Comprehensive stability analysis of 13 β-lactams and β-lactamase inhibitors in in vitro media, and novel supplement dosing strategy to mitigate thermal drug degradation

ABSTRACT

Non-clinical antibiotic development relies on in vitro susceptibility and infection model studies. Validating the achievement of the targeted drug concentrations is essential to avoid under-estimation of drug effects and over-estimation of resistance emergence. While certain β-lactams (e.g., imipenem) and β-lactamase inhibitors (BLIs; clavulanic acid) are believed to be relatively unstable, limited tangible data on their stability in commonly used in vitro media are known. We aimed to determine the thermal stability of 10 β-lactams and 3 BLIs via LC-MS/MS in cation-adjusted Mueller Hinton broth at 25 and 36°C as well as agar at 4 and 37°C, and in water at −20, 4, and 25°C. Supplement dosing algorithms were developed to achieve broth concentrations close to their target over 24 h. During incubation in broth (pH 7.25)/agar, degradation half-lives were 16.9/21.8 h for imipenem, 20.7/31.6 h for biapenem, 29.0 h for clavulanic acid (studied in broth only), 23.1/71.6 h for cefsulodin, 40.6/57.9 h for doripenem, 46.5/64.6 h for meropenem, 50.8/97.7 h for cefepime, 61.5/99.5 h for piperacillin, and >120 h for all other compounds. Broth stability decreased at higher pH. All drugs were ≥90% stable for 72 h in agar at 4°C. Degradation half-lives in water at 25°C were >200 h for all drugs except imipenem (14.7 h, at 1,000 mg/L) and doripenem (59.5 h). One imipenem supplement dose allowed concentrations to stay within ±31% of their target concentration. This study provides comprehensive stability data on β-lactams and BLIs in relevant in vitro media using LC-MS/MS. Future studies are warranted applying these data to antimicrobial susceptibility testing and assessing the impact of β-lactamase-related degradation.

INTRODUCTION

Efficiently identifying and developing viable therapeutic candidates against multi-drug resistant bacteria is critical to combating the antimicrobial resistance crisis (1–4). Preclinical drug development of antibiotics and dosage regimen optimization strategies commonly involve minimum inhibitory concentration (MIC) testing, static and dynamic in vitro assays, and murine infection models (5). Reliable translation of in vitro antimicrobial activity to in vivo efficacy can only be achieved by attaining targeted drug concentrations and by accounting for potential drug degradation. For dynamic in vitro and murine infection models, LC-MS/MS bioanalysis is often deployed to validate the achieved antibiotic concentration profiles (5–10). However, for static in vitro time-kill (SCTK) experiments, drug concentrations are rarely quantified (11, 12). Instead, limited degradation during a traditional 24 h study is assumed. If present, extensive thermal drug degradation would lead to under-estimating bacterial killing and potentially result in bacterial survival with regrowth. The latter may be misinterpreted as the emergence of resistance, especially if resistance of re-growing bacteria is not confirmed by viable counts on antibiotic-containing agar with subsequent MIC testing (5).

β-Lactams and older β-lactamase inhibitors (BLI), such as clavulanic acid, are generally thought to be among the least stable antibiotics (13, 14) due to thermal instability of the β-lactam ring. Degradation may be more extensive at higher pH due to increased nucleophilic attack on the carbon of the carbonyl group in the β-lactam ring by water (15). The stability of several β-lactams has been assessed at high concentrations using in vitro infusion solutions at 25 and 37°C (15–22). The drug solutions in these studies used optimized conditions (e.g., pH) for each β-lactam, which may not accurately reflect the stability of β-lactams in broth and agar media traditionally employed in antimicrobial in vitro studies and in clinical MIC testing.

A few prior studies have quantified the degradation of meropenem, ceftazidime, and aztreonam in Mueller-Hinton broth by HPLC-UV or LC-MS/MS (11, 12, 23). Traub and Leonhard (24) assessed MIC shifts (using 2-fold dilution series) to study the heat inactivation of 62 antimicrobials during exposure at 56°C for 30 min or at 121°C for 15 min (i.e., at autoclave conditions). At these high temperatures, some β-lactams degrade extensively (≥16-fold MIC shift), whereas other β-lactams remain relatively stable. However, the study conditions of this very extensive data set (24) are less relevant for SCTK traditionally performed at physiological temperatures of 35–37°C over 24 h. Moreover, quantitative concentration data (e.g., from LC-MS/MS) are needed to develop supplement dosing algorithms that can offset thermal degradation, as proposed previously (25). To the best of our knowledge, such comprehensive stability data that compare multiple β-lactams within the same study in relevant in vitro media are lacking.

The primary aim of this study was to determine the stability of 10 β-lactams and 3 BLIs in water, cation-adjusted Mueller-Hinton broth (CA-MHB), and cation-adjusted Mueller-Hinton agar (CA-MHA) at various temperatures relevant for in vitro drug screening studies and clinical MIC testing. Clavulanic acid stability (14th drug) was additionally assessed in CA-MHB only. Second, we sought to explore the stability of β-lactams at varying pH in vitro which can be observed at different sites of infection in vivo (e.g., acidic pH in setting of pulmonary empyema and alkaline pH in setting of urinary tract infection/pyelonephritis associated with staghorn calculi) (26–28). Our third aim was to develop a new dosing algorithm to offset thermal degradation via supplement doses. This approach allows one to achieve drug concentrations in CA-MHB close to the target concentration in SCTK over 24 h and to thereby minimize the impact of thermal degradation. Moreover, the generated stability data will help to inform optimal storage conditions and incubation durations of antibiotic-containing agar plates, and the stability of dosing solutions to support dynamic in vitro infection models (5).

