Divergent genetic landscapes drive lower levels of AmpC induction and stable de-repression in Serratia marcescens compared to Enterobacter cloacae

ABSTRACT The chromosomally encoded AmpC beta-lactamase is widely distributed throughout the Enterobacterales. When expressed at high levels through transient induction or stable de-repression, resistance to ceftriaxone, a commonly used antibiotic, can develop. Recent clinical guidance suggests, based on limited evidence, that resistance may be less likely to develop in Serratia marcescens compared to the better-studied Enterobacter cloacae and recommends that ceftriaxone may be used if the clinical isolate tests susceptible. We sought to generate additional data relevant to this recommendation. AmpC de-repression occurs predominantly because of mutation in the ampD peptidoglycan amidohydrolase. We find that, in contrast to E. cloacae, where deletion of ampD results in high-level ceftriaxone resistance (with ceftriaxone MIC = 96 µg/mL), in S. marcescens deletion of two amidohydrolases (ampD and amiD2) is necessary for AmpC de-repression, and the resulting ceftriaxone MIC is 1 µg/mL. Two mechanisms for this difference were identified. We find both a higher relative increase in ampC transcript level in E. cloacae ΔampD compared to S. marcescens ΔampDΔamiD2, as well as higher in vivo efficiency of ceftriaxone hydrolysis by the E. cloacae AmpC enzyme compared to the S. marcescens AmpC enzyme. We also observed higher relative levels of transient AmpC induction in E. cloacae vs S. marcescens when exposed to ceftriaxone. In time-kill curves, this difference translates into the survival of E. cloacae but not S. marcescens at clinically relevant ceftriaxone concentrations. In summary, our findings can explain the decreased propensity for on-treatment ceftriaxone resistance development in S. marcescens, thereby supporting recently issued clinical guidance.

Nevertheless, when AmpC is expressed at high levels, hydrolysis of ceftriaxone or cefotaxime can result in resistance (8).This high-level expression can be triggered by two mechanisms: transient induction or stable mutational de-repression.Transient induction occurs upon exposure to a beta-lactam antibiotic; the resulting derangement in cell wall remodeling causes a buildup in peptidoglycan fragments that bind to the AmpR transcriptional regulator, leading to increased ampC transcription (9)(10)(11).When cell wall homeostasis is restored, AmpC returns to basal levels.In contrast, stable de-repression of AmpC most often results from mutations in the peptidoglycan recycling pathway itself, especially in the peptidoglycan amidohydrolase AmpD.By this mechanism, AmpC remains expressed at high levels, and MIC values are often elevated, even in the absence of antibiotic (10,12).
There is a concern that treating AmpC-producing Enterobacterales infections with ceftriaxone or cefotaxime, even if initial MIC testing indicates susceptibility, can lead to treatment failure due to either transient induction, mutational derepression, or both.On-treatment emergence of resistance has been observed in clinical series most frequently in E. cloacae and less so for S. marcescens (13)(14)(15).Recent analyses of in vitro mutation rates reveal considerably lower rates of the development of resistance to ceftriaxone in S. marcescens isolates compared to E. cloacae isolates (16).Based on these findings, there is now guidance that S. marcescens infections may be treated with ceftriaxone if the isolate tests susceptible, in contrast to infections with E. cloacae which should not (17).
However, the mechanisms responsible for the lower tendency for S. marcescens to develop ceftriaxone resistance remain poorly characterized.Possible explanations include variations in the amount of independent mutational events necessary for the emergence of ceftriaxone resistance, differences in the degree of AmpC overexpression required to produce resistance, or differential enzyme induction by ceftriaxone.Here, we investigate these potential mechanisms.

