AMR Genomic Intelligence: The Use of Whole Genome Sequencing to Inform Phenotypic Testing and Rapid Treatment Protocols

This Clinician’s View is written by Jose Alexander, MD, Medical Director of Clinical Microbiology at AdventHealth Central Florida.

The use of whole genome sequencing (WGS) plays a significant role in understanding the magnitude of antimicrobial resistance at AdventHealth. It should not be seen only as a research or diagnostic tool, but as a clinical solution for rapid deployment, use, and analysis of active cases. For the last two years, we have been able to implement multiple next-generation sequencing (NGS) solutions with results available within 24 hours.

Although it has played a direct role in patient treatment by assisting in defining the best combination or synergistic antimicrobial approach, its role is more strategic, as a tool for developing genomic-assisted phenotypic screening, testing, and treatment protocols for future cases. Its impact on understanding AMR epidemiology, including trends, prevalence, and distribution among organisms, is well known, as is its contribution to molecular epidemiology. Moreover, NGS may play a more direct role in guiding and adapting laboratory protocols to the changing landscape of antimicrobial resistance.

At the Microbiology Lab in AdventHealth Central Florida, we rely on a tight and well-defined system for AMR screening, confirmation, and reporting. These criteria are designed to rapidly identify significant resistance mechanisms that may compromise the effectiveness of our standard antimicrobial options and to ensure appropriate and timely infection prevention interventions. These AMR classifications follow antimicrobial phenotypic patterns (based on breakpoints or MIC distributions) and genetic markers.

The AMR classification of organisms allows us to divide the population of a specific species into those with and without specific resistance mechanisms. These two populations, although the same organism, are phenotypically different and clinically represent different therapeutic and epidemiological challenges that should be communicated and addressed differently by the clinical team.

AMR Classification Is the basis for an Appropriate Screening Protocol

For example, P. aeruginosa, classified as Difficult-to-Treat (DTR) according to the IDSA GN AMR guidance, is non-susceptible to all first- and second-line active antimicrobials. These criteria directly correlate with the absence of effective first- and second-line agents and trigger a laboratory reflex protocol for ceftolozane/tazobactam. If it is non-susceptible, we release ceftazidime/avibactam, and if that is also non-susceptible, we release cefiderocol.

The P. aeruginosa DTR classification is clear, and because of that, we have defined and developed our own criteria for the A. baumannii complex DTR. The purpose of this classification is not only to define a strain lacking effective agents but also to guide the laboratory in automatically testing and releasing the next active and appropriate agents, thus informing providers of the next best therapeutic option. This also improves upon the carbapenem-resistant A. baumannii (CRAB) definition, which lacks clear therapeutic guidance, as some CRABs can still be susceptible to ampicillin/sulbactam.

We define A. baumannii as DTR when it is non-susceptible to ampicillin/sulbactam and meropenem. Since ampicillin/sulbactam and meropenem are the two most active beta-lactams and indicated agents, their resistance directly correlates with the need for escalation. For an A. baumannii DTR, our lab automatically releases sulbactam/durlobactam, and if it is minocycline-resistant, eravacycline. In sulbactam/durlobactam non-susceptible strains, cefiderocol is the next agent released.

In Carbapenem Resistant Enterobacterales the Gene Is the Key for Rapid Escalation

In the case of CRE, we define Enterobacterales species by the specific gene present. We have created phenotypic AST profiles for KPC, NDM, OXA-48, and, more recently, IMP. These are the only carbapenemases currently seen in Enterobacterales across our Central Florida institutions.

A K. pneumoniae KPC strain is phenotypically different from NDM or OXA-48 from the perspective of anti-CRE active agents and appropriate therapeutic options. This level of definition and antimicrobial profiling allows us to quickly define additional antimicrobial reflex testing and provide clinicians with the most appropriate agent for rapid escalation based on our cumulative antimicrobial data guided by the gene.