MATERIALS AND METHODS

Chemicals and reagents

Imipenem monohydrate, meropenem, and clavulanic acid were purchased from AK Scientific (Union City, CA). Biapenem, doripenem monohydrate, and avibactam sodium salt were obtained from MedChem Express (Monmouth Junction, NJ). Ceftazidime pentahydrate, cefepime dihydrochloride monohydrate, cefsulodin sodium salt hydrate, piperacillin sodium salt, ticarcillin disodium salt, aztreonam, and tazobactam sodium salt were purchased from Chem-Impex International (Wood Dale, IL) and sulbactam from TCI America (Portland, OR). The CA-MHB and CA-MHA were obtained from Becton, Dickinson and Company (Sparks, MD). Hydrochloric acid and LC/MS grade water and methanol were purchased from Fisher Scientific (Fair Lawn, NJ), sodium hydroxide from VWR International (Radnor, PA), and formic acid from Sigma-Aldrich (St. Louis, MO). As internal standards for LC-MS, diclofenac and diphenhydramine were obtained from Sigma-Aldrich (St. Louis, MO).

Broth stability

The pH of CA-MHB was adjusted by sodium hydroxide and hydrochloric acid to an average ± SD of 6.80 ± 0.0045, 7.00 ± 0.010, 7.25 ± 0.037 (i.e., unadjusted pH), 7.40 ± 0.0050, and 7.80 ± 0.0055. Stability was determined at five pH for all drugs excluding cefepime and cefsulodin which were tested at three pH conditions (6.80, 7.25, 7.80). The pH adjusted CA-MHB was then sterilized through a 0.22-µm filter (Merck Millipore, Tullagreen, Carrigtwohill, Ireland) and pre-warmed in a shaking water bath at 36°C. The β-lactams and BLIs were dissolved in water, sterile-filtered, and then added to the pre-warmed CA-MHB to reach a final concentration of 8 mg/L. The latter falls in the clinically relevant range against Gram-negative bacteria for many β-lactams (29). At least seven serial samples over 24–26 h were obtained to quantify drug concentration profiles. Methanol with the internal standards (diclofenac or diphenhydramine, Table 1) was added to the collected samples for protein precipitation, followed by centrifugation at 17,968g for 5 min.

TABLE 1.

Classifications and mass spectrometric conditions of tested compounds on the Agilent 6460 and Sciex 6500+ instruments^a,b

Compound name	Drug class	Abbreviation	Mass transition (m/z)	LLOQ (mg/L)	Instruments	Internal standard
Imipenem	Carbapenem	IPM	300.1 → 98.0300.1 → 126.1	0.05 0.30	Sciex 6500+, Agilent 6460	DIC
Biapenem		BPM	351.1 → 110.4	0.25, 0.30	Sciex, Agilent	DIC
Meropenem		MEM	384.2 → 141.1	0.05, 0.30	Sciex, Agilent	DIC
Doripenem		DOR	421.1 → 274.1	0.05, 0.30	Sciex, Agilent	DIC
Ceftazidime	Cephalosporin	CAZ	547.1 → 468.0	0.05, 0.30	Sciex, Agilent	DIC
Cefepime		FEP	481.1 → 396.0	0.05, 0.30	Sciex, Agilent	DIC
Cefsulodin		CFS	533.0 → 123.0	0.05, 0.30	Sciex, Agilent	DIC
Piperacillin	Penicillin	PIP	518.5 → 143.1	0.05, 0.30	Sciex, Agilent	DIC
Ticarcillin		TIC	385.0 → 160.1	0.025, 0.30	Sciex, Agilent	DIC
Aztreonam	Monobactam	ATM	436.1 → 313.1	0.05, 0.30	Sciex, Agilent	DIC
Tazobactam	β-Lactamase inhibitor	TZB	301.0 → 168.1 301.2 → 122.1	0.025 0.30	Sciex, Agilent	DPH
Sulbactam		SUL	232.0 → 124.1 232.0 → 140.0	0.25 0.30	Sciex, Agilent	DPH/DIC
Avibactam		AVI	264.0 → 96.8	0.25, 0.30	Sciex, Agilent	DPH/DIC
Clavulanic acid		CLA	198.3 → 107.9	0.25	Agilent 6460	DPH
Diclofenac	Internal standard	DIC	296.0 → 215.1
Diphenpydramine		DPH	256.2 → 167.2

^a The lower limit of quantification (LLOQ) was not optimized for extremely high sensitivity, since this was not needed at the studied relatively high drug concentrations.

^b DIC, diclofenac; DPH, diphenpydramine.

Additional stability studies were performed at room temperature (i.e., 25°C) for 11 drugs at concentrations of 1 and 16 mg/L in CA-MHB at pH 7.25 (i.e., the unadjusted pH after autoclaving) in triplicate. Serial samples were obtained at 0, 5, 24, 48, 72, and 96 h. These studies were meant to evaluate the possibility of storing drug-containing broth for dosing in a dynamic in vitro infection model at room temperature (i.e., avoiding refrigeration) if a continuous infusion dosage regimen is to be simulated.

Agar stability

The CA-MHA was prepared following the manufacturer’s instructions and pH adjustment was not performed. After tempering the autoclaved CA-MHA to 55°C in a water-bath, the respective β-lactam or BLI was added at concentrations of 1, 3, 10, or 30 mg/L. Agar stability studies were performed in three independent vials (i.e., triplicates) at each concentration. The agar was placed into a 2-mL homogenization tube and cooled at 4°C in a refrigerator for 10 min to solidify. Subsequently, agar was either continuously stored at 4°C or incubated at 37°C.

