CD437

Structure−Activity Relationship and Anticancer Profile of Second- Generation Anti-MRSA Synthetic Retinoids

Ana V. Cheng,† Wooseong Kim,‡ Iliana E. Escobar,‡ Eleftherios Mylonakis,‡ and William M. Wuest*,†,§

ABSTRACT: We previously reported the antibacterial activity of CD437, a known antitumor compound. It proved to be a potent antimicrobial agent effective against both growing and persister cells of methicillin-resistant Staphylococcus aureus (MRSA). Herein, we report the synthesis of a panel of analogs and their effect on both MRSA and cancer cells. The hydrophobic group of the parent compound was varied in steric bulk, and lipid-mimicking analogs were tested. Biological assessment confirmed that the adamantane moiety is the most effective substitution for antibacterial activity, and some preferential action in cancer over MRSA was achieved.

Retinoids are molecules derived from vitamin A, which is essential to mammalian growth, immunity, and vision.1
Figure 1. Resistance and persistence. (a) In a population of cells, only those with a beneficial mutation survive antibiotic treatment and then grow a population of genetically identical, resistant cells. (b) Persister cells withstand treatment then revert to their active state. They grow a population of susceptible cells, but retreatment is met with new persisters, resulting in a recurrent infection.
Several synthetic retinoids, such as adapalene and tazarotene, are used for treatment of skin disorders, including psoriasis and acne.2 CD437, a third generation synthetic retinoid, has been studied for its potent anticancer activity.3,4 Recently, Han and
Figure 2. Structures of CD437, its primary alcohol derivative, and ST1926 (a related compound studied for its anticancer activity).
Figure 3. Steric bulk panel. (a) Analog structures. (b) Biological results: minimum inhibitory concentration (MIC) against MW2MRSA cells, median lethal concentration (LC50) in human hepatoma cells (HepG2), and median hemolytic concentration (HC50). For reference: CD437 MW2 MIC = 1 μg/mL, HepG2 LC50 = 1.02 μg/mL, HC50 > 64 μg/mL; Analog 5 HepG2 LC50 = 1.7 μM co-workers showed that CD437 exerts its anticancer activity by directly binding to POLA1,5 contradicting previous theories that it kills cancer cells by agonizing the gamma isoform of retinoic acid receptor (RARγ).6 POLA1, a subunit of DNA polymerase α, initiates DNA synthesis by extending the RNA primer by 10−20 nucleotides. Through a seemingly selective mechanism that has yet to be elucidated, this promotes apoptosis in cancer cells but not in healthy cells.

We previously reported the antibacterial activity of CD437, which we identified using a whole animal Caenorhabditis elegans−MRSA infection model.7 Staphylococcus aureus has emerged as a serious global health threat. Thirty percent of the population is colonized by S. aureus, and pathogenic infections can cause a variety of diseases from impetigo and osteomyelitis to toxic shock and sepsis.8 MRSA strains possess the staphylococcal cassette chromosome mec (SCCmec). This Figure 4. Lipid-mimicking analogs. (a) Structures of unsaturated and saturated analogs. (b) Biological results: minimum inhibitory concentration (MIC) against MW2 MRSA cells, median lethal concentration (LC50) in human hepatoma cells (HepG2), and median hemolytic concentration (HC50) cassette contains the mecA gene, which encodes penicillin- binding protein (PBP) 2a, an altered PBP with decreased affinity for all β-lactam antibiotics.9,10 Resistance to vancomy- cin, a glycopeptide antibiotic and the next line of defense against MRSA, has also emerged as a result of acquisition of the vanA gene cluster from vancomycin-resistant enterococci (VRE).11 Bacterial persistence (a form of tolerance) can also complicate the treatment of S. aureus and many other infections. First observed in 1944,12 these “dormant” cells are nongrowing and tolerant of most antibiotics. They are distinct from resistant cells (Figure 1a) in that their physiological state is reversible, and they are genetically identical to susceptible cells. Colonies regrown from a persister population are indeed susceptible to antibiotic, as opposed to a population borne of resistant cells, which retain the resistance mutations of their predecessors (Figure 1b).