RESULTS
We investigated differences in AmpC biology in the well-characterized type strains E. cloacae ATCC 13047 and S. marcescens ATCC 13880.We began by constructing compa rable markerless deletion mutants and determining MICs by Etest for ceftriaxone and related beta-lactams as well as the unrelated agent tobramycin.Wild-type E. cloacae had a higher baseline ceftriaxone MIC than wild-type S. marcescens (Table 1).Expression of AmpC appears to be responsible for much of this difference since deletion of ampC in E. cloacae and in S. marcescens resulted in similar MICs.It has been previously shown using broth microdilution that in S. marcescens, like in the closely related Yersinia enterocolitica (18), deletion of both ampD and its orthologous peptidoglycan amidohydrolase amiD2 is necessary for substantial elevation in ceftriaxone MIC.We confirm this result here using Etest gradient diffusion.Clinical & Laboratory Standards Institute (CLSI) breakpoints for "Intermediate" and "Resistant" correspond to a ceftriaxone MIC of 2 and 4 µg/mL, respectively.In wild-type S. marcescens ATCC 13880, deletion of ampD or amiD2 alone results in small MIC elevations; deletion of both is necessary for an MIC of 1 µg/mL (representing a 16-fold increase from wild-type) (Table 1).It has been shown that OmpF is the S. marcescens porin most important for ceftriaxone permeation (19).In addition to ampD and amiD2, deletion of ompF is required for ceftriaxone resistance (with MIC = 8 µg/mL).In contrast, deletion of ampD alone in E. cloacae ATCC 13047 is sufficient for high-level ceftriaxone resistance (MIC = 96 µg/mL, representing a 128-fold increase from wild-type).Modest increases in cefepime and imipenem MICs were observed in S. marcescens and in cefepime MICs in E. cloacae mutants; however, all remained in the susceptible range.No substantial differences in tobramycin MICs were observed in either organism (Table 1).
We hypothesized that one contribution to these AmpC-dependent differences in ceftriaxone MIC could be the greater efficiency of the E. cloacae AmpC in ceftriaxone hydrolysis.The k cat./K m (a measure of the catalytic efficiency for a given substrate) for the closely related cephalosporin, cefotaxime, is known to be about 10-fold higher in purified AmpC enzyme from E. cloacae compared to that from S. marcescens (7).To test this hypothesis in vivo, we performed scarless replacement of the S. marcescens ampC protein coding sequence with that of E. cloacae, and vice versa, preserving native upstream regulatory elements.We predicted that S. marcescens strains with the E. cloacae ampC would show increased MICs, and E. cloacae with the S. marcescens ampC would show decreased MICs.However, only a small increase in ceftriaxone MIC in S. marcescens with the E. cloacae ampC was observed (Table 2).Ceftriaxone MICs increased, as expected, in S. marcescens with ampD and amiD2 deletion, but MICs were only slightly higher in those with the E. cloacae ampC.However, E. cloacae with S. marcescens ampC exhibited pronounced decreases in ceftriaxone MICs.This difference was particularly notable in E. cloacae ΔampD; the strain with the S. marcescens ampC had a 16-fold lower ceftriaxone MIC than the strain with the native E. cloacae ampC (Table 2).These findings suggest that the higher hydrolytic efficiency of the E. cloacae AmpC for ceftriaxone contributed at least in part to the higher ceftriaxone MICs we observed.
Relative levels of de-repression of ampC expression could also contribute to AmpCdependent differences in ceftriaxone MICs.To test this hypothesis, we performed reverse transcription relative quantitative real-time PCR (relative RT-qPCR) to compare wild-type ampC expression to those in de-repressed deletion mutants.As a negative control, we found, as expected, that ampC transcripts had very low abundance in the deletion mutants (Table 3).Similar to our MIC data which showed only small MIC increases in single S. marcescens amidohydrolase mutants, relative transcript levels were similar or only slightly increased compared to the wild-type in these strains.The S. marces cens double mutant had ~40 fold increased relative levels.In contrast, the E. cloacae single ampD deletion mutant had a marked ~500 fold increased relative level of ampC transcripts.Thus, E. cloacae's greater relative ampC expression upon de-repression, as well as its higher intrinsic AmpC enzyme ceftriaxone hydrolytic efficiency