Serial samples were obtained at 0, 2, 5, 8, 24, 48, and 72 h to mirror typical durations of incubating antibiotic-containing agar for viable counting of resistant bacteria. For each sample, water (200 µL) and methanol (200 µL) were added, followed by breaking the agar texture using vigorous shaking with lysing beads via a Cryolys Evolution Precellys Evolution homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France). The homogenized samples were centrifuged at 17,968g for 5 min. Then, 400 µL of methanol (containing the internal standards, Table 1) was added to the supernatant for subsequent LC-MS/MS analysis.

Water stability

The stability of β-lactams and BLIs in purified water (LC-MS/MS grade, at neutral pH) was studied at −20, 4, and 25°C and drug concentrations of 10 and 1,000 mg/L using 6 replicates (i.e., 3 replicates each on different days) for all drugs except ticarcillin, sulbactam and avibactam which were studied in triplicate. Serial samples were obtained at 5, 24, 48, 72, and 96 or 120 h from each vial. Methanol (400 µL) containing the internal standards (Table 1) was added for LC-MS/MS analysis.

LC-MS/MS analysis

Bioanalyses for stability in CA-MHB were performed on an Agilent 6460 LC-MS/MS. Stability in CA-MHA and water was assessed on a Sciex 6500+ LC MS/MS. All LC-MS/MS analyses employed an internal standard (Table 1). For studies in CA-MHB, chromatographic separation was achieved using an Agilent 1260 series LC system (Santa Clara, CA) with an analytical column (Luna Omega Polar C18, 150 × 2.1 mm, 5 µm; Phenomenex, Torrance, CA). A gradient mobile phase method was employed, with the aqueous phase (solvent A) containing 0.1% formic acid in water and organic phase (solvent B) containing methanol. The total flow rate was 0.3 mL/min, and total run time was 12 min. Mass spectrometric determination of sample eluents was carried out using an Agilent 6460 triple-quadrupole system equipped with an electrospray ionization source. The mass transitions via multiple reaction monitoring (MRM; Table 1) of different analytes were monitored in positive ion mode for all drugs, except avibactam which was studied in negative mode (due to its better sensitivity in this mode). The peak areas of the respective analyte and internal standard were used for integration of chromatograms.

For stability studies in CA-MHA and water, the β-lactams and BLIs were analyzed using an Acquity I-Class UPLC system (Waters, Milford, MA) equipped with a Triple Quad 6500+ mass spectrometry (AB Sciex, Framingham, MA). We used a Kinetex Polar C18 100 × 2.1 mm, 2.6 µm column (Phenomenex, Torrance, CA) at 40°C with a run time of 4 min. The mobile phase consisted of 0.1% formic acid in water (A) and 0.1% formic acid in methanol (B) at a flow rate of 0.4 mL/min. The source and gas parameters were curtain gas (CUR), 30; collision gas (CAD), 9; ionspray voltage (IS), 5,000 V in positive mode and −4,500 V in negative mode; temperature (TEM), 450; ion source gas 1 (GS1), 25; ion source gas 2 (GS2), 25. Optimized MRM conditions are summarized in Table 1. The Analyst software (AB Sciex) was used for data analysis using peak areas of the respective analyte and internal standard.

Estimating degradation half-lives

Population pharmacokinetic (PK) modeling was used to simultaneously estimate the first-order degradation half-lives of all 13 drugs (plus clavulanic acid as 14th drug in CA-MHB only) and to assess the impact of pH in broth. Initially, we estimated the stability of all studied drugs in CA-MHA and water simultaneously. However, we identified noticeable differences in the impact of temperature and drug concentrations on the degradation half-lives between the studied drugs. Therefore, the stability data in CA-MHA and water were modeled separately for each drug. We did not merge the broth, agar, and water data sets, as the data in CA-MHB included varied pH at a constant temperature, whereas the data sets in CA-MHA and water assessed different temperatures and drug concentrations at a constant pH.

Estimations were performed via the importance sampling algorithm (pmethod = 4) in the S-ADAPT software (version 1.57) and the SADAPT-TRAN facilitator tool (30–32). Competing models were evaluated by assessing the individual and population fitted degradation profiles and via the objective function (negative log-likelihood in S-ADAPT), and other standard diagnostic plots (33, 34).

Supplement dosing algorithms

A re-dosing strategy was developed to offset the thermal degradation of β-lactams via supplement doses. Two scenarios were considered with one supplement dose at 9 h or two supplement doses at 6 and 10 h. These time-points of supplement dosing were chosen because they can be readily implemented in in vitro time-kill experiments without requiring a supplement dose at night. Re-dosing sought to achieve drug concentrations close to the target concentration and minimize the impact of thermal degradation over time. The mean deviation (Dev) of the drug concentration relative to the targeted concentration (C₀) over 24 h was calculated as:

$${\displaystyle \text{Dev}=\frac{\text{1}}{\text{24}\text{h}}\cdot {\int }_{\text{0h}}^{\text{24h}}\frac{|C_{\text{t}}-C_{\text{0}}|}{C_{\text{0}}}\cdot \text{dt}}$$

The C_t denotes the actual drug concentration which changes over time. Based on the estimated degradation half-lives, these simulations and calculation of deviations were coded in the Berkeley Madonna software (version 8.3.18, University of California, Oakland, CA). The model code is provided in the supplementary materials and can be readily adapted for drugs with other degradation half-lives and supplement dosing times.

For the single-dose supplement dosing scheme at 9 h, a series of supplement dosing fractions (i.e., 0%–60% of the original dose, with a 1% step size) was simulated. For the two-supplement dose scheme, a range of supplement dose fractions (i.e., 0%–55%, with a 1% step size) was simulated for the first supplement dose [A] at 6 h. The second supplement dose [B] at 10 h used supplement dose fractions of 0%–60% (with a 1% step size). The combination of the two supplement doses that provided the smallest mean deviation from the target concentration was obtained from the combinations of supplement doses.

