Azithromycin was purchased from AK-scientific (Union City, CA, USA). Synthesis of analogues is described previously (Yan et al., 2017) with Figure 1A and Supplementary Tables 1–4 providing further details of chemical structure and the origin of each analogue. Drug stocks of azithromycin (100 mM) (AK Scientific) and all analogues (10 mM) were made in ethanol. Drugs were added such that the vehicle was diluted >1000-fold for intracellular growth assays to minimise non-specific inhibition by the vehicle.

Green fluorescent protein (GFP) expressing P. falciparum (D10-PfPHG) (Wilson et al., 2010) and P. knowlesi (PkYHI) parasites (Lim et al., 2013) were cultured in human O⁺ red blood cells (RBCs) (Australian Red Cross Blood Service) in RPMI-HEPES culture medium (Thermo Fisher Scientific) supplemented with 0.5% v/v Albumax (Gibco), 52 μM gentamycin (Gibco), 367 μM hypoxanthine (Sigma-Aldrich), and 2 mM sodium bicarbonate (Thermo Fisher Scientific), adjusted to a pH of 7.2-7.4. Cultures were grown in sealed containers with 1% O₂, 5% CO₂ and 94% N₂ (BOC Gases) at 37°C (Trager and Jensen, 1976). Synchronization of D10-PfPHG parasites for growth inhibition assays was achieved using heparin sodium (Pfizer) as previously described (Boyle et al., 2010; Wilson et al., 2013). PkYH1 parasites were passaged over a gradient of 70% Percoll (Sigma-Aldrich) to purify late stage schizonts which were then allowed to rupture for ~4 hrs in the presence of fresh RBCs prior to ring-stage treatment with 5% w/v sorbitol (Sigma-Aldrich), enabling effective synchronisation of 0-4 hrs old rings.

Growth assay protocols for measuring drug inhibition of one cycle, ring to schizont stages (approximately 44 hrs post-invasion for P. falciparum and 28 hrs post-invasion for P. knowlesi) ( Figure 1B ), and two cycle (120 hrs post-invasion for P. falciparum and 78 hrs post-invasion for P. knowlesi, 2 cycles of replication) ( Figure 1C ) have been described previously (Wilson et al., 2013; Wilson et al., 2015; Lyth et al., 2018). For the assessment of stage specificity for azithromycin or analogues during the blood stage lifecycle of P. falciparum, each respective drug was removed at the specified time point (0-6 hrs or 0-12 hrs). To remove drug, cultures underwent three consecutive washes with 200 μl medium (centrifuged at 300 x g for 2 mins) with the final resuspension in 100 μl of fresh medium. Parasite growth at late trophozoite/schizont stages (44-48 hrs post invasion for P. falciparum; 24-30 hrs post invasion for P. knowlesi) was quantified using flow cytometry of parasites stained with ethidium bromide (EtBr) (10 μg/mL for 1 hr) prior to washing with PBS.

Apicoplast null (D10-PfPHG^{apicoplast-null}) parasites were generated as previously described (Yeh and DeRisi, 2011; Uddin et al., 2018). Briefly, the culture medium was supplemented with 200 μM isopentenyl pyrophosphate (IPP) (NuChem Therapeutics, Canada) and 0.35 μM (5x IC₅₀) of azithromycin for a minimum of 6 days (~three cycles) and parasites were cultured continuously thereafter with IPP. Successful removal of the apicoplast was assessed by growing D10-PfPHG^wildtype and D10-PfPHG^{apicoplast-null} parasites with reducing concentrations of azithromycin for ~120 hrs (delayed death), which confirmed a loss of sensitivity to azithromycin as observed by a ~64 fold-change in the IC₅₀ with apicoplast removal (D10-PfPHG^{apicoplast-null} IC₅₀, 4.5 μM; D10-PfPHG^wildtype IC₅₀, 0.07 μM) ( Supplementary Figure 1 ). To assess the inhibitory activity of analogues, D10-PfPHG^{apicoplast-null} and D10-PfPHG^wildtype parasites were grown in the presence of the IC₉₀ obtained for D10-PfPHG^wildtype parasites in in cycle (0-44 hrs) or delayed death (0-120 hrs) assays, or a dilution series of the respective drug. Drugs were added to tightly synchronised ring stage D10-PfPHG^{apicoplast-null} (+ 200 μM IPP) or D10-PfPHG^wildtype (no IPP) parasite lines and assays were incubated for ~44 hrs (in-cycle) or 120 hrs (two cycle delayed death) as specified with the resulting parasitemia quantitated by flow cytometry.