likely both contribute to its increased resistance to ceftriaxone.Recently, Kohlmann et.al. determined species-specific mutation rates leading to cefotaxime resistance in clinical isolates with baseline susceptibility to cefotaxime (16).They observed 100-fold lower mutation rates in S. marcescens isolates compared to E. cloacae isolates.We predicted that one explanation for this finding is the greater number of mutations necessary for AmpC de-repression.Consistent with that hypothesis, we found that, despite a large inoculum of wild-type S. marcescens, we were unable to detect any ceftriaxone mutants after plating on 1 µg/mL ceftriaxone agar (Table 4).As expected, the mutant frequency was higher in wild-type E. cloacae for ceftriaxone concentrations of 1, 2, and 4 µg/mL.Deletion of both ampD and amiD2 in S. marcescens was necessary for the recovery of a similar number of ceftriaxone mutants as in wild-type E. cloacae.Deletion of S. marcescens ompC and/or ompF alone had little influence on mutant frequency.Similar to their influence on MIC (Table 1), only amidohydrolase deletions in combination with porin deletions led to relatively high mutant frequencies.To ensure that mutant frequencies did not appear low in S. marcescens because of fitness defects in deletion strains, we compared their growth in rich and in minimal media; we  b did not detect differences in growth rates in either single or compound S. marcescens deletion mutants (data not shown).
The genetic selection and outgrowth of ceftriaxone mutants with stable de-repres sion of ampC may be facilitated in vivo by transient induction of ampC upon ceftriaxone exposure.In this way, transient increases in AmpC levels could support initial survival in ceftriaxone, supporting a sufficient inoculum for the appearance of de-repressed mutants.Though ceftriaxone and cefotaxime have historically been considered to be relatively poor inducers of ampC transcription (6), it has recently been shown that E. cloacae does undergo ampC induction at high concentrations of cefotaxime (that may be encountered in vivo) (20).We predicted that ceftriaxone might have similar potential for induction and sought to compare ampC induction in S. marcescens and E. cloacae.We exposed either wild-type E. cloacae or S. marcescens to therapeutic concentrations of ceftriaxone or cefotaxime (21)(22)(23) for 1 hour and determined the increase in relative levels of ampC transcription.We observed larger increases in relative ampC expression in E. cloacae compared to S. marcescens (Fig. 1A).At 10 µg/mL of ceftriaxone, ampC was induced at 10-fold higher levels in E. cloacae compared to in S. marcescens, and at 100 µg/mL of ceftriaxone, 40-fold higher levels were observed in E. cloacae compared to in S. marcescens.This difference was even more pronounced when ampC induction was normalized to baseline ceftriaxone MIC (Fig. 1B).Results were similar for cefotaxime (Fig. 1C and 1D ) and for ceftriaxone induction over 15 minutes as opposed to 1 hour (data not shown).Thus, ceftriaxone and cefotaxime stimulate greater induction of ampC expression in E. cloacae vs S. marcescens.
We hypothesized that these large differences in ceftriaxone-triggered induction of ampC expression would allow wild-type E. cloacae to survive better than wild-type S. marcescens when exposed to ceftriaxone.We performed ceftriaxone time-kill curves with two different inocula to test this idea.A 10 5 CFU/mL inoculum, typical of a urinary tract infection and higher than most blood stream infection (24,25), and a 10 8 CFU/mL inoculum that replicates the high local concentration of bacteria that might occur on a prosthetic device or in a heart valve vegetation (26) were tested.We used experimen tal ceftriaxone concentrations of 10 µg/mL and 100 µg/mL that are well above the MICs of both organisms, and would approximate achievable in vivo trough and peak concentrations, respectively (23).We found that at the 10 5 CFU/mL inoculum, wild-type S. marcescens was killed at both ceftriaxone concentrations (Fig. 2A).In contrast, E. cloacae, after undergoing an initial decline in CFU at 6 hours after exposure to ceftriaxone 10 µg/mL, recovered to levels comparable to the no-ceftriaxone control by 24 hours (Fig. 2B).In the 10 8 CFU/mL inoculum, both E. cloacae and S. marcescens had similar levels of viable bacteria compared to control by 24 hours at both ceftriaxone concentrations (Fig. 2C and 2D ).