NGS and Culture Integration

The beauty of the NGS implementation at AdventHealth is its seamless integration with cultures and AST as part of a continuous process. If you work today at one of our bacteriology, mycology, or AFB benches, you may be running and analyzing NGS results tomorrow, not only for identification but also for epidemiology, S. pneumoniae serotyping, or genomic intelligence. Clinical Lab Scientists own the NGS process.

Genomic intelligence is an essential part of our responsibility to own and use effectively in keeping our patients safe, and this has never been more evident than in the current changes within the carbapenemase landscape. How do we ensure and confirm that our current laboratory screening protocols can detect emerging threats or carbapenemase variants? Are our current rapid escalation agents against CRE still the most appropriate?

NGS has demonstrated a significant upcoming threat, one that will impact on the way we use genetic information to guide rapid escalation and challenge our current rapid diagnostic tests. This impact could not be more significant than in those carrying NDM.

The wave of new anti-KPC agents and their use, even when appropriate, has created significant selective pressure for NDM. The current CDC report indicating a 460% increase in metallo-β-lactamases (MBL), driven by NDM, reflects the broader spread of this gene across the Enterobacterales group, where KPC was previously more confined to K. pneumoniae and E. cloacae complex.

NDM is typically found co-carried with carbapenemase genes other than KPC. This change in epidemiology and the rise of MBL have required the use of new agents and treatment strategies.

Aztreonam + ceftazidime/avibactam (now aztreonam/avibactam) and cefiderocol are our best bets against these new resistance challenges, and thankfully, guidelines and clinical data have made them the clear choice over traditional options like polymyxins and tigecyclines. However, bacteria are adapting and evolving.

The bacterial response to these anti-NDM approaches will be overwhelming, multi-mechanistic, and far more complex to define than our current clear carbapenemase detection path. This is why, when a CRE presents with unexpected resistance to some of the anti-KPC or anti-NDM agents, a full resistome analysis by whole genome sequencing is our best approach to learn, define, and deploy new testing and treatment strategies. This is the core of AdventHealth AMR Genomic Intelligence.

These strategies start with a case-by-case and out-of-the-box thinking approach to study and challenge KPC- and NDM-resistant organisms when encountered, but they must also inform the definition of new phenotypic patterns, the development of screening workflows, and rapid therapeutic escalation.

At this point, it is time to ask: aren’t we overdue for adopting multi-beta-lactam treatment regimens for these highly resistant organisms? Should we be considering more dual-beta-lactam plus beta-lactamase inhibitor combinations (2BL/BLI)?

This is what AMR Genomic Intelligence looks like.

Case 1: Keep an Eye on non-CP-CRE

A K. pneumoniae non-CP-CRE isolated from blood culture was not a surprise. Typical mechanisms behind this pattern are a combination of ESBL or AmpC plus porin restriction. A low meropenem MIC (between 2 and 8 mg/mL) commonly correlates with the lack of a carbapenemase gene. In this isolate, the meropenem MIC was resistant at 4.

For this case, we tried to avoid meropenem/vaborbactam due to the porin restriction on carbapenems. Although its MIC was susceptible at 2, from a meropenem MIC of 4, it was not significantly recovered by the inhibitor. There was no clear inhibitor activity, demonstrating the porin mechanism.

Ceftazidime/avibactam is, from a rational and mechanistic approach, a good option. It is not affected by porins, and avibactam has a strong inhibitory spectrum against ESBL and AmpC, the other possible mechanisms present. Ceftazidime/avibactam had a susceptible MIC of 2.

Cefiderocol is also a rational option. If porins are the problem, why not simply bypass them through the iron-transport system? But to our surprise, its MIC was resistant at 32.

This isolate was immediately sequenced on our Oxford Nanopore Technologies GridION sequencing platform, and the data were analyzed using the BugSeq database. We intended to learn from it, why is cefiderocol resistant? How is the phenotypic expression of porin restriction encoded in its DNA, and how can we predict it, if possible, from the AST pattern? Also, what should be recommended as a therapeutic option for rapid escalation reducing drug/bug mismatch?

