WhiB7 expression results in enhanced FF-NH2 activity against M. abscessus

To assess the impact of the WhiB7 regulon on CAM activity in M. abscessus, we conducted susceptibility testing using M. abscessus ATCC19977 (wild type (WT)) and an isogenic MAB_3508c deletion strain (∆whiB7). The potency of CAM increased fivefold against ∆whiB7, like that of amikacin (AMK) control (3.7-fold), which is subject to WhiB7-dependent N-acetylation encoded by MAB_4532c (eis2) (Fig. 1 and Extended Data Fig. 1a). This suggested a WhiB7-dependent element impacted CAM susceptibility. To investigate this possibility, we performed RNA sequencing (RNA-seq) in the WT strain using a sub-minimum inhibitory concentration (MIC) of CAM for 30 min. CAM strongly induced the WhiB7 regulon, notably increasing the expression of a WhiB7-regulated O-acetyltransferase MAB_2989 (log fold change (FC) 2.7, false discovery rate (FDR) <0.01; Extended Data Fig. 1b), which has previously been inferred as a probable CAM acetyltransferase (further denoted as cat). Transformation of Mycobacterium smegmatis with the multicopy vector pOLYG-aac(3)IV-cat decreased CAM activity eightfold compared with pOLYG-aac(3)IV vector control (MIC of 128 µg ml versus 16 µg ml) (Supplementary Table 1). An isogenic MAB_2989 deletion was then constructed in M. abscessus (∆cat) showing an increase in CAM activity when compared with the WT strain (Fig. 1a), suggesting that cat mediates CAM resistance. Similarly, clinical isolates of the M. abscessus complex that possessed cat frameshift mutations were more susceptible to CAM (Supplementary Table 1). To determine whether this observation was CAM specific or applies more broadly to other phenicols, we then tested florfenicol (FF) and thiamphenicol (TAM), which differ by a single fluorine replacement, against these strains. While TAM activity was enhanced against ∆cat, similar to ∆whiB7, FF activity remained unchanged (Fig. 1a), indicating that the fluorine substitution of the C3-hydroxyl in FF prevents inactivation by Cat.

To facilitate future chemistry campaigns, we synthesized amine derivatives of these phenicols to serve as versatile intermediates for subsequent chemical modifications. Routine M. abscessus susceptibility testing of these intermediates revealed FF-NH, a major metabolite of FF in many domesticated animals, to have similar activity compared with FF in WT and ∆cat strains, displaying a MIC of 64 µg ml or 225 µM (Fig. 1b). In contrast, chloramphenicol amine (CAM-NH) and thiamphenicol amine (TAM-NH) were only active against the ∆cat strain (Fig. 1b). Surprisingly, the activity of FF-NH was dependent on WhiB7 but in an unexpected manner: FF-NH was 7.6-fold more potent against the WT (half-maximum inhibitory concentration (IC) of 17.8 µg ml or 62.7 µM) compared with the ∆whiB7 strain (IC of 136 µg ml or 479 µM). Similar decreases in activity were observed for CAM-NH and TAM-NH in the ∆whiB7 strain compared with the ∆cat strain, despite WhiB7's established function as a positive regulator of cat expression (Fig. 1b). These results suggest that WhiB7 expression is a vulnerability in the context of amine phenicol treatment. Given the role of WhiB7 in antibiotic resistance in M. abscessus, we sought to explore the mechanism by which WhiB7 enhances FF-NH activity.

