A long-term survey of Serratia spp. bloodstream infections revealed an increase of antimicrobial resistance involving adult population

ABSTRACT Serratia spp. is a well-recognized pathogen in neonates; however, limited data are available in adults. We studied microbiological and clinical characteristics of Serratia spp. causing bloodstream infections (BSI) in our institution (January 2005–July 2020). Overall, 141 BSI episodes affecting 139 patients were identified and medical records reviewed. Antimicrobial susceptibility was recovered from our informatics system and 118 isolates from 116 patients were available for further microbiological studies. Whole genome sequencing (WGS) was completed in 107 isolates. Incidence of Serratia BSI was 0.3/1000 overall admissions (range 0.12–0.60), with maximum prevalence (27 episodes, 19.1%) during 2017–2018. Relevant patients’ clinical characteristics were 71.9% ≥60 years (n = 100), with high comorbidity rates (49%, ≥2), 23 (74.2%) of them died within 1 month of the BSI episode. WGS identified all isolates as Serratia marcescens when Kraken bioinformatics taxonomic tool was used despite some which were identified as Serratia nematodiphila (32/118) or Serratia ureilytica (5/118) by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Nevertheless, when using MASH distance, Serratia nevei (63/107), S. ureilytica (38/107), and S. marcescens (6/107) were assigned. Carbapenemase (blaVIM-1) and extended-spectrum β-lactases (ESBL) (blaSHV-12) genes were found in seven and three isolates, respectively, one of them expressing both genes. The worldwide-disseminated IncL/M scaffold plasmid was identified in six VIM producers. Four genotypes were established based on their virulence factors and resistome. Serratia spp. emerged as a relevant nosocomial pathogen causing BSI in elderly patients in our hospital, particularly in recent years with a remarkable increase in antibiotic resistance. ESBL and carbapenemases production related to plasmid dissemination are particularly noteworthy. IMPORTANCE Serratia spp. is the third most frequent pathogen involved in outbreaks at neonatal facilities and is primarily associated with bacteremia episodes. In this study, we characterized all causing bloodstream infection (BSI) in patients admitted to our hospital during a 16-year period (2005–2020). Despite having no neonatal intensive care unit in our hospital, this study revealed that Serratia spp. is a relevant pathogen causing BSI in elderly patients with high comorbidity rates. A significant increase of antimicrobial resistance was detected over time, particularly in 2020 and coinciding with the coronavirus disease (COVID-19) pandemic and nosocomial spread of multidrug-resistant Serratia spp. isolates. extended-spectrum β-lactases and carbapenemases genes associated with plasmid dissemination, typically detected in other Enterobacterales species, were also identified, reinforcing the role of Serratia spp. in the antimicrobial resistance landscape. Additionally, this work highlights the need to reclassify the species of Serratia, since discrepancies were observed in the identification when using different tools.

of Serratia, since discrepancies were observed in the identification when using different tools.
KEYWORDS bacteremia, bloodstream infections, Serratia spp., retrospective study, whole genome sequencing B loodstream infections (BSIs), mainly caused by Gram-negative pathogens, are a major cause of morbidity and mortality worldwide, being also a priority for the World Health Organization (https://www.paho.org/es/70-asamblea-mundial-salud),with variations in incidence and epidemiology between countries (1)(2)(3).Globally, Serratia is low represented among the BSI pathogens, although it has a significant burden in the neonatal population, especially among preterm infants (4,5).However, the microbio logical and clinical features of Serratia spp.and infections caused by this pathogen have been little explored in the adult population, with a lack of a trend analysis of its antibiotic resistance.Serratia species harbor intrinsic antibiotic-resistant genes, such as those responsible for the production of the chromosomal AmpC β-lactamases, which when hyperproduced confers resistance to extended-spectrum cephalosporins (6).It also carries other intrinsic resistant determinant affecting aminoglycosides and polypepti des (7,8).The acquisition of plasmid-mediated resistance mechanisms, such as those involving extended-spectrum β-lactases (ESBLs) and carbapenemases, has also been reported (9).
In the present work, we aimed to assess the incidence of Serratia spp.causing BSI in our institution, which do not have a neonatal intensive care unit (ICU), over a 16-year period (2005-2020), characterizing its distinctive epidemiological, clinical, and microbio logical features.
Patients exhibited high rates of comorbidities, 80 of them (57.6%) with two or more comorbidities highlighting hypertension and oncologic diseases.It should be noted that 31 patients died (22.3%) within the first month of BSI diagnosis, 23 (74.2%) of them had ≥60 years.(Table 1).After bacteremia diagnosis, most of the patients were treated in monotherapy with fluoroquinolones or carbapenems alone.Combination therapy was used in 25.5% of the patients.No significant differences were obtained regarding the antimicrobial treatment and clinical outcome (Table S1).We did not detect any case of selection of AmpC hyperproduction associated with the treatment with cephalosporins.