The findings were unexpected and concerning. Whole genome sequencing revealed a K. pneumoniae strain type (ST) 393 without significant virulence factors detected. ST393 is a common circulating strain, not reported or described as highly virulent or as a typical outbreak or spreader strain. At this point, nothing epidemiologically significant.

The resistome analysis showed blaSHV-12, an ESBL gene, and Ompk35/Ompk36 knockout, indicating the genetic inactivation or deletion of these significant porin channels used by carbapenems. But how can this explain cefiderocol resistance when it is able to overcome porin restrictions by using the iron-transport system?

Several articles describe cefiderocol resistance in K. pneumoniae carrying blaSHV-12 and knockout or loss of outer membrane proteins OmpK35 and OmpK36. The resistance is explained by the capability of SHV-12 to hydrolyze cefiderocol by accommodating it into its active site, which is facilitated by its unique amino acid sequence and its overexpression.

Although cefiderocol can overcome porin restriction, the activity of SHV-12 appears to be the main driving mechanism behind the resistance. Avibactam can inhibit and restore cefiderocol activity, just as it does with ceftazidime.

The most concerning aspect is that this K. pneumoniae was not a hypervirulent or rapidly spreading strain such as ST11, ST23, or ST395, but a common circulating strain that was able to evolve through selective pressure, adapting into a clinically significant AMR strain.

This exemplifies that multi-mechanistic and complex resistances are not limited to highly virulent or outbreak strains but can occur in any species that undergoes appropriate selective pressure.

What we learned

Ceftazidime/avibactam should be the initial and default rapid escalation agent for non-CP-CRE infections (so far). The drivers of this resistance pattern, porin restriction and overexpression of ESBL/AmpC, can be overcome by avibactam, and ceftazidime is less impacted by porins.

Our NGS results demonstrate that SHV-12 is a common circulating gene, and it must be assumed that it is spreading. Additionally, these non-CP-CRE strains fly under the radar of the BCID platforms and will not be identified until 48 hours later when the full AST panel is performed.

This AMR evolution reinforces the case for implementing rapid AST directly from positive blood cultures. This could be the only option to rapidly identify and appropriately escalate treatment in patients infected with organisms carrying this combination of mechanisms. This is an example of how NGS can inform and define the use case for new diagnostic tools, such as rapid phenotypic AST from blood cultures.

Case 2: The Rise of the Variants, KPC-31

In 2024, before the official introduction of our genomic intelligence, we had a glimpse of what we needed to do. We encountered our first ceftazidime/avibactam-resistant KPC (and at this point, not the only one). Although initially thought to be a major error, it was later confirmed phenotypically and by whole genome sequencing.

A Gram-negative rod from a positive blood culture was identified as K. pneumoniae carrying a KPC gene by BCID. As part of our escalation protocol, the patient was placed on meropenem/vaborbactam. In addition to the first- and second-line AST, we rapidly set meropenem/vaborbactam, ceftazidime/avibactam, and imipenem/relebactam as part of our reflex protocol for KPC. Both carbapenems (meropenem and imipenem) and their combinations with inhibitors were susceptible, but ceftazidime/avibactam was resistant.

Thought to be a major error, it was repeated, again resulting as resistant, while meropenem and imipenem were also confirmed as susceptible. Therefore, we have a K. pneumoniae carrying a KPC, initially detected by BCID and confirmed by PCR (CARBA-R) directly from the colony, which is not hydrolyzing carbapenems but is resistant to avibactam. The only explanation was a KPC variant resistant to ceftazidime/avibactam.

