidino-2-phenylindole,62248; Thermo Scientific) for viability and 1:1000 goat anti-mouse IgG2b-Alexa Fluor 647 (A21242; Invitrogen) in FACS buffer for 5 min at RT, washed with FACS buffer, then finally fixed in 2% paraformaldehyde for 30 min at RT.

72 h viability was assessed as described previously using a flow cytometric assay measuring update of a vital dye. This assay was previously validated against the standardly used MTT assay⁶. Briefly, donor-matched mock infected cells were collected 72 h after treatment. Cells were suspended in 4 ng/mL DAPI in FACS buffer for 20 min on ice, washed once with FACS buffer and fixed in 2% paraformaldehyde for 30 min at RT.

CD4 expression was assessed by staining mock transduced primary T cells that were CD8 depleted and activated with IL-2 and PHA. Cells were suspended in anti-human CD4-APC conjugated antibody (17-0048-42; Invitrogen, 1:400) in FACS buffer and incubated for 30 min. The cells were washed once in FACS buffer. After washing, the cells were resuspended in 4 ng/mL DAPI in FACS buffer for 5 min RT, washed with FACS buffer and fixed in 2% paraformaldehyde for 30 min. Primary T cells prepared for these experiments were >90% CD4⁺ (Supplemental Figure 2).

In all experiments, flow cytometry data was collected with a BioRad Ze5 cytometer and data was analyzed using FlowJo software (BD Life Sciences). As shown in Supplemental Figure 1, Cells were gated sequentially by forward scatter vs. side scatter for cells, doublet exclusion (forward scatter area vs. height), and exclusion of DAPI for viable cells. GFP negative (uninfected) and GFP positive (infected) populations were analyzed for expression levels of MHC-I (HLA-A2). 72 h viability was assessed as shown in Supplemental Figure 1 using donor-matched uninfected cells.

Calculation and Statistical Analysis.

$$\text{MFI}=\text{median}\text{fluorescence}\text{intensity}\text{of}\text{MHC-I}(\text{HLA}-\mathrm{A}2)$$

$$\text{Fold}\text{downmodulation}={\mathrm{M}\mathrm{F}\mathrm{I}}_{\text{uninfected}}/{\mathrm{M}\mathrm{F}\mathrm{I}}_{\text{infected}}$$

$$\text{Normalized}\text{Nef}\text{activity}=\text{Fold}\text{downmodulation}{\text{MHC-I}}_{\text{sample}}/\text{Fold}\text{downmodulation}{\text{MHC-I}}_{\text{solvent}}$$

All statistical analyses were performed using GraphPad Prism. Curves were generated using GraphPad Prism software using [Inhibitor] vs. response with variable slope (four parameters).

Outliers were removed with the Prism algorithm ROUT. IC₅₀s and ratios of Lysotracker IC₅₀ to Nef inhibition IC₅₀ and viability TC₅₀ to Nef inhibition IC₅₀ were calculated for individual experiments. Statistical significance for CMA comparisons was assessed using 1 way ANOVA with Brown-Forsythe and Welch’s correction for unequal standard deviations and Dunnett’s multiple comparisons test as described in the legends. Where there were sufficient donor-matched data points for bafilomycins, CMB and CMF and their derivatives, a 1 way ANOVA mixed effects analysis with Dunnett’s multiple comparisons test was performed as described in the legends.

We assessed the activity of each inhibitor on a minimum of three donors. In cases where some inhibitors were tested on additional donors, sometimes as controls in follow up experiments, all the data from all donors was included. In our analysis, “n” represents the number of unique individual donors. In some cases where there are different n values reported for IC₅₀, neutralization ratio and therapeutic ratio, we either did not perform one of the tests or there were technical problems with the test such that the data could not be analyzed.

Purity Analysis.

Purity of active compounds was assessed by analytical HPLC (Shimadzu autoinjector). A Luna 5μ C18 (250 × 4.6 mm, 5μ) column was used. Compounds were visualized at 280 nm unless otherwise noted. Samples were prepped in a mixture of DMSO and water unless otherwise noted. The flow rate was 2 mL/min and the column was warmed to 40°C with and injection volume of 10 μL. Four methods were used. Method 1. Solvent A= water; B=MeCN with the following gradient in relation to %B: 10% (0–2 min), 10–60% (2–4.3 min), 60% (4.3–10 min), 60–100% (10–12 min), 100% (12–18 min). Compounds were visualized at 280 nm unless otherwise noted. Method 2: Solvent A= water (0.1% FA); B=MeCN (0.1% FA). Gradient in relation to %B: 5% (0–1 min), 5–100% (1–6 min). Compounds were visualized at 280 nm unless otherwise noted. Method 3: Solvent A= water (0.1% FA); B=MeCN (0.1% FA). Gradient in relation to %B: 15% (0–1 min), 15–100% (1–6 min), 100% (6–9 min). Visualization wavelength is denoted with each trace. Method 4: Solvent A= water (0.1% FA); B=MeCN (0.1% FA). Gradient in relation to %B: 15% (0–1 min), 15–100% (1–10 min), 100% (10–12 min). Visualization wavelength is denoted with each trace.

Abbreviations

CMA

concanamycin A

CMB

concanamycin B

CMC

concanamycin C

CMF

concanamycin F

CMX

concanamycin X

Baf

bafilomycin

DIAD

Diisopropyl azodicarboxylate

EDC

1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide

FA

formic acid

FCC

flash column chromatography

IPA

isopropyl alcohol

Lyso

lysotracker

pTLC

preparative TLC

RP

reverse phase

SM

starting material

TPAP

Tetrapropylammonium perruthenate