Effects of Microplate Type and Broth Additives on Microdilution MIC Susceptibility Assays

The determination of antibiotic potency against bacterial strains by assessment of their minimum inhibitory concentration normally uses a standardized broth microdilution assay procedure developed more than 50 years ago. However, certain antibiotics require modified assay conditions in order to observe optimal activity.

The published reference procedures for broth dilution assays do not generally specify the type and nature of the container in which the assays should be conducted, with the CLSI M07-A11 document listing "sterile 13 ϫ 100-mm test tubes" for the macrodilution procedure and "plastic microdilution trays that have round or conical bottom wells" for the microdilution procedure (2). The one exception is a CLSI-EUCAST working group recommendation in 2016 that surfactants should not be included in the reference broth microdilution method for colistin, and that untreated polystyrene (PS) trays should be employed (7). Agar and broth dilution methods are also reported in a 2008 Nature Protocols article (8), which is one of the very few published protocols that mentions the potential for plate composition effects on MIC potency. This report showed that cationic antimicrobial peptides (AMPs) have reduced MICs in tissue culture-treated PS plates compared to those in polypropylene (PP) plates, and that the addition of acetic acid/bovine serum albumin (BSA) alleviated this effect (8). The adherence of cationic AMPs to PS (particularly tissue-culture treated PS) is also mentioned in a note in a 2007 Methods in Molecular Biology chapter, which recommended the use of PP plates (9).
Microtiter plates used in broth microdilution assays are generally made from PS, but several different types of surface modifications are available. For example, Corning offers over 10 types of surface treatments for microplates, many of them designed to specifically bind cells or biomolecules. Untreated PS is considered a medium binding surface that is hydrophobic and binds biomolecules through passive interactions. The standard tissue culture-treated (TC-treated) surface, used for the attachment and growth of anchorage-dependent cells, is created by applying a corona discharge that grafts oxygen atoms onto the surface PS chains (Fig. 2) so that the surface becomes hydrophilic and negatively charged. Other binding surfaces include a high binding surface to bind biomolecules that possess ionic groups and/or hydrophobic regions, and surfaces coated with poly-D-lysine, sulfhydryl, carbohydrate, amine, or photoreactive groups that can be used to covalently immobilize biomolecules.
Some surfaces are designed to minimize binding, including Corning's nonbinding surface (NBS), a proprietary treatment technology that creates a polyethylene oxide-like nonionic hydrophilic surface to minimize nonspecific molecular interactions (Fig. 2). The NBS surface has been compared to untreated PS and PP for the binding of radiolabeled proteins; BSA bound to PS at 450 ng/cm 2 and to PP at 440 ng/cm 2 but to the NBS-coated PS at Ͻ2.5 ng/cm 2 (78), so the NBS plate is recommended to reduce protein binding during assays.
Similarly, Thermo Scientific Nunc offers a range of treated plates, generally based on PS, with various degrees of absorption characteristics; the Nunc MiniSorp and GeNunc module surfaces have very low nonspecific binding characteristics, due to specially formulated polyethylene resin. Unfortunately, some plates have optical characteristics unsuitable for MIC determinations requiring optical density or visual readouts of turbidity.
In the course of developing third-generation semisynthetic lipoglycopeptide antibiotics designed to selectively target bacterial membranes (24), our laboratory noted that nonbinding surface (NBS) plates provided significantly improved microdilution MIC values compared to other types of plates, and we also observed significant variations caused by added protein or surfactant depending on the plate type. As a result, we initiated a systematic comparison of several different plate types for microdilution assays, comparing various antibiotics against both Gram-positive and Gram-negative strains, and with/without a number of common additives. We were particularly interested in determining if a common plate type could prevent the need for specialized assay conditions for individual lipophilic antibiotics, driven by our internal drug discovery program on synthetic lipoglycopeptide vancomycin derivatives. These "vancapticins" increase the selectivity of vancomycin toward bacterial membranes by using an attached cationic "associative" peptide sequence terminated with an "insertive" lipophilic group (Fig. 3) (24). They possess high protein binding and a propensity to adhere to plastic surfaces, similar to the second-generation lipoglycopeptides telavancin (compound 2), dalbavancin (compound 3), and oritavancin (compound 4). The ability to avoid additives would simplify assay preparation, preventing errors due to incorrect concentrations of polysorbate. It would also avoid potential unexpected effects of added surfactants, given that a nonionic polyethylene glycol surfactant (Triton X-100) similar to Tween 80 (T80) used to avoid compound aggregation during other types of screening assays has been shown to unpredictably affect assay results (25). More importantly, it would enable better standardization of conditions and comparison of antimicrobial activity profiles between laboratories if a single plate type, with no need for broth additives, could be adopted.