Antimicrobial susceptibility and resistance phenotypes
The antimicrobial susceptibility profile is shown in Table 2.The last 6 years of the study (2015-2020) showed the highest rates of antimicrobial resistance (Table S2).Overall, tobramycin was the aminoglycoside that presented the highest rate of resistance from this group (48.3%).Cefotaxime had a resistance rate of 15.6%, and the resistant rate for cefepime was 7.8%.Carbapenems (imipenem, meropenem, and ertapenem) were the antimicrobial group with the greatest rate of susceptibility (96.5%, 96.5%, and 94.3%, respectively).ESBL production was corroborated in two isolates, carbapenemase production in six isolates, and both enzymes in one isolate.A positive result in the phenotypic AmpC induction test was detected in 87 out of 97 cefotaxime susceptible isolates (89.7%) (Fig. S1).The induction positivity was not observed in ESBL or carbapenemase produc ers, even in the former with cefepime, due to the expanded cephalosporin resistance conferred by these enzymes.Moreover, according to the AmpC induction test and not considering the result of ESBL and carbapenemase producers, 12 isolates of all 118 isolates (10.2%) hyperproduced the AmpC enzymes, all of them being susceptible to cefepime.

Bacterial typing, whole genome sequencing (WGS), and bioinformatics analysis
The WGS process was successfully completed in 107 of the 118 isolates.Eleven isolates were excluded for the bioinformatics analysis due to low sequence quality.

Genome characteristics and bacterial classification
The median genome size of our isolates was 5 Mb-5.1 Mb, with a G + C content of 59.5% and an average of 4,727 protein-coding sequences.Genomic information based on contig size ≥500 bp is detailed in Table S3.Interestingly, Kraken identified all the isolates as S. marcescens, despite some of these isolates previously identified as S. nematodiphila (32/118) or S. ureilytica (5/118) by MALDI-TOF.Nevertheless, using PATO tool, which is based on MASH distance, Serratia nevei (58.9%, 63/107) was the most prevalent species detected, followed by S. ureilytica (35.5%, 38/107) and S. marcescens (5.6%, 6/107).Interestingly, only five isolates were identified as S. marcescens with both WGS tools and MALDI-TOF MS and only two S. ureilytica were identified with both PATO and Kraken.Results comparing the three identification tools are detailed in Table S4.

Population structure
A similarity tree was constructed with the 107 isolates (Fig. 2) revealing that our Serratia spp.BSI isolates did not present a clonal origin.Moreover, we did not find any clear proximity in the isolates according to their year of isolation.Nevertheless, they were mainly grouped in resistotypes I and II.Distribution of resistotypes was irrespective of Serratia spp.identification.

Plasmid gene content
Plasmid replicon content was identified in 28% (30/107) of the isolates.The families detected belonged to various incompatibility group (Inc) replicons: IncL/M, IncFII, IncY, and IncF [IncL/M (pMU407), IncFII (pCRY), IncY_1, pSM22_1, respectively].In addition, the worldwide spread of IncL/M (pOXA-48) was detected in six isolates, with a coverage and identity of 100%, in all cases related to those strains in which bla VIM-1 production was detected (genotype VI).Regarding isolate 101, the only one co-producing ESBL and carbapenemase, the IncY_1 plasmid was identified.
Col440I_1, belonging to the Col-plasmid family, was detected in five strains, four of them in VIM-1 producers associated to IncL/M (pOXA-48).