Whole genome sequencing, at this time performed on our old workhorse sequencer, the Nanopore MK1C, and analyzed using the bioMérieux EPISEQ database, identified a K. pneumoniae ST307, a well-known hypervirulent global strain, carrying a ceftazidime/avibactam-resistant KPC, specifically blaKPC-31. In addition, a full arsenal of beta-lactamases was detected, including ESBLs such as blaCTX-M-15, blaTEM-6, and blaSHV-28. Penicillinases such as blaOXA-1 and blaTEM-1, along with other resistance markers, were also present across three different plasmids.

Additional phenotypic testing showed aztreonam/avibactam to be susceptible, and curiously, ceftazidime was not recovered even when combined with vaborbactam or relebactam. The resistance of KPC-31 to avibactam inhibition may be dependent on the antimicrobial combination rather than the inhibitor itself, since neither vaborbactam nor relebactam inhibited it in the presence of ceftazidime.

Although more data is needed, it seems to suggest that KPC-31 is an inhibitor-resistant and antimicrobial-specific beta-lactamase. The concerning aspect of these variants (like others such as KPC-44, KPC-93, KPC-203, KPC-205, etc.) is their ESBL-like pattern. Although carbapenemases and molecularly belong to the KPC class, they present themselves phenotypically as ESBLs.

A routine AST showing 3rd- and 4th-generation cephalosporin resistance and carbapenem susceptibility, from any source other than blood, may result in this isolate being reported as ESBL. If our KPC-31 had been isolated from any source other than blood, it would have been missed, misidentified, and reported as ESBL.

Mutations of the omega-loop of these KPC variants, especially in the residues 164-179, and the 270-loop, are a well-known evolution of resistance against ceftazidime/avibactam.

What we learned:

1- It is critical that we adapt our CRE screening to new KPC variants. We have implemented new AST panels that include ceftazidime/avibactam and meropenem/vaborbactam, not only for rapid reporting when a KPC or OXA-48 is detected but also to be included in the routine CRE screening performed through our CRE algorithm. A K. pneumoniae isolate from any culture with an ESBL pattern that tests resistant to ceftazidime/avibactam and susceptible to meropenem/vaborbactam, even when not reported, would be automatically flagged as “possible KPC,” requiring PCR confirmation.

2- All infections caused by KPC-producing Enterobacterales should be rapidly escalated to meropenem/vaborbactam instead of ceftazidime/avibactam. The use of ceftazidime/avibactam in a KPC-carrying Enterobacterales should begin only after susceptibility testing is available.

3- The use of rapid AST from blood culture may provide the evidence needed to rapidly move away from ceftazidime/avibactam in a KPC-positive isolate.

Case 3: The Storm is Coming

If we are waging a war against bacteria and NDM represents the state-of-the-art of their arsenal, E. coli is leading the charge.

We have experienced, sporadically, NDM-producing Enterobacterales that are non-susceptible to aztreonam + ceftazidime/avibactam or cefiderocol, but until now, an isolate resistant to both had not been seen.

An E. coli, isolated from a urine culture, resulted as CP-CRE with an NDM gene detected. As expected, this organism was resistant to every single beta-lactam (including aztreonam), quinolones, trimethoprim/sulfamethoxazole, and nitrofurantoin. It was typically susceptible to fosfomycin.

Our reflex protocol calls for aztreonam + ceftazidime/avibactam, which resulted as resistant with an MIC of 16. Immediately, cefiderocol was reflexed, also resulting as resistant with an MIC of 128.

What would be our treatment options if this organism were isolated from a respiratory or bloodstream infection?

WGS was performed, demonstrating an E. coli ST66 O8:H9. It has been detected in molecular epidemiology studies of food-borne Shiga toxin-producing E. coli (STEC) and hemolytic uremic syndrome-associated E. coli (HUSEC), but it was its resistome analysis that set it apart from any other CRE seen in our institutions.

Carrying four plasmids (two conjugative and two mobilizable), we looked for the main target of our analysis, the NDM. Two beta-lactamases, blaNDM-5 and blaCMY-42 (AmpC), were detected. The presence of these two beta-lactamases explains the broad hydrolytic spectrum against beta-lactams but not against the two anti-NDM agents.