RESULTS
Our studies were designed to examine the effects of different plate types on broth microdilution MIC determinations. Initially, we selected seven different PS-based plate types, as follows: Corning untreated flat-bottom, Corning TC-treated flat-bottom, Corning NBS-treated flat-bottom, Nunc untreated flat-bottom, Nunc TC-treated U-bottom, Nunc TC-treated flat-bottom, and Trek Diagnostics untreated flat-bottom plates. For each plate type, one Gram-positive organism was tested against seven antibiotics and one Gram-negative organism against seven antibiotics (see Fig. S1 in the supplemental material for experimental design). Antibiotics were selected to include those previously reported to exhibit plate-or surfactant-based MIC variations, as well as examples expected to not show an effect. All assays were conducted in both Mueller-Hinton broth (MHB) and in MHB supplemented with 0.002% T80 to see if the plate type could obviate surfactant. The second set of assays (Fig. S1) examined whether differences between plates seen in the first experiments were consistent across bacterial strains. Three plate types from one manufacturer (Corning untreated, TC treated, and NBS treated) were compared using the same sets of antibiotics but with six additional Gram-positive and three additional Gram-negative strains, again using MHB medium with and without T80. The final experiments (Fig. S1) compared the effects of different additive used to assess the effectiveness of antibiotics under physiological conditions, 50% human serum or 2% lung surfactant. These were done in four plate types (Trek PS untreated or Corning PS untreated, TC treated, and NBS treated), and were compared in MHB, MHB with added 0.002% T80, and MHB with added 2% LHB. Experiment 1, initial plate comparison with or without Tween 80. The first set of experiments compared MIC values determined in seven different plate types against one Gram-positive (methicillin-resistant Staphylococcus aureus [MRSA] ATCC 43300) and one Gram-negative (Escherichia coli ATCC 25922) organism conducted in MHB with and without 0.002% T80, using two sets of the following seven antibiotics for the two different classes of bacteria: vancomycin, ciprofloxacin, telavancin, dalbavancin, MCC233, MCC310, and MCC520 for MRSA; and colistin, ciprofloxacin, oxacillin, trimethoprim, polymyxin B, penicillin, and rifampin for E. coli (note that oxacillin and penicillin G generally have poor Gram-negative activity; they were included to assess if surfactant additives or plate types had synergistic effects, via damage to the bacterial membranes, that increased their potency by providing greater access to the periplasm/peptidoglycan). Literature value ranges for these antibiotics against MRSA and E. coli are listed in Table 1, with plate results tabulated in Table  2 and visualized in Fig. 4 Given that the control antibiotics with low protein binding and nonlipophilic characteristics showed little variation, it would appear to be unlikely that the plate or Tween itself is having a synergistic effect on antibacterial activity. For the assays against MRSA, telavancin in all 3 untreated PS plates showed values around 0.06 g/ml with and without added T80, similar to levels in NBS plates, while TC plates without additive produced higher values (0.5 to 1 g/ml), which were reduced when T80 was added (0.125 g/ml). Dalbavancin and the vancapticins showed much more pronounced variations, with Ͼ4-fold improvements (and up to Ͼ1,000-fold) when T80 was added to PS plates from two of the three manufacturers; improvements  in Nunc PS plates were substantially less striking (1-to 8-fold). The Corning NBS plates without additive gave values comparable to those with PS plus T80 (e.g., Յ0.003 to 0.03 g/ml). The addition of T80 to NBS plates resulted in 2-fold or greater reductions in potency. The TC plates consistently gave MICs of Ն2 g/ml for these antibiotics, with minimal improvements upon the addition of T80. Similar variations were seen in the antibiotics tested against E. coli (Table 3 and Fig.  5). Ciprofloxacin, oxacillin, penicillin G, trimethoprim, and rifampin showed little variation across plate types, with or without added T80. In contrast, colistin (and, to a lesser extent, polymyxin B) showed the same trends as the lipoglycopeptides, as follows: PS plates gave more potent values (Յ0.03 to 1 g/ml) than TC plates (0.5 to 4 g/ml), and the addition of T80 generally improved activity in PS plates (Յ0.03 to 0.25 g/ml) but not in TC or NBS plates. The NBS plates gave values equivalent to the best PS plus T80 results (Յ0.03 g/ml). Experiment 2, plate plus Tween 80 comparison versus