We began by selecting FF-NH-resistant M. abscessus mutants on agar plates, which arose at a frequency of 1 × 10 cells. At a concentration of 64 or 128 µg ml FF-NH, there were noticeable differences in colony size and morphology reflecting two separate populations: (1) a large, rough morphotype and (2) a small, smooth morphotype (Fig. 2a and Extended Data Fig. 2a). At a concentration of 256 µg ml FF-NH, only the small morphotype was observed. Sequencing analysis revealed that the large colonies harboured mutations in the transcription factor whiB7, and the small colonies had mutations in the WhiB7-dependent N-acetyltransferase eis2 (Fig. 2a-c and Supplementary Table 2). Notably, these phenotypic changes were transient, as passaging whiB7 and eis2 mutants in antibiotic-free media reverted the colonies to the smooth morphotype seen with the WT strain (Extended Data Fig. 2b,c). For the whiB7 mutants, we identified multiple non-synonymous single nucleotide polymorphisms (SNPs) and insertion-deletions impacting conserved WhiB7 elements. These included mutations near the iron-binding cysteine residues (H16P, L37P and A51T), GVWGG β-turn motif (G59R and G62C/A) and the DNA-binding AT-hook (R76del and P77fs) (Fig. 2b). Among the eis2 mutants, a substantial proportion (27 out of 45 unique mutations) exhibited substantial structural changes due to frameshift mutations and the introduction of premature stop codons (Supplementary Table 2). Additional SNPs identified in the eis2 gene conferred amino acid changes near the acetyl-CoA binding site (V83D, R90W, E120K and G128V) (Fig. 2c). Notably, no mutations were found in the 23S rRNA CAM binding site, indicating that FF-NH resistance is probably mediated through whiB7 or eis2 disruption.

To confirm Eis2 expression is responsible for amine phenicol activity, we conducted susceptibility testing using both isogenic deletion strains (∆eis2 and ∆whiB7) and spontaneously generated missense mutants (Eis2 T258R and WhiB7 A51T), each complemented with either the multicopy pOLYG-aac(3)IV-eis2 vector or the pOLYG-aac(3)IV control vector. Introduction of pOLYG-aac(3)IV-eis2 restored FF-NH activity in each strain tested (Fig. 2d,e). CAM-NH also showed improved activity in the pOLYG-aac(3)IV-eis2 complemented strains relative to pOLYG-aac(3)IV control, whereas the antimicrobial activity of clarithromycin (CLR) and linezolid (LZD) remained unchanged (Fig. 2d,e and Extended Data Fig. 3a,b). In contrast, AMK was less potent in strains complemented with pOLYG-aac(3)IV-eis2 than in those with the pOLYG-aac(3)IV control vector (Fig. 2d,e). These results confirm that Eis2 increases efficacy of FF-NH and decreases AMK activity.

To demonstrate that FF-NH is enzymatically modified by Eis2, we performed a biochemical kinetic assay using Eis2 purified from Escherichia coli. Our results confirm that FF-NH is a substrate for Eis2, with a k of 11.7 ± 0.9 s and a specificity constant (k K) of 6.7 × 10 ± 5.1 × 10 M s (Fig. 2e, Supplementary Fig. 4 and Supplementary Table 3). Consistent with previous findings, the aminoglycoside controls AMK and hygromycin B were also shown as substrates of Eis2 (ref. ), as well as CAM-NH and TAM-NH. Notably, these compounds all displayed higher catalytic efficiencies compared with FF-NH (Fig. 2f, Extended Data Fig. 4a-c and Supplementary Table 3). This observation, alongside the enhanced activity of CAM-NH and TAM-NH against the ∆cat strain (Fig. 1b), suggest that these compounds would display antimicrobial activity if they were not also substrates for Cat. In contrast, the dichloroacetyl parent molecules FF, CAM and TAM were not identified as substrates for Eis2 (Fig. 2f, Extended Data Fig. 4a-c and Supplementary Table 3). These findings highlight how specific chemical modifications to antibiotics govern substrate specificity and antimicrobial efficacy.