Parasitaemia was measured on an LSR Fortessa (Becton Dickinson) using a 96-well plate reader. Mature (>36 hrs post-invasion) P. falciparum D10-PfPHG parasites were counted using Fl-1-high (GFP; excitation wavelength, 488 nm) and Fl-2-high (EtBr; excitation wavelength, 488 nm) (Wilson et al., 2013). Mature parasites of the PkYH1 line (>24 hrs post-invasion) were gated with a forward scatter (FSC) and FL-2-high (EtBr) gate. Typically, 20,000-40,000 RBCs were counted in each well. All samples were analysed using FlowJo software (TreeStar Inc, Ashland, OR, USA) and growth of drug treatments were normalised against growth of media control wells to calculate the percent survival of drug treated parasites. To address the phenotypic effects of drugs, thin smears were fixed with fresh methanol and stained in fresh 10% Giemsa (Merck) for 10 mins before images of drug treated parasites were taken with an Olympus BX51/BX52 light microscope with immersion oil using 100X magnification.

Growth assays for drug inhibition of Toxoplasma gondii were conducted using RH-strain parasites expressing yellow fluorescent protein (Gubbels et al., 2003). Freshly-lysed parasites were used to infect HFF monolayers at 2x10³ parasites/well in 96-well plates, in FluoroBrite DMEM culture medium (Invitrogen) supplemented with 10% FCS. Drugs were added and plates were incubated at 37°C in 5% CO₂. At completion of the incubation periods (72 hrs and 120 hrs), fluorescence read-outs were obtained using the BMG LabTech PHERAstar FS microplate reader set at an excitation of 485 nm and emission of 520 nm.

Toxicity against mammalian cells was determined using the Huh-7D cell line derived from human hepatocellular carcinoma cells (Sigma-Aldrich). Huh-7D cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS) and non-essential amino acids. Huh-7D cells were grown in an atmosphere of 5% CO₂ in a 37°C incubator. Cultures were seeded to 40,000 cells in round bottom 96-well microtiter plates (Corning) and incubated with two-fold serial dilutions of drug for 24 hrs in 5% CO₂ at 37°C. Post incubation 1:1 addition of CellTiter-Glo Reagent (Promega, USA) was added to each well to lyse HuH-7D cells and release ATP for detection by luminescence (Posimo et al., 2014). Plates were incubated for 10 mins to allow the luminescent signal to stabilise, which was then detected using a Phera Star FS using the luminescent module (Lum Plus, spectral wavelength 230 nM to 750 nM). Cell viability with drug treatment was assessed by comparing cell replication in drug treated wells and normalizing this growth against non-inhibitory control wells (media).

All IC₅₀, IC₉₀ and cytotoxicity concentration (CC₅₀) estimates were determined using GraphPad Prism (GraphPad Software) according to the recommended protocol for nonlinear regression (constrained to top= 100 and bottom= 0) of a log-(inhibitor)-versus-response curve. Statistical significance between drug treatments were determined with the GraphPad Prism software using the log-(inhibitor)-versus-response curve with Extra Sum-of-Squares F Test (best-fit LogIC₅₀). P values were considered significant if P=<0.05.

Minimum inhibitory concentration (MIC) assays for assessment of Streptococcus pneumoniae sensitivity to azithromycin and analogues were performed as described (Wiegand et al., 2008). Briefly, the antibacterial activity of azithromycin and all analogues were assessed with a two-fold serial dilution in the presence of macrolide sensitive S. pneumoniae D39. Cells were inoculated at a final concentration of approximately 10⁶ CFU/mL in Mueller Hinton Broth supplemented with 5% lysed horse blood. The MIC was determined to be the concentration of drug that inhibited bacterial growth within a 96-well microtiter tray after 24 hrs incubation at 37°C. Drug activity was assessed by determining the minimal inhibitory concentration (MIC) that stopped bacterial growth, as indicated by a media colour change. MICs are expressed as µM.