RESULTS

Bioanalytical assays

The LC-MS/MS assay performance data are provided in the supplementary materials. In CA-MHA, the median (5th–95th percentile) accuracy of the studied drugs (i.e., the ratio of the measured divided by the nominal concentration) was 98.8% (93.1%–103.1%) at different concentrations and across several days, and precision was 2.8% (0.4%–9.4%) (Table S1). In CA-MHB, these statistics were 100.2% (91.5%–106.3%) for accuracy, and 7.0% (2.6%–12.6%) for precision (Table S2) on the Sciex 6500+ LC MS/MS. On the Agilent 6460 LC-MS/MS, accuracy was 99.8% (91.8%–104.4%) and precision 1.8% (<1%–16.1%) (Table S3). In water, accuracy was 100.9% (91.7%–112.3%) and precision 6.7% (2.3%–10.1%) across the studied drugs, different concentrations, and assay days (Table S4). These assay performance results were suitable for the present study.

Stability in CA-MHB at different pH

The average ± SD of pH in CA-MHB after autoclaving without pH adjustment was 7.25 ± 0.0370. For all drugs, the degradation half-lives were shorter at higher pH (Table 2). When compared to pH 7.25, population modeling estimated the degradation half-lives to be on average 1.51-fold longer (SE: 4.7%) at pH 6.80 and 1.24-fold longer (7.7%) at pH 7.00. In contrast, the average degradation half-lives were 1.14-fold shorter (7.8%) at pH 7.40 and 1.42-fold shorter (3.4%) at pH 7.80 compared to those at pH 7.25.

TABLE 2.

Population PK estimated degradation half-lives (i.e., posthoc estimates) of β-lactams and β-lactamase inhibitors at different pH values in CA-MHB at 36°C^a,b,c

Compound name	Degradation half-life (h) at:
	pH 6.80	pH 7.00	pH 7.25	pH 7.40	pH 7.80
Imipenem	23.7	19.8	16.9	15.1	10.3
Biapenem	33.0	25.6	20.7	16.4	10.0
Doripenem	59.7	49.7	40.6	34.7	23.3
Meropenem	62.4	53.4	46.5	38.9	29.2
Cefsulodin	36.8	NS	23.1	NS	14.9
Cefepime	80.9	NS	50.8	NS	28.1
Ceftazidime	266	217	176	148	102
Piperacillin	95.8	81.3	61.5	53.3	32.9
Ticarcillin	282	234	188	165	109
Aztreonam	1,824	1,488	1,190	1,025	685
Clavulanic acid	45.0	38.1	29.0	25.7	19.3
Tazobactam	211	172	138	121	78.4
Sulbactam	568	460	374	321	210
Avibactam	673	552	448	386	256

^a Notably, some of the estimated degradation half-lives substantially exceeded the 24 h study duration.

^b The population mean half-life (relative standard error) at pH 7.25 was 16.8 h (8.4%) for imipenem, 20.0 h (10.5%) for biapenem, 40.3 h (6.2%) for doripenem, 45.4 h (6.3%) for meropenem, 24.1 h (14.4%) for cefsulodin, 51.0 h (7.2%) for cefepime, 175 h (11.9%) for ceftazidime, 61.8 h (7.1%) for piperacillin, 189 h (15.8%) for ticarcillin, 1,190 h (8.4%) for aztreonam, 30.3 h (9.3%) for clavulanic acid, 139 h (45.0%) for tazobactam, 372 h (14.1%) for sulbactam, and 446 h (14.0%) for avibactam.

^c NS, not studied at this pH.

Among the carbapenems, degradation in broth at 36°C was fastest for imipenem, followed by biapenem, and then meropenem and doripenem (Table 2; Fig. 1). Cefsulodin was the least stable cephalosporin, whereas cefepime and ceftazidime were fairly stable. Clavulanic acid was substantially less stable than tazobactam and sulbactam. Penicillins were fairly stable, and aztreonam was very stable (Table 2). The very long degradation half-lives for some drugs in broth were reported to compare stability results; however, it should be noted that the long half-lives were considerably longer than the 24 h observation period in CA-MHB.

Fig 1.

Degradation of tested β-lactams and β-lactamase inhibitors in CA-MHB at 36°C and different (adjusted) pH values.

Supplement dosing algorithms

Based on the estimated degradation half-lives at pH 7.25 (i.e., the unadjusted pH), imipenem degraded by 63% within 24 h in CA-MHB (i.e., from 100% to 37%; Fig. 2), if no supplement dose was used. Meropenem degraded by 30%, clavulanic acid by 44%, cefepime by 28%, and piperacillin by 24% (Fig. 2). Due to this noticeable degradation, we developed a supplement dosing algorithm for any drug that degraded more than 10% after 24 h of incubation in CA-MHB. When one supplement dose was allowed at 9 h, the optimal supplement was 61% of the original dose for imipenem, 51% for biapenem, 46% for cefsulodin, and smaller for the other drugs (Table 3). For example, at a targeted imipenem concentration of 10 mg/L, this supplement dosing algorithm would propose adding the appropriate amount of imipenem at 9 h to increase the imipenem concentration by 6.1 mg/L (i.e., 61% of 10 mg/L). This amount needs to account for the potentially smaller broth volume at 9 h compared to that at 0 h since viable count (and other) samples may have been taken before 9 h.

Fig 2.