DISCUSSION
Bacteremia is associated with high mortality, ranging between 10% and 30% depending on patient risk factors, the origin of the infection, and antimicrobial treatment.Nosoco mial infections due to Serratia spp.have increased, highlighting its pathogenic role, particularly in ICUs.Serratia spp.constitutes the third most frequent pathogen involved in outbreaks at neonatal facilities and is primarily associated with bacteremia episodes (10,11).In our series, 22.3% overall mortality within 1 month of BSI diagnosis was observed, although they were vulnerable given their elderly age and clinical conditions.
Despite having no neonatal ICU in our hospital, S. marcescens represents a frequent pathogen causing BSI in our patients (0.30/1,000 hospitalizations, range 0.12-0.60),71.9% of them with ≥60 years and 22.3% admitted in ICU.In our context, as well as reported by other authors, the most frequent origins of Serratia bacteremia are the genitourinary tract, surgical wounds, respiratory tract, and intravascular catheters; however, in up to 29.1% (41/141) of the cases, the source is unknown (10).This figure is similar to previous studies, in which the source of the infection was not determined in a high proportion of the bacteremia episodes.
Although Kraken, a taxonomic sequence classification system, confirmed that all isolates were identified as S. marcescens, MASH distance tool also revealed a majority presence of S. nevei and S. ureilytica besides a minority presence of S. marcescens.Also, MALDI-TOF MS identified some of them as S. nematodiphila and S. ureilytica.Differences on identification when using different tools have been described in this study.Never theless, MALDI-TOF is currently the reference method in clinical microbiology laborato ries.These species are closely related to S. marcescens, and phylogenetic studies have proposed re-classifying them as S. marcescens based on their core genome phylogeny (8,12).Identification obtained with "classifier" function of PATO tool was employed to search for the closet reference species in the NCBI (National Center for Biotechnology Information), highlighting that the recently described S. nevei is also closely related to S. marcescens, demonstrating the complexity in Serratia spp.taxonomy.
The increase in antimicrobial resistance is one of the most important public health challenges in our society, and its spread is mainly driven by horizontal transfer of resistance genes (13).Intrinsically, S. marcescens has an inducible chromosomal βlactamase, which confers resistance to aminopenicillins, first-and second-generation cephalosporins, remaining susceptible to third-and fourth-generation cephalosporins, monobactams, and carbapenems (14).S. marcescens is characterized by its rapid acquisition of antibiotic resistance, mainly due to mutations in regulatory genes affecting derepression of its inducible chromosomal AmpC β-lactamase but also due to plasmid acquisition.In our series, we did not document cases of selection of AmpC hyperproduc tion after treatment with cephalosporins.This situation coincided with the low risk for significant AmpC production indicated by Tamma et al. (15).
Isolates from our study exhibited high resistance rates, particularly during the last 6 years (2015-2020).This is also the studied period with the highest number of Serratia spp.isolates (n = 67), with resistance rate to cefotaxime (12.8%) and cefepime (7.8%), reaching up to 71.4% in the nine isolates from 2020.This period was also coincident with the description of carbapenemase producing S. marcescens in our hospital (16).Tobramycin is the aminoglycoside that showed the highest resistance rate in this period due to the chromosomal aac6-Ic gene that is related to this intrinsic resistance profile, although other aminoglycosides as gentamicin were not affected (17).
A major limitation of our study could be that the antimicrobial susceptibility results were retrospectively retrieved from the laboratory records and were obtained with the automated systems routinely used in the clinical microbiology laboratory of our hospital.Nevertheless, these data represent real-life susceptibility testing results reported to the clinicians.Broth microdilution, the reference standard method, was only performed prospectively for meropenem, while other newly developed molecules were not tested.Notably, the latest isolates recovered in 2020 exhibited the highest antibiotic resistance rates, demonstrating a significant increase in antimicrobial resistance over time, which also coincides with the first year of the coronavirus disease (COVID-19) pandemic and overuse of antimicrobials in the hospital setting (18).
On the other hand, the spread of antimicrobial resistance mechanisms such as carbapenemases is typically associated with Enterobacterales species, mostly Escherichia coli and Klebsiella pneumoniae.However, recent studies performed at our institution confirmed the contribution of other species, such as S. marcescens and Enterobacter spp (16,19).In our study, WGS showed that all isolates harbored intrinsic or acquired genes conferring resistance to various antimicrobials (defining six distinct genotypes), including β-lactamases.The bla SST-1 and bla SRT-1 genes encoding for the inducible AmpC β-lactamases in S. marcescens were found in all isolates.Their phenotypic induction was observed in 89.7% of cefotaxime-susceptible tested isolates (87 out of 97 isolates).This was not the case in ESBL or carbapenemase producers in which these enzymes hid the potential positive result in the induction test.Even using cefepime, it cannot be ruled out as these resistance mechanisms might also make difficult the correct interpretation of the AmpC induction phenotypic test (15,20,21).ESBL and carbapenemase producers (except the isolate that produced both enzymes) were mostly recovered in the last 4 years of the study.Moreover, we did not detect the bla SME gene, a class A carbapene mase gene previously found inserted in the S. marcescens chromosome (22).
The virulome characterization showed a narrower trait.Only virulence factors related to flagella, proteins of chemotaxis, and enterotoxin were present.Although Serratia species are opportunistic pathogens, their mechanisms of invasion and pathogenesis remain to be elucidated (7,23).
Various plasmid incompatibility groups were detected, and replicons belonging to the IncL/M family were the most frequent.The WGS comparative analysis confirmed that six Serratia spp.BSI isolates (VIM-1 producers recovered during the last years of studied period) carried the IncL/M plasmid, harboring the same class I integron with bla VIM-1 carbapenemase gene, which had been found in previous studies from our institution (13,16), suggesting a considerable ability to persist (24).
In conclusion, this retrospective study helps to understand the role of S. marcescens as a pathogen in the antimicrobial resistance landscape of BSIs, particularly in recent years, with the acquisition of ESBL and carbapenemase genes associated with plasmid dissemination.Moreover, we provide new insights of BSIs due to this pathogen in the elderly population.