DISCUSSION
We have identified fundamental differences in the AmpC-related biology of two important Gram-negative pathogens, E. cloacae and S. marcescens.We find that E. cloacae has an increased propensity toward the development of ceftriaxone resistance due, at least in part, to two factors: the E. cloacae AmpC appears to have higher hydrolytic efficiency for ceftriaxone, and ampC transcript levels are an order of magnitude higher upon maximal de-repression.We also observed higher levels of ceftriaxone-stimulated AmpC induction in E. cloacae; these higher levels of enzyme, combined with its higher catalytic efficiency, led to increased survival of wild-type E. cloacae compared to S. marcescens in time-kill curve experiments.
Our findings generally fit well within the existing literature.Our measurement of lower ceftriaxone MICs in E. cloacae with its AmpC coding sequence replaced by that of S. marcescens is consistent with in vitro enzyme kinetics (7,27,28).Similar to our results with ceftriaxone, Guerin et al. observed comparable similar increases in cefotaxime MICs in an independently generated ΔampD mutant, as well as comparable levels of ampC induction upon exposure to cefotaxime (20).Literature is sparse for S. marcescens; to our knowledge, ours are the first data quantitatively examining relative ampC transcript levels in response to experimental de-repression and to induction with ceftriaxone.The underlying mechanism responsible for lower levels of both de-repression and induction in S. marcescens remains unknown.Previous work has revealed that, in contrast to the Enterobacteriaceae, the S. marcescens ampC 5' untranslated region has a stem-loop structure that serves to increase the half-life of the S. marcescens ampC transcript (29).This suggests that differences in transcript levels are presumably not due to transcript stability.
Our findings are in support of recent IDSA guidance that cautions against ceftriax one use in general in E. cloacae infections (17).Given the lower levels of ampC induc tion, ampC de-repression, and the negligible rate of selection for ceftriaxone mutants, their recommendation that ceftriaxone may be used in S. marcescens infections when antimicrobial susceptibility testing indicates susceptibility is reasonable.However, our time-kill curves revealed poor killing of S. marcescens by ceftriaxone at high inocula.It thus may remain prudent, as suggested (17), to consider alternative treatment in S. marcescens infections with a lack of source control or the potential for biofilm formation (in which organisms are generally expected to respond relatively poorly to beta-lactams).This is not a rare situation; S. marcescens can be an exuberant biofilm former (30) and is an important cause of endocarditis, particularly in persons who use intravenous drugs (31)(32)(33), though in the largest case series of S. marcescens endocarditis to date, the microbiological cure rate with ceftriaxone (for susceptible isolates) was high (34).
Our studies have limitations.We chose E. cloacae ATCC 13047 and S. marcescens ATCC 13880 because they are the type strains for the species, but other strains may differ, particularly due to the substantial genetic heterogeneity observed in S. marcescens clinical strains (35).Relevant to this point, the lower baseline ceftriaxone MIC observed here in S. marcescens compared to E. cloacae could have important influences on the mutant frequencies observed.There are isolated reports where S. marcescens with higher baseline cefotaxime MICs have developed resistance through amidohydrolase mutation alone (in the absence of apparent porin deficits) (36,37).However, baseline cefotaxime and ceftriaxone MICs generally tend to be low in S. marcescens clinical strains; in one collection of S. marcescens clinical isolates, only 6% tested ceftriaxone non-susceptible (38).In the past 5 years of SENTRY data, only 14% of submitted S. marcescens isolates tested non-susceptible compared to 23% of E. cloacae isolates (39).Additionally, our kill-curve experiments, while comprehensive, do not replicate in vivo conditions.
Future studies to uncover the mechanisms of differential ampC induction and de-repression between S. marcescens and E. cloacae will be worthwhile.It appears the additional AmpD ortholog in S. marcescens is unlikely to be an important factor, as we have found that deletion of both orthologs did not lead to similar transcript levels.If the S. marcescens AmpC has lower catalytic efficiency for ceftriaxone, as our experiments suggest, then one would predict an overall higher level of peptidoglycan turnover, leading to greater (not lower) levels of AmpC induction and derepression (12).Instead, we would speculate that the lower level induction of ampC expression in S. marcescens vs. E. cloacae may be due to differences in AmpR binding sequences, which are known to be divergent between these species (29), or perhaps due to lower affinity of peptidogly can muropeptides for AmpR itself.Furthermore, it remains to be determined if instead of multiple amidohydrolase mutations, other single mutations can be selected in vitro and in vivo that result in resistance to ceftriaxone in S. marcescens.As an example, one recent clinical report implicated the Cpx envelope stress response in resistance (40).
In conclusion, we have described fundamental differences in the propensity for E. cloacae and S. marcescens toward AmpC de-repression and induction.These findings support clinical guidance, which should enable treatment of S. marcescens infections with narrower-spectrum beta-lactams, preserving cefepime and carbapenems for infections with organisms with higher risk of development of resistance to third-genera tion cephalosporins.