Simulated drug concentration time profiles without supplementation, as well as with one or two supplement doses using the optimized re-dosing fractions compared to the targeted drug concentration (baseline). The simulation was based on drug degradation half-life in CAMHB at pH 7.25. The numbers at the right represent the fractions of the initial drug concentration remaining at 24 h if no supplement dose was used. Results for additional drugs are shown in Fig. S1.

TABLE 3.

Optimal dosing schemes with one supplement dose at 9 h

		With one supplement dose				No supplement dose
Compound	Degradation half-life at pH 7.25 in CAMHB (h)	Re-dosing fraction (%)a	Average deviation (%)	Lower bound (%)b	Upper bound (%)c	Average deviation (%)	Lower bound (%)b,d
Imipenem	16.9	61	15.6	69.1	130	36.4	37.4
Biapenem	20.7	51	12.9	74.0	125	31.3	44.8
Doripenem	40.6	27	6.73	85.7	113	18.0	66.4
Meropenem	46.5	24	5.89	87.4	111	15.9	69.9
Cefsulodin	23.1	46	11.6	76.3	122	28.7	48.7
Cefepime	50.8	22	5.40	88.4	110	14.7	72.1
Clavulanic acid	29.0	37	9.33	80.6	117	23.9	56.3
Tazobactam	138	6	1.48	96.7	103	5.79	88.6
Piperacillin	61.5	18	4.48	90.4	108	12.4	76.3

^a Redosing fraction at 9 h that minimizes the mean deviation.

^b The lowest predicted drug concentration relative to the targeted concentration during a 24 h period.

^c The highest predicted drug concentration relative to the targeted concentration during a 24 h period.

^d The upper bound for the no supplement dose case was 100% for all drugs.

After this supplement dose, the drug concentrations exceeded the targeted concentration by up to 30% for imipenem (Fig. 2; Fig. S1) and to a lesser degree for the other drugs (see upper bound in Table 3). The average deviation was 15.6% for imipenem, 12.9% for biapenem, 11.6% for cefsulodin, and less than 10% for more stable drugs. As expected, the supplement doses for the latter drugs were smaller and drug concentrations closer to the targeted value (referred to as 100% of the target in Fig. 2; Fig. S1). Compared to the reference scenario without a supplement dose (Table 3), giving one supplement dose allowed one to achieve the targeted concentration ~2 to 3 times more precisely (based on the average deviation); and the trough concentration at 24 h was closer to the targeted concentration by up to 33% (for imipenem, i.e., 37% remaining at 24 h for no supplement dose vs 70% for the supplement dosing; Fig. 2; Fig. S1).

As expected, when using two supplement doses, the average deviations were smaller with 11.8% for imipenem, 9.72% for biapenem, 8.70% for cefsulodin, and less than 7% for all other drugs (Table 4). Imipenem had optimal supplement doses of 30% at 6 h and 36% at 10 h. For an imipenem target concentration of 10 mg/L, appropriate amounts of imipenem would need to be added to increase the imipenem concentration by 3.0 mg/L at 6 h and by 3.6 mg/L at 10 h (while accounting for the broth volume at these times). Thus, when the supplement doses were split into two smaller doses, the sum of the supplement doses (here 66%) was slightly larger than the single supplement dose option (61% for imipenem, Table 3). The two-supplement dose scheme was most beneficial for the less stable drugs (i.e., imipenem, biapenem, and cefsulodin).

TABLE 4.

Optimal dosing schemes with two supplement doses^e,f

Compound	Redosing A (%)a	Redosing B (%)b	Average deviation (%)	Lower bound (%)c	Upper bound (%)d
Imipenem	30	36	11.8	72.0	127
Biapenem	23	31	9.72	76.8	122
Doripenem	12	16	5.03	87.8	111
Meropenem	11	14	4.12	90.4	110
Cefsulodin	22	27	8.70	79.2	120
Cefepime	9	14	4.08	90.7	110
Clavulanic acid	17	22	6.98	83.2	116
Tazobactam	4	5	1.58	96.3	104
Piperacillin	8	11	3.33	92.2	108

^a Redosing fraction at 6 h that minimizes the mean deviation.

^b Redosing fraction at 10 h that minimizes the mean deviation.

^c The lowest predicted drug concentration percentage during a 24 h study.

^d The highest predicted drug concentration percentage during a 24 h study.

^e Please see Table 3 for degradation half-lives used during simulations.

^f The reference results for the no supplement case were the same as those shown in Table 3.

Broth stability at room temperature

The degradation half-lives of the β-lactams and BLIs at 25°C were approximately 3 times (median) longer than those at 36°C in CA-MHB at pH 7.25 (Table 5). Imipenem was the least stable drug with a half-life of 48.0 h followed by biapenem (95.7 h), doripenem (119 h), and cefsulodin (148 h). All other drugs had degradation half-lives longer than 160 h in CA-MHB at pH 7.25 and 25°C.

TABLE 5.

Thermal degradation half-lives of ten β-lactam antibiotics and three β-lactamase inhibitors in different matrices relevant for in vitro infection model experiments and susceptibility testing at different temperatures and pH^c,d