expanded set of bacteria. In order to examine whether the plate variations in MIC extended across multiple types of bacteria, three plate types (PS, TC, and NBS) from the same manufacturer were compared, using the same combinations of seven antibiotics against either seven  Gram-positive (Table S1 and Fig. S2) or four Gram-negative (Table S2 and Fig. S3) bacteria. The results were very similar to those observed from the first experiments across all strains tested, in that more polar antibiotics were generally unaffected by the plate type/additive, while lipoglycopeptides or lipopeptides were least potent in TCwithout T80 plates and most potent in PS with T80, with equivalent activity in NBS without T80. Experiment 3, plate plus additives comparison. In assessing the activity of potential antibiotics, it is important to conduct assays in the presence of biological components that the antibiotics will encounter in the human body, namely, human serum (inactivation by protein binding) and lung surfactant (encountered when treating pneumonia). We therefore assessed four plate types (PS, TC, and NBS from Corning, and PS from Trek) in MHB, MHB plus 50% human serum (HS), and MHB plus 2% artificial lung surfactant (LS). For this experiment, in addition to testing the activity with no additive, and with T80, we also tested in the presence of 2% lysed horse blood (LHB), which was previously reported to have the same blocking effect as surfactant (10).
The Gram-positive antibiotic panel against MRSA ATCC 43300 (Table 4 and Fig. 6) demonstrated again that vancomycin and ciprofloxacin had little variation with plate type or any combination of T80, LHB, HS, or LS additive. Dalbavancin was strikingly inactivated in the presence of 50% HS under all plate type and additive conditions, with telavancin and vancapticin activities reduced to a lesser extent. In TC plates, T80 and LHB improved the activities of telavancin and dalbavancin but had little effect on vancapticin activity. In PS plates, telavancin was generally unaffected by the additives, while dalbavancin and the vancapticins showed significant improvement (Ͼ10-fold). NBS plates gave the most potent activity for all lipoglycopeptides but with reductions in activity when T80 or LHB was added.
In the Gram-negative antibiotic panel (Table 5 and Fig. 7), ciprofloxacin, oxacillin, penicillin G, trimethoprim, and rifampin again showed little variation even with HS or LS additives. Trimethoprim was marginally more active in their absence across multiple plate types, though this difference disappeared once T80 or LHB was added. Colistin and polymyxin B were most active in NBS plates and least active in TC plates regardless of T80, LHB, HS, or LS additive. Curiously, LHB in Corning PS plates greatly reduced colistin and polymyxin B activity compared to T80 but not in Trek PS plates, and in TC plates, the effect was reversed.

DISCUSSION
We have, for the first time, systematically evaluated plate-based effects on MIC determinations during broth microdilution assays. While MIC assay guidelines cover a range of experimental parameters, the composition of the assay vessel is generally not specified, other than the CLSI-EUCAST recommendations for colistin that state that plain PS should be employed (7). The majority of antibiotic stock solutions for testing are prepared in water, phosphate buffer, or pH-adjusted aqueous solution (106 out of 139 reported in CLSI M100 [see Table 6A, "Solvents and diluents for preparation of stock solutions of antimicrobial agents"]) (3). However, several antibiotics, notably the lipophilic lipoglycopeptides telavancin (compound 2), dalbavancin (compound 3), and oritavancin (compound 4) (Fig. 1), must be solubilized in DMSO or 0.002% polysorbate 80 (Tween 80) in water (3). Furthermore, the CLSI reference MIC quality control range tabulated for dalbavancin and oritavancin are obtained in cation-adjusted Mueller-Hinton broth (CAMHB) supplemented with 0.002% polysorbate 80 (see Table 5A-1 in CLSI M100, "MIC QC ranges for nonfastidious organisms and antimcirobial agents") (3). This requirement is consistent with published reports that 0.002% (final concentration) of polysorbate 80 is required for reproducible MIC testing of dalbavancin without substrate or medium constituent interference, due to poor antibiotic solubility and facile absorption to plastic surfaces (13). If the dalbavancin dilutions were in contact with plastic for as little as 30 min before inoculation with surfactant-containing media, the measured MIC against S. aureus ATCC 29123 rose from CLSI-consistent values of 0.06 g/ml to values of 2 to 8 g/ml. Similarly, oritavancin (compound 4) MIC values were underestimated by 16-to 32-fold in the absence of added polysorbate 80, again due to the depletion of free drug onto plastic surfaces (10,11). The loss to PS microtiter plates was quantified using [ 14 C]oritavancin; at 16 g/ml, the concentration of oritavancin was approximately 70% of that expected, but at 4 g/ml, it was only 35%, and at 1 g/ml, it was Ͻ10% (10). The addition of 2% lysed horse blood (LHB) was found to have the same blocking effect as surfactant (10). A similar effect, with added protein reducing antibiotic adherence to plastic, was observed with the glycolipodepsipeptide complex ramoplanin (compound 6), where the addition of 0.02% bovine serum albumin (BSA) resulted in more potent MICs (15,16). In 2014, the CLSI methods for determining the MIC of telavancin (compound 2) were revised to include the addition of polysorbate 80, with DMSO used during stock preparation (12,14). Notably, the closely related glycopeptide vancomycin (compound 5), without a lipophilic moiety, does not require surfactant supplement. A recent report discussed the effects of solvent (DMSO, ethanol, and methanol) on bacterial growth and found 20% reductions in growth across five organisms at concentrations of Ͼ3% DMSO, Ͼ3% methanol, or 1% ethanol, so solubilizing additives may also affect assay results (26). Members of the lipopeptide polymyxin class of antibiotics (polymyxin B [compound 7a] and polymyxin E [or colistin, compound 7b]) are important last-line therapeutic agents against many multidrug-resistant Gram-negative bacteria (27)(28)(29)(30)(31). They consist of an N-terminal fatty acid side chain that is attached to a polycationic deca-peptide backbone ( Fig. 1)   peptides. These structural features confer amphipathicity, which is a key feature of many other cationic antimicrobial peptides (CAMPs) (32)(33)(34). As mentioned earlier, plate types have been reported to affect CAMP MIC values (8). In 2012, the addition of 0.002% polysorbate was reported to improve the MIC results for colistin and polymyxin B, with 4-to 8-fold more potent MICs against over 200 strains with surfactant present (19), with the results confirmed in clinical isolates (17,18,20). Greater differences were observed when the initial MIC was lower (Ͻ2 g/ml). As for oritavancin, measurement of colistin concentrations in MHB following incubation in PS, PP, and glass tubes showed substantial time-dependent depletion at lower concentrations, with only 8%, 23%, or 25% of the initial 0.125 g/ml concentration detected after 24 h, but 84%, 90%, and 80% of an expected 8 g/ml concentration detected, respectively (21). Dramatic differences were observed between different brands of untreated PS microwell plates, comparing those from Greiner (remarkably, only 2% of 8 g/ml after 24 h) and Nunc (70% of 8 g/ml after 24 h). Low-binding PP microtubes showed the least loss at low concentrations (59% of 0.125 g/ml after 24 h) (21). However, a CLSI-EUCAST working group in 2016 determined that surfactants should not be included in the reference broth microdilution method for colistin, and that untreated PS trays should be employed (7). Two new reports in 2018 described container effects on polymyxin activity. Untreated PS Sensititre GNX2F assay plates (Thermo Fisher) were compared to broth  (23). PS and glass-coated plates were compared for broth microdilution assays of colistin and polymyxin against 42 carbapenem-resistant strains of Acinetobacter baumannii (22). For both antibiotics, the PS resulted in greater variability and slightly less potent MICs (glass plate MIC for all 42 strains, 1 or 2 g/ml; PS plates had 3 isolates with MIC of 2 g/ml).
The extent of protein binding of antibiotics is often approximated by a serum reversal MIC instead of standard equilibrium dialysis or ultrafiltration methods. This technique conducts MIC assays without and with added serum proteins, either with broth containing 50 to 95% human or mouse serum, or with added 3 to 4% human serum albumin or bovine serum albumin protein (53,59,60). The ratio of retained activity indicates the extent of unbound antibiotic. However, bacteria do not grow as well in human serum as in standard medium, so high concentrations of serum may have a synergistic antimicrobial effect (59). MIC assays of lipoglycopeptides/lipopeptides that bind to both protein and plastic have the potential to be confounded by the opposing effects. High protein binding means that little free antibiotic is available for antimicrobial activity, reducing their MIC potency, but the added protein also reduces nonspecific binding to plastic, resulting in more potent MIC values.