A PBP3 mosaic insertion ftsI_N337NYRIN clearly defines the resistance against aztreonam and cefiderocol. Here, unlike the two previous cases, we see the inclusion of a different mechanism from beta-lactamase and porin restriction. The PBP mutation alters the active site, decreasing the affinity of the PBP for beta-lactams.

Another interesting mechanism observed directly impacted cefiderocol and may explain, along with the PBP mutation, the high MIC level observed. A CirA knockout deletes or modifies the outer membrane receptor for siderophores. This is a critical and significant resistance mechanism since it specifically targets cefiderocol by disrupting its unique way of entering the cells through hijacking the iron-transport system.

What we learned

Although aztreonam + ceftazidime/avibactam, and now aztreonam/avibactam, form the backbone of our anti-NDM strategy, the option of additional agents such as cefiderocol is key for rapid escalation.

This case challenged both of our options, but fortunately, it was isolated from a UTI where specific agents such as fosfomycin were still active. What are our options when detected from non-urine sources?

We tried multiple combinations without success, such as cefiderocol + ceftazidime/avibactam, thinking outside the box and beyond the common approaches. I like to think of this as the reason we, Medical Microbiologists, exist.

We tested cefepime/zidebactam, a new BL/BLI currently submitted for FDA approval. With only investigational breakpoints, this agent resulted in an MIC of 0.12.

Another interesting result was obtained with sulbactam/durlobactam. With the known antimicrobial activity of durlobactam, this agent tested with an MIC of 0.06 at 4 mg/mL. Although there are no established breakpoints or indications against Enterobacterales, the lack of affinity between durlobactam and NDM, along with the inhibitory activity across non-metallo-β-lactamases such as CMY-42 (AmpC), may explain its activity. Also, the mosaic insertion ftsI_N337NYRIN does not seem to impact the activity of durlobactam.

This case showed us that new agents are needed for these (for now) sporadic cases, where our main anti-NDM agents fail. The adoption of cefepime/zidebactam and an innovative approach for sulbactam/durlobactam are beginning to be rationalized and defined at AdventHealth.

Case 4: Well, well, well, Look Who Is Here, IMP

After seven years of testing and refining CRE screening and confirmation, NGS discovered a new regional challenge, our first IMP (MBL) from a clinical isolate.

An E. cloacae complex isolated from a respiratory source resulted phenotypically as CRE. After the CARBA-R detected a KPC, our bench microbiologist noticed an unusual discrepancy: meropenem/vaborbactam and ceftazidime/avibactam were resistant.

E. cloacae complex is not an unusual suspect. Along with K. pneumoniae, it is one of the most prevalent KPC producers and one of the most common non-CP-CRE with porin restriction mechanisms. So, are we seeing a KPC-31 plus porin restriction? This would explain the susceptibility pattern observed.

As part of our Genomic Intelligence, and thanks to the technical acuity of our team, it was selected for rapid WGS and further review.

The findings were critical for our epidemiology, and it is important to reemphasize that this was only possible because of the technical expertise and attention of our microbiology team.

The isolate was fully identified as Enterobacter hormaechei, the most common species within the E. cloacae complex. It belonged to ST204, without any published significant information, but its complex resistome was the main source of concern.

Three plasmids were identified: two conjugative and one non-mobilizable. One of the conjugative plasmids carried multiple resistances against aminoglycosides, quinolones, and tetracyclines, but the surprise was the presence of blaKPC-45 Like our previous KPC-31, it is a ceftazidime/avibactam-resistant variant. But do Ompk35 and Ompk36 knockouts explain meropenem/vaborbactam resistance?

The non-mobilizable plasmid carried only two resistance genes: blaOXA-2, a penicillinase, and the main actor of the story, blaIMP-18, our first detected MBL of this group. It is important to note, for awareness, the limitation observed in the CARBA-R for detecting some IMP variants. This IMP-18 was not detected by our PCR and may be considered the reason for the low reported prevalence of IMP. Without our genomic protocol based on unusual phenotypic patterns, this carbapenemase would have been missed and still flying under our radar.