Molecular biology
The accession numbers for the S. marcescens ATCC 13880 and E. cloacae ATCC 13047 genomic sequences used for primer design are CP072199.1 and CP001918.1,respec tively.Allelic exchange using pTOX3 was used to make all in-frame deletions as described previously (41).5' and 3' primers (which denote the deletion boundaries), the E. cloacae protein coding sequence 5' and 3' primers (used to amplify this to knock in to S. marcescens), and qRT-PCR primers are in Table 5. Primers were synthesized at Eton Bioscience (Boston, MA).The insert for the E. cloacae ampC (into the S. marcescens locus) was directly synthesized by Twist Bioscience (South San Francisco, CA)

Minimal inhibitory concentration determination
MICs were determined using Etest per the manufacturer's specifications (bioMerieux).The reported MIC is the mode of at least three separate determinations performed on separate days.If a mode could not be determined after the third sample, additional samples were tested to ensure accuracy.

Time-kill kinetic assays
To determine the in vitro bactericidal activity of ceftriaxone and cefotaxime on wild-type S. marcescens and E. cloacae, overnight cultures of S. marcescens in LB media were diluted 30-fold and E. cloacae 25-fold in LB media.S. marcescens overnight cultures were diluted more to account for faster initial growth kinetics so as to ensure a similar initial inoculum.Then, cultures were incubated with shaking at 37°C.At 1 hour, either LB-alone or LB + ceftriaxone was added to yield final concentrations of 0, 10 ug/mL or 100 ug/mL ceftriaxone.Incubation with shaking was continued, and colony counts were performed at 1, 2, 4, 6, and 24 hours on non-selective LB agar to quantitate viable bacteria.

Relative quantitative RT-PCR assays
To determine relative ampC mRNA transcript levels, total E. cloacae and S. marcescens RNA were isolated from log-phase cultures using the RNeasy Midi Kit (Qiagen).Proce dures were generally according to the manufacturer's directions with the following modifications to ensure satisfactory genomic DNA digestion: twice the volume of lysis buffer was used; DNA digestion was performed with twice the suggested DNAse concentration; and the incubation was extended to twice the suggested duration.cDNA was then synthesized from mRNA using the Verso cDNA synthesis kit (Abgene, Thermo Fisher).Real-time reverse transcription relative quantitative PCR was then performed on the resulting cDNA using EvaGreen dye and the CFX96 Real-Time System (Bio-Rad).The housekeeping gene mdh was used to determine ΔCT values for both E. cloacae and S. marcescens.Three technical replicates were performed for each reaction and subsequently averaged.The graphed values result represent the mean and standard error of the mean from three biological replicates performed on three different days.PCR primers are listed in Table 5. Primer efficiency was established using serial 10-fold dilutions of template DNA.Relative RT-PCR was then performed for each primer pair.The slope of the best-fit line for the trend between log(dilution) and average cycle threshold was then calculated.Finally, primer efficiency was calculated using the formula: Efficiency (%) = (10 (-1/slope) −1)*100.All primer pairs had efficiencies that averaged between 102% and 105%.

FIG 1
FIG 1 Reverse transcription relative quantitative real-time PCR (RT-qPCR) of ampC in wild-type S. marcescens and E. cloacae as a function of the inducing ceftriaxone (CRO) (A) or cefotaxime (CTX) (C) concentrations.Panels B and D are data replotted with the Y-axis expressed as a multiple of the MIC for the indicated antibiotic.Log-phase cultures were exposed to the indicated antibiotic concentration and samples isolated for processing after 1 hour.Values represent the mean and standard error of three independent biological samples.

FIG 2
FIG 2 Time-kill assays.Log-phase S. marcescens or E. cloacae were adjusted to the indicated inoculum (either 10 5 or 10 8 CFU/mL) in either 0, 10, or 100 µg/mL ceftriaxone (CRO).Colony counts were performed on nonselective solid media for 24 hours.Values represent the mean and standard error of three independent biological samples.

TABLE 1
Minimum inhibitory concentrations (MIC) of wild-type and selected S. marcescens (SM) and E. cloacae (EC) deletion mutants a

MIC (μg/mL) Ceftriaxone Cefepime Imipenem Tobramycin
MICs were determined by Etest.The values represent the mode of at least three individual biological samples.
a a EC-ampC-KI denotes the E. cloacae AmpC protein coding sequence inserted into the S. marcescens ampC locus by allelic exchange, and vice-versa for SM-ampC-KI.MICs were determined by Etest.The values represent the mode of at least three individual biological samples.

TABLE 3
Reverse transcription relative quantitative real-time PCR (RT-qPCR) of ampC in wild-type and selected S. marcescens (SM) and E. cloacae (EC) deletion mutants a Values represent the mean of three individual biological samples, with standard errors, and are normalized to expression in the corresponding wild-type isolate. a

TABLE 4
Mutant frequency of the indicated strain plated on the indicated ceftriaxone (CRO) concentra tion.At least three individual biological samples of S. marcescens (SM) and E. cloacae (EC) deletion mutants were grown as overnight cultures, diluted as detailed in Methods, and plated on ceftriaxone agar a a The values represent the mean and standard error of the resulting colony counts, divided by the input CFU.ND, not determined.* <6E-10 indicates that no colonies were detected (out of a total inoculum of 1.7E9 CFU plated).

TABLE 5
Primers used in this study.SM, S. marcescens.EC, E. cloacae