Drug	CAMHB (8 mg/L)		CAMHB (1 & 16 mg/L)	CAMHA (1, 3, 10, or 30 mg/L)		Water (10 mg/L)			Water (1,000 mg/L)
	36°C, pH 6.80	36°C, pH 7.25	25°C, pH 7.25	at 37°C	at 4°C	at 25°C	at 4°C	at −20°C	at 25°C	at 4°C	at −20°C
Imipenem	23.7	16.9	48.0	21.8	≥90% at 72 ha	73.7	83.0b	260	14.7	69.7	206
Biapenem	33.0	20.7	95.7	31.6	≥90% at 72 h	>500	>500	>500	215	>500	>500
Doripenem	59.7	40.6	119	57.9	≥90% at 72 h	91.1	151	171	59.5	112	113
Meropenem	62.4	46.5	167	64.6	≥90% at 72 h	>500	>500	>500	297	>500	>500
Cefsulodin	36.8	23.1	148	71.6	≥90% at 72 h	>500	>500	>500	354	>500	>500
Cefepime	80.9	50.8	275	97.7	≥90% at 72 h	>500	>500	>500	>500	>500	>500
Ceftazidime	266	176	512	166	≥90% at 72 h	>500	>500	>500	363	>500	>500
Piperacillin	95.8	61.5	529	99.5	≥90% at 72 h	370	>500	354	258	>500	321
Ticarcillin	282	188	n.s.	>200	≥90% at 72 h	>500	>500	>500	402	>500	>500
Aztreonam	1,824	1,190	1,400	>200	≥90% at 72 h	>500	>500	>500	>500	>500	>500
Tazobactam	211	138	298	>200	≥90% at 72 h	>500	>500	>500	>500	>500	395
Sulbactam	568	374	599	>200	≥90% at 72 h	>500	>500	>500	431	>500	>500
Avibactam	673	448	n.s.	>200	≥90% at 72 h	>500	>500	>500	431	>500	>500

^a The estimated degradation half-life in agar at 4°C was greater than 200 h. Therefore, we listed the average fraction of drug remaining at 72 h as a more practical surrogate measure of stability for refrigerated storage of drug-containing agar plates.

^b When using data from all six replicates, this estimate was 582 h. We reported the estimate from the three less stable replicates to be conservative.

^c Stability in water showed a faster degradation at a high stock or dosing solution concentration of 1,000 mg/L compared to a more clinically relevant concentration of 10 mg/L.

^d n.s., not studied.

Agar stability

Over the studied range of drug concentrations (i.e., 1, 3, 10, and 30 mg/L), noticeable differences in the drug degradation half-lives in agar were not found. When averaged over the three replicates and the four concentrations (i.e., 12 replicates in total), stability in CA-MHA at 4°C was >90% at 72 h for all studied drugs (Tables 5 and 6), and the estimated degradation half-lives were over 200 h.

TABLE 6.

Fraction of drug remaining in agar incubated at 37 and 4°C for different durations^a

	At 37°C			At 4°C
Drug	24 h	48 h	72 h	24 h	48 h	72 h
Imipenem	0.58 ± 0.04	0.26 ± 0.07	0.07 ± 0.005	1.01 ± 0.09	0.94 ± 0.26	0.96 ± 0.09
Biapenem	0.69 ± 0.04	0.38 ± 0.06	0.17 ± 0.01	0.94 ± 0.18	0.96 ± 0.08	0.98 ± 0.25
Doripenem	0.83 ± 0.03	0.59 ± 0.05	0.39 ± 0.01	0.99 ± 0.12	0.98 ± 0.16	1.03 ± 0.11
Meropenem	0.81 ± 0.03	0.62 ± 0.05	0.44 ± 0.01	0.98 ± 0.10	0.99 ± 0.17	1.03 ± 0.10
Piperacillin	0.87 ± 0.05	0.73 ± 0.05	0.56 ± 0.04	0.97 ± 0.10	1.05 ± 0.09	1.04 ± 0.10
Cefsulodin	0.92 ± 0.04	0.68 ± 0.07	0.46 ± 0.03	1.00 ± 0.09	1.05 ± 0.13	1.05 ± 0.12
Cefepime	0.98 ± 0.04	0.73 ± 0.04	0.61 ± 0.03	1.02 ± 0.09	0.99 ± 0.26	1.06 ± 0.10
Ceftazidime	0.97 ± 0.03	0.85 ± 0.06	0.71 ± 0.06	0.99 ± 0.11	1.04 ± 0.09	1.07 ± 0.08
Tazobactam	1.00 ± 0.04	0.92 ± 0.03	0.82 ± 0.02	1.06 ± 0.25	1.00 ± 0.12	1.08 ± 0.19
Ticarcillin	0.91 ± 0.01	0.90 ± 0.03	0.83 ± 0.02	0.96 ± 0.13	1.11 ± 0.1	1.08 ± 0.06
Avibactam	0.96 ± 0.06	0.97 ± 0.07	0.90 ± 0.03	0.99 ± 0.13	1.08 ± 0.12	1.12 ± 0.22
Aztreonam	1.05 ± 0.02	0.97 ± 0.04	0.95 ± 0.02	1.10 ± 0.29	1.08 ± 0.26	1.17 ± 0.26
Sulbactam	1.00 ± 0.04	1.00 ± 0.04	0.95 ± 0.03	1.22 ± 0.30	1.17 ± 0.23	1.31 ± 0.18

^a Data are average ± SE from n = 12 replicates (i.e., three replicates each at 1, 3, 10, and 30 mg/L). Numbers in bold represent drugs that displayed more than 10% degradation at the respective time point.

In contrast, degradation in CA-MHA at 37°C was much faster for some drugs. At this temperature, aztreonam, sulbactam, and avibactam were ≥90% stable for 72 h; ticarcillin and tazobactam for 48 h; as well as ceftazidime, cefepime, and cefsulodin for 24 h. Among the drugs displaying >10% decline by 24 h, the average extent of degradation at 24 h was 42% for imipenem, 31% for biapenem, 19% for meropenem, 17% for doripenem, and 13% for piperacillin (Table 6). The degradation half-lives in CA-MHA at 37°C were comparable to those in CA-MHB at pH 6.80 and 36°C (Table 5, with exception of cefsulodin and ceftazidime).