For metabolomics experiments, two 150 mL flasks at 6% haematocrit containing tightly synchronised parasites 28-34 hrs post-invasion (5-6 hrs rupture window), were harvested via magnet purification (Miltenyi Biotech). Infected RBC density was quantitated by flow cytometry (Tham et al., 2010) and 2 mL of 3x 10⁷ parasites were added into the wells of 24 well microtiter plates. Parasites were incubated for 1 hrs at 37°C to stabilise the culture. Following this initial incubation, 5x IC₅₀ of the azithromycin analogue C1, the control drugs chloroquine, dihydroartemisinin (DHA) and azithromycin, and the vehicle control ethanol were added and incubated for 2 hrs. Supernatant was removed and parasites washed twice with 800 μL ice-cold 1 x PBS, with cells pelleted via centrifugation at 400 x gs for 5 mins at 0°C. Cell pellets were resuspended in 200 μL of ice-cold extraction buffer (CHCl₃/MeOH/water (1:3:1 v/v)) containing 1 µM internal standards, CHAPS and PIPES, and then incubated on ice for 1 hrs with shaking at 200 rpm. Cell debris was pelleted with centrifugation at 14800 x gs for 10 mins at 0°C. The resulting supernatant (180 µL) was transferred to Eppendorf tubes and the remaining ~20 µL were combined to make a pooled QC sample. Extraction blank samples (without cells) were prepared alongside and samples were stored at -80°C until analysis.

Liquid chromatography-mass spectrometry (LC-MS) data was acquired on a Q-Exactive Orbitrap mass spectrometer (Thermo Scientific) coupled with high-performance liquid chromatography system (HPLC, Dionex Ultimate^® 3000 RS, Thermo Scientific) as previously described (Creek et al., 2016). Briefly, chromatographic separation was performed on ZIC-pHILIC column equipped with a guard (5 µm, 4.6 × 150 mm, SeQuant^®, Merck). The mobile phase (A) was 20 mM ammonium carbonate (Sigma Aldrich), (B) acetonitrile (Burdick and Jackson) and needle wash solution was 50% isopropanol. The column flow rate was maintained at 0.3 mL/min with temperature at 25°C and the gradient program was as follows: 80% B decreasing to 50% B over 15 min, then to 5% B at 18 min until 21 min, increasing to 80% B at 24 min until 32 min. Total run time was 32 min with an injection volume of 10 µL. The mass spectrometer operated in full scan mode with positive and negative polarity switching at 35 k resolution at 200 m/z with detection range of 85 to 1275 m/z, AGC target was 1e⁶ ions, maximum injection time 50 ms. Electro-spray ionization source (HESI) was set to 4.0 kV voltage for positive and negative mode, sheath gas was set to 50, aux gas to 20 and sweep gas to 2 arbitrary units, capillary temperature 300°C, probe heater temperature 120°C. The samples were analyzed as a single batch to avoid batch-to-batch variation and randomized to account for LCMS system drift over time.

The acquired LCMS data was processed in untargeted fashion using the open source software, IDEOM (Creek et al., 2012) (http://mzmatch.sourceforge.net/ideom.php). Initially, ProteoWizard was used to convert raw LC-MS files to mzXML format and XCMS (Centwave) to pick peaks. Mzmatch.R was used to convert to peakML files, align samples and filter peaks using minimum detectable intensity of 100,000, relative standard deviation (RSD) of <0.5 (reproducibility), and peak shape (codadw) of >0.8. Mzmatch was also used to retrieve missing peaks and annotation of related peaks. Default IDEOM parameters were used to eliminate unwanted noise and artefact peaks. Loss or gain of a proton was corrected in negative and positive ESI mode, respectively, followed by putative identification of metabolites by accurate mass within 3 ppm mass error searching against the Kyoto Encyclopedia of Genes and Genomes (KEGG), MetaCyc, HMDB and LIPIDMAPS databases and a theoretical database of short peptides. To reduce the number of false positive identifications, retention time error was calculated for each putatively identified metabolite using IDEOM’s built-in retention time model which is based on actual retention time data of authentic standards (~350 standards). Statistical analysis on filtered data was performed using the Matboanalyst web interface (Chong et al., 2018).

All 22 azithromycin analogues (structures available in Supplementary Tables 1–4 ) were initially assessed for growth inhibition at 10 μM with in vitro in cycle assays ( Figure 1B ), from early rings to early schizonts, against P. falciparum (D10-PfPHG, 44 hrs) (Wilson et al., 2010) or P. knowlesi (PkYH1, 28 hrs) (Lim et al., 2013) [in cycle ( Supplementary Tables 1–4 )]. This primary screen identified 17 of 22 analogues for P. falciparum (growth inhibition of between 98% to 69% for these 17 compounds) and 18 analogues for P. knowlesi (growth inhibition of between 99% to 76% for these 18 compounds) that inhibited growth by >40% under these conditions. The 17 analogues inhibitory against P. falciparum also inhibited P. knowlesi and one analogue (A4) that was active in P. knowlesi but not P. falciparum were all prioritised for further evaluation.