We hypothesized that Eis2, a GCN5-related N-acetyltransferase (GNAT), acetylates FF-NH in M. abscessus cells, producing the active metabolite FF acetyl (FF-ac). To test whether Eis2-mediated acetylation of FF-NH is necessary for ribosomal inhibition, we conducted a cell-free translation inhibition assay using S30 extracts from M. smegmatis. The results revealed that FF and CAM display similar potency for inhibiting mycobacterial ribosomes (absolute IC of 16.9 and 20.2 µM, respectively), whereas FF-ac showed slightly reduced potency (absolute IC of 39.2 µM) (Fig. 2g and Supplementary Table 4). However, FF-NH displayed markedly decreased activity (absolute IC > 800 µM) (Fig. 2g and Supplementary Table 4), confirming that acetylation of FF-NH to FF-ac by Eis2 is crucial for ribosome inhibition and antimicrobial activity.

Given that FF-NH is a weak ribosomal inhibitor whose activity depends on WhiB7 and Eis2 expression, we hypothesized that its activation is WhiB7 and time dependent. To test this, we took a multifaceted approach utilizing a platform consisting of RNA-seq, quantitative proteomics and an accumulation/conversion assay, all conducted in exponentially growing M. abscessus cells over a 3 h period (Fig. 3a).

Our RNA-seq analysis revealed that FF-NH treatment at 100 µM (approximately 1/2MIC) upregulates WhiB7 target genes at 0.5 and 3 h compared with vehicle control (Fig. 3b,c). Specifically, at 0.5 h, 104 differentially expressed genes (logFC >|2|, FDR <0.01) were WhiB7 targets, which increased to 128 differentially expressed genes after 3 h. These results demonstrate that FF-NH strongly induces the transcription of genes involved in amine phenicol activity, including eis2, cat and whiB7 (Fig. 3b,c). We also compared the transcriptomic profiles for M. abscessus cells treated with 25 µM FF-NH (approximately 1/8MIC) or FF. While FF treatment induced similar levels of whiB7 and eis2 at both 0.5 h and 3 h, FF-NH treatment resulted in a more pronounced upregulation of these genes over the same period, highlighting the time-dependent nature of M. abscessus's response to FF-NH (Extended Data Fig. 5a,b). The direct comparison of the 3 h FF and FF-NH treatment profiles identified only 11 significantly dysregulated genes, suggesting a consistent mechanism of action for both compounds (Extended Data Fig. 5c,d).

To confirm that FF-NH-induced transcriptomic changes were reflected at the protein level, we analysed the proteomes of M. abscessus cells treated with 100 µM FF-NH at 0.5 h and 3 h utilizing data-independent acquisition mass spectrometry (MS). The results matched the transcriptomic profiles, showing 18 WhiB7 target proteins were differentially expressed (logFC >|1|, FDR <0.01) at 0.5 h, increasing to 39 WhiB7 target proteins by 3 h after FF-NH treatment (Fig. 3d,e). The abundance of Eis2, Cat and WhiB7 were significantly increased at both timepoints (FDR <0.01; Fig. 3d,e), with Eis2 induction rising from logFC of 1.45 at 0.5 h to 3.4 at 3 h. Together, these transcriptomic and proteomic studies demonstrate that FF-NH induces its activating enzyme, Eis2, in a time-dependent manner.