Aqueous stability

The stability in water was evaluated at three relevant temperatures (25, 4, and −20°C) at concentrations of 10 and 1,000 mg/L since dosing solutions often contain high drug concentrations. All but two drugs had degradation half-lives under all six studied conditions over 200 h. Only imipenem and doripenem degraded faster in water with degradation half-lives of 14.7 h for imipenem and 59.5 h for doripenem at 25°C and a drug concentration of 1,000 mg/L. At 10 mg/L and 25°C, the degradation half-life was 73.7 h for imipenem and 91.1 h for doripenem (Table 5). Other drugs also tended to show more degradation at 1,000 mg/L compared to that at 10 mg/L, but their degradation half-lives were longer than the 96 h stability study duration performed in water.

DISCUSSION

The β-lactam antibiotics and BLIs, such as clavulanic acid, are perceived to be thermally unstable (13). However, there is limited tangible data quantifying their stability using contemporary LC-MS/MS assays in matrices relevant for in vitro susceptibility testing and infection models. The present study comprehensively evaluated the stability of 13 β-lactams and BLIs in common in vitro broth and agar media, and water, under various clinically relevant conditions (e.g., physiologic temperatures, varying pH, and different drug concentrations). Additionally, clavulanic acid was studied in broth only.

Here, our results reveal carbapenems (especially imipenem and biapenem), cefsulodin, and clavulanic acid had significantly diminished stability in CA-MHB at 35 ± 2°C where their degradation half-lives ranged from 16.9 to 29.0 h at pH 7.25; this degradation may impact MIC interpretation (Table 2). Moreover, the degradation half-life at 35 ± 2°C was 21.8 h for imipenem and 31.6 h for biapenem in CA-MHA (Table 5). Meropenem, doripenem, cefepime, and piperacillin showed slower degradation half-lives of 40.6–61.5 h during incubation in CA-MHB at pH 7.25, with longer half-lives of 57.9–99.5 h in CA-MHA. All other drugs were more stable (Tables 2 and 5).

For less stable drugs, we proposed a new feasible and flexible supplement dosing algorithm allowing investigators to achieve more constant drug concentrations in SCTK studies (Fig. 2; Fig. S1); this may be important to assessing bacteriostatic versus bactericidal activity by an antibiotic. The supplement doses are intended to offset thermal degradation during 24 h of incubation, but to not overshoot the targeted concentration by more than 30% (ideally less) due to supplement dosing (Fig. 2). As expected, larger supplement doses were required for drugs with faster degradation half-lives such as imipenem (Tables 3 and 4). In contrast, more stable drugs (degradation half-life ≥139 h) such as ceftazidime, ticarcillin, aztreonam, tazobactam, sulbactam, and avibactam did not require supplement dosing.

When compared to the relatively rapid thermal degradation half-lives of β-lactams during incubation in broth or agar, stability was markedly improved under refrigeration at 4°C in CA-MHA, with at least 90% stability for 72 h for all studied drugs (Table 5). Under these conditions, the degradation half-lives were not affected by the drug concentrations (i.e., 1–30 mg/L). It still seems advisable though to prepare and use antibiotic-containing agar within 24–48 h to minimize the impact of any drug degradation (Fig. S2) and to consider stability in CA-MHA during incubation when evaluating bacterial growth on agar.

Interestingly, all but two drugs had degradation half-lives >200 h in water at 10 and 1,000 mg/L. Only imipenem and doripenem displayed faster degradation half-lives, and their degradation was noticeably faster at 1,000 mg/L (relevant for dosing solutions) than at 10 mg/L (Table 5), suggesting the possibility for self-mediated β-lactam hydrolysis at higher concentrations. Most drugs were stable at both 10 and 1,000 mg/L over 96 h in water only showing slight concentration-dependent degradation.

Furthermore, drug stability improved at lower temperatures (Table 5) in water and at mildly acidic pH (i.e., 6.8) compared to that at pH 7.25 in broth (Table 2). We observed shorter degradation half-lives for all drugs at higher pH and longer half-lives at lower pH (Fig. 1, Table 2). This suggested that acidic and alkaline pH should be considered due to its impact on drug stability in in vivo infection models and patient clinical trials since pH is known to vary based on the site of infection (with acidic pH in pneumonia and a range of different pH in urinary tract infections) (35–38), bacteria and bacterial density (26–28). Other considerations reflective of the in vivo environment apart from temperature and pH, and their impact on drug stability, which were not evaluated in our in vitro investigation, include sodium chloride composition, ionic strength, and protein content present in serum or plasma. Some caution should be applied when extrapolating our stability results to other matrices and conditions.

The proposed supplement dosing algorithms minimize the effect of drug degradation during each 24 h period, and thereby offer a path to extend the duration of SCTK studies beyond 24 h. To precisely re-establish the targeted drug concentrations, centrifugation and re-suspending all bacteria in fresh, antibiotic-containing broth with the targeted drug concentrations every 24 h has been used to extend SCTK studies over several days (25, 39–44). This further removes bacterial waste products and provides fresh nutrients and oxygen. While SCTK studies are efficient to assess the time-course of bacterial growth and killing, achieving the targeted drug concentrations is important to distinguish between regrowth due to drug degradation or emergence of resistance (5).

Drug stability becomes even more important when evaluating slow growing intracellular Mycobacteria such as Mycobacterium avium complex and Mycobacterium tuberculosis complex where drug MIC values are interpreted following 7–21 days of incubation (45, 46). Several reports highlight concerns regarding the impact of drug stability on MIC testing and SCTK results (13, 47, 48). While the present study was not designed to address drug stability for long-term incubation and MIC testing of Mycobacteria, our stability results and supplement dosing algorithms may support research on Mycobacterium abscessus complex where β-lactams and BLIs are frequently used in therapy and MICs are assessed following 48–72 h of incubation (49).