The in cycle IC₅₀ values for the 17 and 18 analogues identified from primary screens were then determined against D10-PfPHG and PkYH1 lines, respectively. All analogues showed improved quick-killing IC₅₀ values compared to azithromycin (azithromycin in cycle IC₅₀- P. falciparum 11 μM; P. knowlesi 13 μM). There was a 1.6 to 30-fold improvement in quick-killing activity against P. falciparum with four analogues showing >10-fold greater potency than azithromycin (IC₅₀; A13, 0.72 μM; B2 0.56 μM; C1, 0.33 μM; and D1, 0.67 μM) ( Table 1 and Supplementary Table 5 ). There was a 1.7 to 20-fold improvement in quick-killing potency against P. knowlesi (PkYH1) with five analogues demonstrating >10-fold activity compared to azithromycin (IC₅₀; A2, 0.67 μM; A3, 1.11 μM; A8, 0.81 μM; A13, 0.64 μM; B2, 0.69 μM) ( Table 1 and Supplementary Table 5 ).

Recently, we showed that azithromycin and analogues are active against early ring stage development (<12 hrs of treatment) as well as broadly inhibitory throughout the blood stage lifecycle with ~12 hrs treatment intervals (Burns et al., 2020). Such broad activity is of interest for clinical treatment as most antimalarials exhibit limited efficacy against early rings (Zhang et al., 1986; Cowman et al., 1988; Wilson et al., 2013). To assess whether this held true for the analogues tested in this study, early ring stage D10-PfPHG parasites (0-4 hrs post invasion) were treated for 6 hrs and 12 hrs with the analogues showing the highest in cycle potency against P. falciparum (A13, B2, C1 and D1). All four analogues exhibited activity against both early (0-6 hrs) and late (0-12 hrs) ring stage treatments with a higher potency seen for longer 12 hrs treatments (A13 (IC₅₀; 0-6 hrs, 2.3 μM; 0-12 hrs, 1.4 μM; 0-44 hrs, 0.72 μM), B2 (IC₅₀; 0-6 hrs, 2.8 μM; 0-12 hrs, 1.4 μM; 0-44 hrs, 0.56 μM), C1 (IC₅₀; 0-6 hrs, 1.5 μM; 0-12 hrs, 0.7 μM; 0-44 hrs, 0.33 μM) and D1 (IC₅₀; 0-6 hrs, 2.1 μM; 0-12 hrs, 1.5 μM; 0-44 hrs, 0.67 μM)) ( Figure 2A ). For analogues A13, B2, C1 and D1 the IC₅₀ of 12 hrs treatment differed <3-fold in comparison to the IC₅₀ of 44 hrs treatment. Shorter 6 hrs ring stage treatments exhibited between a 3.1 and 5-fold rise in IC₅₀ over 44 hrs treatments, however, the activity of these analogues with a 6 hrs treatment was still significantly better than that recorded for azithromycin (IC₅₀; 0-6 hrs, 30 μM; 0-12 hrs, 16 μM) (Burns et al., 2020).

We next examined the effect of 6 hrs and 12 hrs ring stage drug treatments on parasite morphology at a 2 x IC₉₀ concentration (0-44 hrs) for the most potent analogues A13, B1, C1 or D1 using light microscopy ( Figure 2B ). No aberrant growth phenotype was obvious for 6 hrs treatments of early ring stage parasites with any drug. Examination of 12 hrs treatments showed evidence of underdeveloped parasites for all four azithromycin analogues, indicating parasite stress in the face of drug pressure. These data demonstrate that the most potent analogues have activity against early ring stage parasite growth.

Previously, we showed that azithromycin and analogues have moderate invasion inhibitory activity against Toxoplasma gondii (Wilson et al., 2015). We tested whether our most potent analogues also displayed evidence of improved quick-killing activity against T. gondii. Compounds A13 (IC₅₀ 2.3 μM), B2 (IC₅₀ 1.4 μM), C1 (IC₅₀ 3 μM) were 3, 2.5 and 10-fold less potent with 72 hrs T. gondii treatment than for 44 hrs P. falciparum treatment. Further, they all exhibited lower potency than azithromycin (IC₅₀ 0.5 μM) with 72 hrs treatment, indicating that the analogues have minimal quick-killing activity against T. gondii.