Study design, bacterial isolates selection, and patients' characteristics
Between January 2005 and July 2020, a total of 363,168 blood cultures (median 22,698 per year) were processed at Ramón y Cajal University Hospital, a tertiary-level public health center with 1,155 beds that serves approximately 600,000 inhabitants in the northern area of Madrid (Spain).A total of 141 of Serratia spp.BSIs episodes affecting 139 patients were detected in this period.To avoid the study of duplicate isolates, only one positive sample per patient and BSI episode was considered.Nevertheless, isolates obtained 1 month apart from the same patient, which occurs in two patients, were also separately considered.Finally, 118 of the isolates, recovered from 116 patients, were available from the strain collection of the Microbiology Department at the HRyC for further studies.All available isolates (n = 118), originally identified by biochemical tests at the time of their isolation, were re-identified by MALDI-TOF MS, library IVD 4194, and RUO 4274 (Bruker-Daltonics, Germany), and the latter with WGS and bioinformatics analysis (see below).Clinical and epidemiological data were retrospectively reviewed from the clinical chart of all patients (n = 141).For comorbidity, we considered medical conditions that normally influence clinical outcomes such as diabetes or hypertension that exists simultaneously but independently of those associated with BSI.
Bacteremia episodes were classified as HO-HCA, CO-HCA, or CA, according to Friedman's criteria (25).Briefly, episodes that occurred at least ≥48 h after hospitalization with no symptoms of infection at hospital admission were considered to be HO-HCA.Episodes within 48 h of admission were considered as CO-HCA or CA if the patient had had recent contact with a healthcare setting or not, respectively.The study was approved by the hospital's Ethics Committee (reference 136/22).

Antimicrobial susceptibility
Antimicrobial susceptibility MIC values (12 antimicrobials) obtained by broth microdilu tion in automated systems (Wider, Soria Melguizo, 2005-2011 or MicroScan, Beckman Coulter, 2011-2020) were retrieved from our laboratory informatics system from all cases (n = 141).In addition, meropenem was latter tested with standard manual broth microdilution in all available isolates (26).Due to several modification of breakpoints over the study period, the EUCAST 2020 criteria (https://www.eucast.org/)were used for the interpretation of MIC data presented in the manuscript.

Phenotypic and molecular antimicrobial resistance gene detection
The screening to detect ESBL production was performed with the double disc diffusion synergy test (27).Inducible AmpC production was tested with disc approximation test, using cefoxitin and imipenem as inducers, with cefotaxime, ceftazidime and cefepime (28).Carbapenemase production was phenotypically determined with the KPC/MBL/ OXA-48 Confirm Kit (Rosco Diagnostica, Denmark).

Statistical analysis
Differences of clinical characteristics, comorbidities, and antibiotic therapy treatment between survivor and non-survivors were checked by Fisher's exact test.P-values of <0.05 were considered significant at a 95% CI.Statistical analyses were performed with IBM SPSS 23 (Armonk, NY, USA) and GraphPad Prism 8 (San Diego, CA, USA) software.The sample size (n) as well as the specific statistical test in each analysis are shown in the tables.

FIG 2
FIG 2 Similarity tree construction by iTOL with the whole genome of the 107 BSI Serratia isolates.O, oncology; IM, internal medicine; NRS, neurosurgery; ICU, intensive care unit; A, anesthesia; E, emergency; U, urology; CS, cardiac surgery; PN, pneumology; C, cardiology; GS, general surgery; ID, infectious diseases; N, nephrology; PS, plastic surgery; NR, neurology, G, gastroenterology; H, hematology.Number of isolate with year of isolation are shown in brackets.

TABLE 1 Demographic and clinical characteristics of patient's population with bloodstream infections due to Serratia spp. a Patients (n = 139) Survivors (n = 108) Non-survivors (n = 31) P-value
a Data expressed as mean ± SD or n (percentage).P-values in Fisher's exact test between survivors and nonsurvivors are shown.b CVD, cardiovascular disease; COPD, chronic obstructive pulmonary disease.