Persisters are implicated in chronic infections and are found in biofilms.13 Unfortunately, most existing therapies target metabolic processes that are suspended in these transient subpopulations of bacteria. We have previously shown that CD437 (Figure 2) displays activity against MW2, a community-associated MRSA strain, with a minimum inhibitory concentration (MIC) of 1 μg/mL and a low probability of resistance selection. Addi- tionally, CD437 kills stationary phase cells (used to approximate persister cells) from 13 clinical isolates (at 10× MIC) as well as 90% of persisters in MRSA biofilms (at 16× MIC). Preliminary structure−activity relationship (SAR) studies show the importance of the phenol and yielded a primary alcohol derivative (Figure 2), which retained antibacterial activity (MIC = 2 μg/mL) with significantly decreased toxicity in primary human hepatocytes.7 CD437 possesses dual applications, it kills by different mechanisms in cancer and bacteria. Thus, we could feasibly tune the structure for activity in one over the other. Previous SAR studies of a closely related scaffold (ST1926, Figure 2) in various cancer cell lines determined that removal of adamantane or replacement with other hydrophobic groups led to a decrease in anticancer activity.

Antibacterial mechanism of action studies indicated that the retinoids target bacterial membranes. Based on all atom molecular dynamics simulations,7 we hypothesized that the hydrophobic adamantyl moiety is essential to this action and functions by interacting with the lipid tails in the membrane. Considering these two studies, we chose the hydrophobic adamantane as the focal point of our dual SAR study. To explore its role in both MRSA and cancer, we synthesized analogs with varied groups at this position. The role of three-dimensional bulk was probed by analogs featuring both smaller and larger hydrophobic groups than the adamantane present in CD437. A line of lipid- mimicking analogs with unsaturated and saturated alkyl chains of varying lengths was also synthesized with the goal of altering protein binding in polymerase α while inducing a stronger interaction between the alkyl chains and the lipid tails in bacterial membranes. For all analogs, both the carboxylic acid and primary alcohol derivatives were synthesized to investigate whether the decreased toxicity of the primary alcohol derivative of CD437 was a trend, and whether it resulted in lowered anticancer potency as previously observed15 in similar retinoids. To this end we synthesized and tested 20 analogs and highlight the findings below.

Each analog was made in 3−6 steps following the same general synthesis (Scheme 1): the 2-substituted 4-bromophe- nols were assembled and protected, borylated for use in a Suzuki cross coupling, then deprotected. Each analog was also reduced to yield the primary alcohol derivatives. For the first group, commercially available 4-bromophenol, 4-bromo-2- methylphenol, and 4-bromo-2-(tert-butyl)phenol were pro- tected with 2-methoxyethoxymethyl ether (MEM) before Miyaura borylation and subsequent Suzuki coupling to methyl-6-bromo-2-naphthoate. The phenol and acid were deprotected, and a portion of each was reduced to yield analogs 1−3 and 5−7. Similarly, 4-bromo-2-(3,5-dimethyl-1-
adamantyl)phenol was synthesized first using a Friedel−Crafts alkylation, then was carried through the aforementioned route. For the lipid-mimics, 5-bromosalicylaldehyde was MEM protected before Wittig reactions were used to append ylides Figure 5. Additional biological testing. (a) Minimum inhibitory concentration (MIC) of analog 4 in additional MRSA strains (μg/ mL). (b) Time killing of persister cells. MRSA MW2 persisters were treated with the indicated concentrations of each retinoid. Viability was measured by serial dilution and plating on TSA plates. The data points on the x-axis are below the level of detection (2 × 102 CFU/ mL).

Results are shown as means ± s.d.; n = 3. (c) Killing of HeLa cervical cancer cells hydrophobic group intercalates into the membrane, facilitating retinoid embedment.7 Analogs possessed cLogP values between 3.75 and 8.75 (Table S1), with 1 and 5 the lowest, making them the least likely to be localized in the lipid bilayer (see Table S1). The larger substitutions in analogs 3, 4, 7, and 8 displayed modest activity against MRSA, indicating that increased steric bulk leads to increased antibacterial activity. However, none were more active and less toxic than CD437 and its alcohol derivative, proving that the original adamantyl substitution is optimal for membrane disruption. In the case of the lipid-mimics (Figure 4a), we were curious about whether the increased rigidity of the unsaturated analogs had any effect since we had previously observed favorable results with a benzyl substitution.7 We also sought to understand what effect chain length might have on anti-MRSA activity. It has been previously observed that small molecules that function by lysing bacterial cells benefit from a 10−12 carbon length chain mimicking the phospholipid composition.16 However, we have shown that retinoids do not lyse bacterial cells,7 thus shorter carbon chains were used. Again, modest activity (in comparison to CD437) was observed (Figure 4b). From analogs 9−11 it might be deduced that increased chain length correlates with antibacterial activity. However, this trend is less pronounced between analogs 15−17 and in the primary alcohol analogs.