Azithromycin and analogues previously found to have quick killing activity (Burns et al., 2020) have regularly been reported to have cytotoxicity levels in vitro (CC₅₀) of >10 μM, and often show no toxicity to mammalian cells at >50 μM (Bukvic Krajacic et al., 2011; Peric et al., 2012; Peric et al., 2021). Therefore, we next investigated the potential of mammalian cell cytotoxicity for a focused group of analogues on the human hepatocellular carcinoma, Huh-7D, cell line (McKim, 2010). Inhibition of Huh-7D cell growth for analogues featuring an IC₅₀ of <1 μM in either D10-PfPHG or PkYH1 parasite lines (A2, A8, B2 and C1) was assessed using an ATP-based luminesce detection assay (Posimo et al., 2014). Two analogues, A13 and D1, with IC₅₀ values of <1 μM were excluded from this analysis due to limited sample. Compounds A2, A8 and B2 had cytotoxicity levels in vitro (CC₅₀) of >8.5 μM with >5-fold and >10-fold higher selectivity index in comparison to the IC₅₀s observed for D10-PfPHG or PkYH1, respectively ( Table 2 ). The selectivity index of C1 (CC₅₀, 2.6 μM) was 7.9 for P. falciparum (IC₅₀, 0.3 μM) but dropped to 1.6 for P. knowlesi.

We next used an untargeted metabolomics approach to identify changes in the metabolomic signatures of late trophozoites stages treated for 2 hrs at 5x IC₅₀ (44 hr) with the most potent analogue in P. falciparum, namely C1. Few metabolic pathways were impacted by C1, with the major impact being upregulation of a range of peptides, consistent with the profile observed for chloroquine and azithromycin ( Supplementary Table 6 ). Interestingly, most of these peptides are not derived from haemoglobin (the primary source of peptides in the acidic digestive vacuole of the parasite), and the source of these peptides cannot be determined due to their short sequence length. Nevertheless, this metabolic signature of elevated non-haemoglobin-derived peptides has been reported before for 4-aminoquinolines such as chloroquine and closely related analogues, suggesting that this profile is a biomarker of chloroquine-like activity in the parasite (Creek et al., 2016; Birrell et al., 2020). C1 also upregulated some unique peptides that were not impacted by azithromycin and chloroquine ( Supplementary Table 7 ), which may be derived from haemoglobin. Overall, the metabolomics profile supports our earlier proposition that azithromycin and many analogues have broad activity against trophozoite stage parasites, including the food vacuole (Burns et al., 2020).

After identifying azithromycin analogues with improved quick-killing activities against the blood stages of malaria, we next assessed apicoplast-targeting delayed death activity for each analogue. Given that the delayed death IC₅₀ of azithromycin is 0.02 μM for P. falciparum and 0.09 μM for P. knowlesi parasites, we reasoned that the delayed death activity of analogues would most likely be evident at drug concentrations below 1 μM across 2-cycles of parasite growth. All 22 analogues were screened for potential delayed death activity by treating P. falciparum (120 hrs treatment) and P. knowlesi (78 hrs treatment) parasites with 1 μM of drug and quantifying growth inhibition after two rupture cycles. This screen identified 10 of 22 analogues that inhibited growth by >30% at 1 μM across 2-cycles in both Plasmodium spp., with two analogues, A10 and A11, only active against P. knowlesi ( Supplementary Tables 1 - 4 ). We then evaluated the IC₅₀s for these prioritised drugs across 2-cycles of parasite growth in the respective Plasmodium spp. The majority of analogues featured high-nanomolar activity against P. falciparum across 2-cycles of parasite growth and were between 6 to 140-fold less potent (IC₅₀ range; C4, 2.8 μM to C1 and D1, 0.12 μM) than azithromycin (0.02 μM) ( Table 1 ; Supplementary Table 8 ). Similarly, the analogues tested against P. knowlesi showed 1.1 to 9.9-fold higher IC₅₀s (IC₅₀ range; A11, 0.89 μM to B2, 0.09 μM) than azithromycin (0.09 μM) ( Table 1 ; Supplementary Table 8 ). Notably, the analogues showing a 2-cycle IC₅₀ similar to azithromycin were also the most potent in 44 hr treatments, opening up the possibility that the improved activity over 120 hrs treatments was due to the cumulative activity of quick-killing across two growth cycles.