We next aimed to test whether the accumulation of active metabolite FF-ac also occurred over time compared with FF by using a MS-based accumulation/conversion assay. M. abscessus WT cells were exposed to 100 µM FF or FF-NH for up to 3 h, and cell lysates were analysed for FF, FF-NH and FF-ac. We find that accumulation of FF occurs rapidly (within 10 min) and remains consistent (180-320 nM recovered) throughout the 3 h experiment (Extended Data Fig. 6a). In contrast, FF-ac levels from FF-NH treatment increased over time, starting at 83.5 ± 25 nM recovered at 10 min and rising to as high as 4,900 ± 2,000 nM at 3 h (Fig. 3f). FF-NH concentrations also showed a modest increase in accumulation throughout the experiment (Fig. 3g). This could result from drug-induced changes in uptake processes or from the conversion of FF-ac back to FF-NH, either through biological mechanisms (for example, enzymatic deacetylation) or technical factors during MS analysis (for example, in-source fragmentation). Regardless, these observations reinforce the central finding that FF-ac is the primary accumulated species in M. abscessus cells. We next assessed the accumulation of FF-ac from FF-NH in the absence of this activation mechanism by testing the ∆whiB7 and ∆eis2 strains. We observed minimal accumulation for either species, with most samples falling below the limit of quantification across the duration of the experiment (Fig. 3f,g). To further verify the role of Eis2 in the accumulation of FF-ac in M. abscessus cells, we repeated the experiment using the ∆eis2 strain complemented with the pOLYG-aac(3)IV-eis2 or pOLYG-aac(3)IV control vectors. The addition of pOLYG-aac(3)IV-eis2 vector restored the time-dependent accumulation of FF-ac and FF-NH, while the control vector had no effect (Extended Data Fig. 6b).

These results highlight the unique mechanism of action displayed by FF-NH, where the Eis2-mediated conversion of FF-NH to FF-ac drives antimicrobial activity by generating a more active ribosome inhibitor, thereby inducing further Eis2 expression through WhiB7. This conversion not only enhances the antimicrobial potency, but also facilitates the accumulation of FF-ac within the cell, creating a perpetual feed-forward loop resulting in the amplified antimicrobial action of FF-NH (Fig. 4).

Broadly, we observed that susceptibility patterns for FF-NH-resistant strains with mutations in either whiB7 or eis2 closely mirrored those of their respective isogenic deletion strain (Fig. 5a). This suggests that these genomic alterations result in loss of function, effectively recapitulating the knockout phenotype. As expected, mutations in either eis2 or whiB7 increased the potency of all tested aminoglycosides/peptide antibiotics (AMK, kanamycin A/B, hygromycin B and capreomycin), with up to a 16-fold increase in potency observed for capreomycin (Fig. 5a,b and Extended Data Fig. 8). Additionally, strains harbouring whiB7 mutations showed increased susceptibility to non-fluorinated phenicols (CAM/TAM) and SPC, whereas strains with eis2 mutations did not (Fig. 5a,b and Extended Data Fig. 8a-s). This differential susceptibility is attributed to the role of WhiB7 in mediating resistance through Cat for non-fluorinated phenicols and the efflux pump TetV for SPC. Increased activity for macrolides -- azithromycin (AZITH), CLR and erythromycin -- was not observed in the whiB7 mutants during the 6 day assay, probably because WhiB7-mediated erm(41) resistance requires longer durations to manifest. However, under standard 14 day MIC conditions, macrolide susceptibility improved 4-32-fold in the ∆whiB7 and WhiB7 A51T strains compared with the WT, ∆eis2 and Eis2 T258R strains (Supplementary Table 6), aligning with previous studies. As no mutations were found in the ribosomal target site, these results suggest that resistance to FF-NH comes at the cost of collateral susceptibility to SOC antimicrobials, highlighting its 'antiresistance' properties.

Given that M. abscessus infections are typically treated using a combination of antimicrobials, we sought to determine the potential of pairing FF-NH with SOC anti-M. abscessus therapeutics through an in vitro high-throughput combination screening platform. Using the response surface method, bivariate response to additive interaction doses (BRAID), we found that FF-NH strongly synergized (κ = 2.16) with LZD, as well as displayed moderate synergy with representative macrolides (AZITH and CLR), and the third generation tetracycline analogue, eravacycline (Fig. 5c). Conversely, FF-NH was found to be antagonistic with AMK and cefoxitin (Fig. 5c). The observed synergy between different classes of protein synthesis inhibitors, excluding aminoglycosides that typically show antagonism with bacteriostatic ribosomal agents, aligns with existing literature. However, it is unclear how FF-NH and LZD, both presumed to act at the ribosomal PTC, display such a high degree of synergy. Nonetheless, at various sub-IC concentrations, FF-NH displayed improved potency and significant synergy with LZD, as well as CLR, as also determined by Bliss independence model (Fig. 5d,e). These findings highlight the potential of FF-NH to be effectively combined with existing anti-M. abscessus therapies.