Analogs 12−14 and 18−20 instead experience a steep drop-off in bacterial killing with the longest length chains. We imagine that the acid analogs more rigidly anchor the retinoids to the bacterial membrane, enforcing this chain length trend, while the primary alcohol analogs are more flexible, rendering the chain length less important. These results indicate that, although the lipid mimics exert anti- MRSA activity, a more voluminous substitution is better able to disrupt the bacterial lipid bilayer. We further tested our most active analog, 4, in another common MRSA strain (JE2), a healthcare-associated MRSA strain (ATCC 33591), and a vancomycin-resistant strain (VRS1) (Figure 5a). Analog 4 maintained a similar activity profile to CD437, with a MIC of 2 μg/mL in all strains. Likewise, 4 killed MW2 persister cells with almost identical speed to CD437, eradicating 5 × 107 CFU/mL MRSA persisters within 1 h (Figure 5b). In most cases, anti-MRSA activity loosely correlates with HepG2 killing. However, primary alcohol analogs 14 and 20 appear to be selective for HepG2 killing over MRSA killing. No trends can be extrapolated on the basis of primary alcohols since the primary alcohol derivative of CD437, which we previously reported, showed the reverse selectivity.7 We further examined the anticancer activity of 4, 14, and 20 as well as that of 16, our most active nontoxic analog, in HeLa cells (Figure 5c). In all cases, we observed lower LC50 values than in HepG2 cells. Most active were 4 (LC50 = 0.49 μg/mL) and 16 (LC50 = with 4-, 6-, and 8-carbon length chains.

A hydrogenation yielded fully saturated analogs 15−20 in addition to 9−14. (Full synthetic details are provided in the Supporting Information.) Antimicrobial and antineoplastic activity of the analogs were tested using MRSA MW2 and a human liver cancer cell line, HepG2. This testing confirmed the importance of the hydrophobic moiety to antibacterial action as analogs 1, 2, 5, and 6 (Figure 3a), which featured no hydrophobic group or only a methyl, exert no killing against MW2 cells (Figure 3b). This finding provides further evidence for the putative method of membrane disruption: previous simulations suggest that the 0.60 μg/mL), two carboxylic acid analogs. However, CD437 remained the most potent (LC50 = 0.21 μg/mL), supporting a previous study of a similar scaffold, which observed decreased apoptotic activity in analogs lacking a carboxylic acid or with different hydrophobic groups.

In conclusion, these results verify the importance of a sterically large hydrophobic group to the anti-MRSA activity of synthetic retinoids. Adamantane was revealed to be the optimal size, as CD437 and its primary alcohol derivative are the most active compared to any other acid-alcohol pair tested, while complete removal of the hydrophobic group resulted in loss of antibacterial activity. Although analogs furnished with lipid- mimicking groups displayed modest bacterial killing, they did not have the predicted selectivity for antibacterial over anticancer activity. Generally, the carboxylic-primary alcohol analog pairs show similar activity (with the exceptions of 11/ 14 and 17/20). Perhaps most interestingly, almost all analogs show low or no hemolytic activity, despite their bacterial membrane disruption ability. However, no clear trend in toxicity or cancer killing can be gleaned. Using the information described here, we will continue to explore synthetic retinoids as an exciting potential MRSA and MRSA persister therapy.

Author Contributions
All authors designed experiments. A.V.C., W.K., and I.E.E. carried out experiments. A.V.C. wrote the manuscript, and all authors have approved the final version.

Funding
This work was funded by the National Institute of General Medical Sciences (GM119426) and the Georgia Research Alliance based in Atlanta, Georgia to W.M.W., and by National Institutes of Health grant P01 AI083214 to E.M. The NMR instruments used in this work were supported by the National Science Foundation (CHE1531620).