The present study evaluated the thermal stability of 13 compounds in-depth under clinically relevant conditions and is an essential first step needed to accurately determine in vitro susceptibility and SCTK studies with these drugs. However, we did not assess the impact of bacteria and did not study β-lactamase-related drug degradation. Thus, β-lactamase activity is not part of the present supplement dosing algorithms (Tables 3 and 4). Our results are applicable to scenarios where β-lactamase-related drug degradation is either absent (due to a lack of bacteria) or negligible. This applies to dosing solutions and to bacterial strains that lack a β-lactamase or that have minimal β-lactamase activity toward the studied drugs (Fig. 3). Moreover, our stability results support SCTK studies at low and normal initial inocula (e.g., ≤10⁶ CFU/mL, depending on strain and pathogen; Fig. 3).

Fig 3.

Overview of different scenarios where the proposed supplement dosing algorithm to offset thermal degradation is fully applicable, helpful and potentially applicable, or should be supported by quantifying extracellular drug concentrations in broth (e.g., via LC-MS/MS). This scheme is meant to provide general guidance since various special cases likely exist.

A potential exception to these studies is strains with inducible β-lactamases, such as the AmpC β-lactamase in P. aeruginosa, when studying β-lactams that inactivate penicillin-binding protein 4 (e.g., carbapenems) (50–52). If extensive bacterial growth occurs, some strains may degrade the studied β-lactam after reaching a sufficiently high bacterial density (e.g., ≥10⁸ CFU/mL). However, in this case, the reason for several log₁₀ of growth or regrowth is more likely ineffective drug treatment, as opposed to extensive β-lactamase activity at low initial bacterial densities which causes a decline of the drug concentration in broth.

Regarding the situation that bacterial strains produce β-lactamase(s) that inactivate the studied β-lactams, we have encountered different scenarios (Fig. 3). In the first, β-lactamase activity is confined to the periplasmic space of Gram-negative bacteria and the studied strain has a low outer membrane permeability toward the studied β-lactams (53). This applies to wild-type P. aeruginosa PAO1 with ceftazidime and aztreonam, where thermal degradation was the main contributor to drug degradation even at high initial inocula (54–56).

The second scenario applies to highly permeable Enterobacterales (e.g., Escherichia coli and Klebsiella pneumoniae) with extensive β-lactamase activity (57–59), and to rapidly penetrating β-lactams, such as imipenem in P. aeruginosa (60) (Fig. 3). When studying a high inoculum (10^7.2 CFU/mL), even wild-type P. aeruginosa PAO1 caused an extensive decline of imipenem concentrations in broth (i.e., from 2 to 0.16 mg/L or from 4 to 2 mg/L, both within 8 h). This is despite >1 log₁₀ initial killing by 2 mg/L imipenem (0.5× MIC) and >2 log₁₀ killing by 4 mg/L (1× MIC) (52). Thus, drug degradation substantially contributed to regrowth at 2 mg/L imipenem. For highly permeable Gram-negatives such as E. coli and K. pneumoniae, β-lactamase producing strains can yield very rapid and extensive degradation of β-lactams in broth (58, 61). In this scenario, our current supplement dosing algorithm might be helpful for low inoculum studies but does not capture the β-lactamase activity at high inocula due to extensive β-lactamase activity that likely affects drug concentrations in broth. Thus, quantifying the achieved drug concentrations (e.g., via LC-MS/MS) is essential to accurately define drug effects in vitro.

In the third scenario, extracellular β-lactamase activity is present, as observed for Acinetobacter baumannii, Staphylococcus aureus, Bacillus, and other pathogens and may involve extracellular vesicles (called outer membrane vesicles in Gram-negatives) (53, 62–71). Moreover, dead bacteria may release β-lactamases into the medium. Here, the impact of β-lactamases is expected to be greatest since drug degradation occurs extracellularly. In this case, the current supplement dosing algorithm does not capture the likely predominant source of drug degradation due to β-lactamase activity in broth, especially at high inocula (Fig. 3). Thus, it is essential to determine drug concentrations for both low and high inoculum studies. Quantifying drug concentrations provided important insights for SCTK studies that assessed β-lactam and BLI combinations (11, 12, 72–74). Taken together, there is an intricate balance between the bacterial inoculum, type of β-lactamase, β-lactamase expression, whether the organism is Gram-positive or Gram-negative, the presence of extracellular/outer membrane vesicles, the site(s) of β-lactamase activity, Gram-negative outer membrane permeability, and the studied β-lactam and BLI concentrations.

In summary, the present study provides the first comprehensive thermal stability data on 13 β-lactams and BLIs (plus clavulanic acid in CA-MHB) under relevant conditions for in vitro susceptibility testing and infection model studies based on contemporary LC-MS/MS. We developed a feasible supplement dosing algorithm to offset thermal degradation and achieve as constant as possible drug concentrations in SCTK studies over 24 h. This approach allows one to extend SCTK studies well beyond 24 h. Imipenem, biapenem, cefsulodin, and clavulanic acid were the least stable compounds when incubated in broth or agar, or at room temperature in water, especially at high carbapenem concentrations. Future research is warranted to evaluate the thermal stability of other β-lactams and develop supplement dosing algorithms that account for β-lactamase-related drug degradation. These will likely require quantifying drug concentration-time profiles in broth and depend on the tested pathogens, strains, and other experimental variables. Overall, this work provides a foundation for confidently designing in vitro infection model studies by laboratories that do not have the associated LC-MS/MS assays readily available.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.01399-23.

Supplemnetary materials. aac.01399-23-s0001.docx.

Additional figures, bioanalysis assay details, and Berkeley Madonna simulation code.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.