The requirement for FF-NH to be activated by Eis2 led us to hypothesize that it would lack the mitochondrial protein synthesis (MPS) inhibition liability of other PTC-targeting agents. We therefore profiled FF-NH, FF-ac and other analogues for MPS inhibition activity. FF-ac exhibited potent MPS inhibition (IC 4.7 ± 1.7 µM), similar to FF (IC 1.2 ± 0.2 µM), CAM (IC 5.5 ± 2.5 µM), TAM (IC 1.4 ± 0.7 µM) and LZD (IC 6.2 ± 1.6 µM) (Supplementary Table 7). However, FF-NH, as well CAM-NH and TAM-NH, displayed markedly decreased activity in this assay (IC >100 µM; Supplementary Table 7). Further, an in vitro cytotoxicity assay showed FF-NH to be non-cytotoxic at the top concentrations tested (IC >200 µM) (Supplementary Table 7). These results indicate that FF-NH's prodrug mechanism may mitigate the toxicity typically associated with PTC-targeting antimicrobials.

This mechanism parallels recent work on 5-aminomethyl oxazolidinone prodrugs, which are also activated by a mycobacterial-specific N-acetyltransferase and show limited MPS inhibition. However, those compounds were found to undergo oxidative deamination in vivo, producing toxic 5-alcohol metabolites with potent MPS activity. To assess whether FF-NH shares this vulnerability, we evaluated its metabolic stability and metabolite profile in hepatocytes, the primary site of drug metabolism. FF-NH demonstrated high stability in human hepatocytes (t > 440 min), with lower stability in rat (t = 131.7 ± 9.0 min) and mouse hepatocytes (t = 92.3 ± 4.4 min) (Supplementary Table 8). Metabolite analysis identified only a reduction product (mouse) and glucuronide conjugate (mouse and rat), with no evidence oxidative deamination or N-acetylation across all three species (Extended Data Fig. 9 and Supplementary Table 9), suggesting a reduced risk of toxic metabolite formation associated with FF-NH.

To further investigate the drug-like properties of FF-NH, we determined it's in vivo pharmacokinetic properties across different doses and routes of administration. The results indicate the absolute bioavailability of FF-NH was relatively high, ranging between 44% and 66% for oral dosing, and reached 86% for subcutaneous administration, with a linear increase in dose-dependent exposure for C and higher than linear increase for the area under the concentration-time profile curve (AUC; Supplementary Table 10). For all routes of administration, FF-NH exhibited limited duration of exposure after administration with terminal half-lives ranging between 0.28 and 1.02 h. The rapid clearance of FF-NH aligns with the natural metabolic pathway of FF, where conversion to FF-NH facilitates elimination in many animal species.

The promising mechanistic results and lack of toxicity associated with FF-NH motivated us to conduct proof-of-concept testing using an in vivo granulocyte macrophage colony stimulating factor (GM-CSF) knock out murine model of acute M. abscessus infection. For optimal exposure, FF-NH was administered subcutaneously at a dosage of 400 mg kg twice daily, which was well tolerated. SOC antibiotics, CLR and LZD, were also included as reference. Following nine consecutive days of therapy, FF-NH treatment led to a statistically significant reduction in M. abscessus bacterial burden in the lung (1.3 log c.f.u. reduction), liver (1.6 log c.f.u. reduction) and spleen (1.2 log c.f.u. reduction) compared with vehicle control (Fig. 5f), achieving reductions comparable to those of CLR and LZD. This demonstration of in vivo efficacy highlights the translational potential of leveraging intrinsic resistance to develop safe and effective therapies against this pathogen.