We now demonstrate that plate type can cause large variations in MIC assay results, not only between different types of plate composition/coatings but sometimes in plates of the same polymer from different manufacturers. This supports the previously reported dramatic variation in colistin concentrations when incubated in untreated PS plates from different manufacturers (21). It is evident that to enable an "apples-toapples" comparison of data from different laboratories, the exact plate type and composition (or vessels used for macrodilution experiments) should be reported when describing MIC assays.
The extent of plate-based variations is highly dependent on the type of antibiotic and appears to correlate with hydrophobic or amphiphilic molecules. We have found that for antibiotics where the addition of Tween 80 leads to more potent activity, the use of NBS plates without additive provides similar results. This suggests that the reduced activity in PS plates is caused by loss of compound due to binding to the plate surface, with even greater loss of compound in TC plates. However, there is a disconnect between the extent of protein binding of antibiotics and their "stickiness" to plastic, based on the observed plate effects. Telavancin (90 to 93% protein binding) showed little alterations in MIC when tested in NBS versus untreated plates, while dalbavancin (93 to Ͼ95%) and colistin (60%) both resulted in large variations.
In summary, plate and additive effects are observed across a range of bacteria, but there are subtle variations depending on the antibiotic, plate type, and additives employed. All broth microdilution MIC determinations should clearly specific the plate type and manufacturer and any additives employed. These studies demonstrate that NBS plates can effectively prevent reductions in MIC due to adsorption of compound to the plate surface and allow for assays of lipophilic antibiotics without the need for added surfactant. We are conducting further investigations against a much larger panel of antibiotics to establish the extent of plate-based variations in MIC determinations. It remains to be determined which plate type provides the "true" MIC value that is most relevant to the clinical activity of the antibiotic. Telavancin (61) was synthesised from vancomycin according to procedures in the literature. MCC223, MCC310, and MCC520 were synthesized from vancomycin and purified by high-performance liquid chromatography (HPLC) to Ͼ95% purity (24). Their purity was ascertained by analytical liquid chromatography-mass spectrometry (LC-MS) and identity confirmed by high-resolution MS (HRMS) and tandem MS (MS/MS) fragmentation.

Materials
The MIC determination via broth microdilution assay. MIC determinations were done in duplicate (n ϭ 2), with vancomycin, telavancin, dalbavancin, and ciprofloxacin used as positive inhibitor controls for Gram-positive bacteria, and colistin, ciprofloxacin, oxacillin, trimethoprim, polymyxin B, penicillin G, and rifampin for Gram-negative bacteria. A positive control of a row of just the bacteria and a negative control of only the medium were included for every plate tested. The antibiotic standards were prepared to 1.28 mg/ml solution in water. MCC223, MCC310, and MCC520 were prepared to 160 g/ml solution in water from a stock solution of 1 mM concentration.
The compounds along with standard antibiotics were serially diluted 2-fold across the 96-well plates. Standards ranged from 64 g/ml to 0.03 g/ml and compounds from 8 g/ml to 0.003 g/ml, with final volumes of 50 l per well. Gram-positive and Gram-negative bacteria were cultured in Mueller-Hinton broth (catalog no. 211443; Bacto Laboratories) with and without 0.002% Tween at 37°C overnight. A sample of each culture was then diluted 40-fold in fresh MH broth (in presence and absence of Tween) and incubated at 37°C for 2 to 3 h. The resultant mid-log-phase cultures were diluted to 1 ϫ 10 6 CFU/ml under the same two conditions, and then 50 l was added to each well of the compound-containing 96-well plates to give a final cell density of 5 ϫ 10 5 CFU/ml. All the plates were covered and incubated at 37°C for 24 h. MICs were determined visually at 24 h of incubation, with the MIC defined as the lowest concentration at which no growth was visible after incubation.
For experiments in the presence of surfactant, 2% beractant (25 mg/ml) was added to the mid-log-phase cultures and mixed gently and added to all the 96-well plates. For experiments in the presence of serum, a mixture of 50% of human serum along with 50% MHB was prepared and used throughout the experiment.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AAC .01760-18.