To address whether the improved activity over 2-cycles of parasite growth was due to quick-killing activity or apicoplast targeted delayed-death activity, the 2-cycle growth inhibitory activity for a panel of analogues was tested against parasites lacking the apicoplast (apicoplast-null) (Yeh and DeRisi, 2011; Uddin et al., 2018). First, the apicoplast was chemically removed and parasite growth rescued with IPP supplementation (Uddin et al., 2018). We next examined whether the quick-killing (in cycle, 44 hrs) activity of analogues was affected by removal of the apicoplast by treating D10-PfPHG^{apicoplast-null} and D10-PfPHG^wildtype lines with the in cycle D10-PfPHG^wildtype IC₉₀ concentration of all quick-killing analogues ( Table 1 ). As growth inhibition for D10-PfPHG^{apicoplast-null} parasites was comparable to D10-PfPHG^wildtype parasites for all drugs tested, this supported our previous observation that the quick-killing mechanism of the analogues is unrelated to apicoplast ribosome inhibition Burns et al. (2020) ( Figure 3A ; Supplementary Table 9 ).

To determine whether analogues could also kill parasites through targeting the bacterium-like ribosome of the apicoplast, we compared growth of D10-PfPHG^wildtype and D10-PfPHG^{apicoplast-null} lines across 2-cycle assays with the delayed death (120 hrs treatment) D10-PfPHG^wildtype IC₉₀ concentration of azithromycin and analogues ( Table 1 ). D10-PfPHG^wildtype parasite growth was abolished when treated with azithromycin, whereas D10-PfPHG^{apicoplast-null} parasites grew normally, consistent with the ‘rescue’ phenotype expected with an apicoplast ribosome targeting drug (Yeh and DeRisi, 2011; Uddin et al., 2018) ( Figure 3B ). However, no rescue was observed for the D10-PfPHG^{apicoplast-null} line with any analogue, results that are consistent with the activity of the non-apicoplast targeting control drug, chloroquine (targets haem-detoxification) ( Figure 3B ; Supplementary Table 9 ). To further confirm this observation, we determined the IC₅₀s for analogues that exhibited micromolar quick-killing activities but nanomolar delayed death IC₅₀ values (A5, A7, A8, A9, A11 and B2), a pattern similar to that of azithromycin, against D10-PfPHG^wildtype and D10-PfPHG^{apicoplast-null} parasites ( Table 1 ). The IC₅₀s of these six analogues against the D10-PfPHG^wildtype and D10-PfPHG^{apicoplast-null} lines were almost identical ( Figure 4 ), confirming that apicoplast targeting delayed-death activity does not contribute to the activity of these compounds.

The azithromycin analogues examined in this study featured a range of modifications that could contribute to improved quick-killing potency, however, we identified that these analogues also lost delayed-death activity. While the activity of these analogues have been addressed previously using a number of macrolide resistant Streptococcus pneumoniae strains, these studies did not test the analogues activity against an erythromycin sensitive S. pneumoniae (Yan et al., 2017), thus precluding direct comparison to our earlier study that assessed quick-killing analogue activity against an erythromycin sensitive strain of S. pneumoniae (Burns et al., 2020). Therefore, to allow clearer comparison of whether analogues with quick-killing antimalarial activity had lost or reduced activity against a bacterial ribosome we tested the potency of all 22 analogues against the gram-positive, azithromycin sensitive, bacteria S. pneumoniae (D39) ( Supplementary Table 9 ). However, none of the analogues showed growth inhibitory activity against S. pneumoniae that was equivalent to azithromycin (MIC 0.125 μM) ( Supplementary Table 9 ). As the intact desosamine sugar is required for binding to the bacterial-like ribosome (Schlunzen et al., 2001; Schlunzen et al., 2003), it was no surprise that Group C and D compounds ( Supplementary Tables 3, 4 ), featuring a desosamine modification, showed limited evidence of targeting bacterium-like ribosomes in this bacterial strain (MICs >4 μM). Group A, featured a cyclic carbonate in the 11, 12 positions and aryl or alkyl carbamoyl substitution in the 3-position of descladinosylazithromycin, and B analogues, featuring aryl or alkyl carbamoyl substitution of both the 3- and the 11- position of descladinosylazithromycin, also showed minimal growth inhibition after treating S. pneumoniae with MICs >1 μM. The only exception in this series was B1 that featured high-nanomolar activity (MIC, 0.25 μM), suggesting this analogue may retain some activity against bacterial ribosomes or has non-specific activity against the bacteria since it was not active against the ribosome of the apicoplast ( Figure 3B ; Supplementary Table 9 ).