The presence of IMP by itself explains the mechanism, clearly an MBL phenotypic pattern. This changes the rules of the game and, more than ever, reinforces our commitment to analyzing and learning about these profiles, translating them into new rapid phenotypic screening and treatment escalation approaches.

What we learned Our use of anti-KPC agents over the last 10 years selected for NDM, but perhaps it extends beyond NDM. Besides the narrowed Enterobacterales species able to modify their porins and the inhibitor-resistant KPC variants, the rise of MBL across the U.S. may have a hidden component, IMP. The lack of detection by the robust and reliable CARBA-R may be the escape route for this gene. We may have witnessed the evolutionary advantage of IMP in evading a widespread CRE PCR assay.

This E. hormaechei was tested against aztreonam + ceftazidime/avibactam with an MIC of 16 and aztreonam/avibactam with an MIC of 8. Cefiderocol tested as susceptible with an MIC of 2 and was our drug of choice for this case.

Additionally, sulbactam/durlobactam showed no potency with an MIC >64, and cefepime/zidebactam, with an MIC of 4, would be considered a reserved last option.

How Has Our Enterobacterales CRE Screening Changed

1- All Enterobacterales that are non-CRE but ESBL or SPICE with ceftazidime/avibactam resistance must be reflexed to CRE confirmatory testing. This is an important pattern for ceftazidime/avibactam-resistant KPC variants, with two variants already circulating in our region (KPC-31 and KPC-45). Further WGS for deeper analysis may be needed.

2- All CRE that are resistant to ceftazidime/avibactam and meropenem/vaborbactam and lack MBL detection by PCR should be reflexed to WGS. We have documented the presence of IMP circulating in our region.

3- Aztreonam/avibactam and/or cefiderocol non-susceptible, MBL-producing CRE should be reflexed to WGS for evaluating their resistance profiles and defining the prevalence of new beta-lactamase variants, PBP, and siderophore mutations.

4- Switching from PCR-based carbapenemase confirmatory testing to enzymatic immunoassay (EIA) testing to strengthen our detection capabilities against IMP is necessary.

These approaches have two main goals: to create rapid phenotypic screening and to provide rapid escalation recommendations based on resistome and in-vitro testing.

Our Genomic Intelligence program is intended to translate the complex genetic profile into rapid and actionable phenotypic patterns that can be deployed at any microbiology lab without access to NGS and expand their screening capabilities for new AMR challenges, providing evidence for rapid therapeutic escalation.

The use of anti-MBL agents is giving rise to multi-mechanistic approaches. PBP modification and siderophore mutations, along with MBL, are clearly limiting our reach in rational and appropriate treatment options.

The adoption of cefepime/zidebactam into our arsenal seems clear. We need an escalation option beyond our latest anti-NDM agents. The in-vitro and genomic results have provided clear evidence that we are at that level of the game.

Exploring a niche for sulbactam/durlobactam along with other beta-lactams is underway. There is evidence of in-vitro activity, and although more testing and analysis are needed, it seems to offer a potential option.

The detection of IMP also increases our awareness of infection prevention. If we do not see it, we do not report it. The next wave of beta-lactam resistance is on its way, and infection prevention actions are the cornerstone of our efforts to slow it down and protect our patients and staff.

NGS is not only filling a clinical diagnostic gap in our institutions but also driving the changes and upgrades to traditional phenotypic testing and screening. Moreover, it is becoming the default tool to define and validate the need for new rapid treatment escalation protocols, reducing the time that patients remain in inappropriate therapy and increasing the likelihood of a positive outcome.

Our adoption of NGS tools by our bench-level scientists is a key example of how NGS should be driven and used by expert microbiologists. NGS, just like MALDI-TOF and PCR, is another essential tool to own and implement.