Bloodstream infections (BSIs) represent a critical category of microbial infections where the rapid identification of causative pathogens is essential for timely and effective clinical decision-making (Peri et al., 2022). Accelerated diagnostic methods play a pivotal role in enabling the prompt initiation of targeted antimicrobial therapy (Lamy et al., 2020), particularly in patients at risk of sepsis, an infectious syndrome associated with alarmingly high morbidity and mortality rates (GBD 2019 Antimicrobial Resistance Collaborators, 2022). Furthermore, the increasing global prevalence of antimicrobial resistance (AMR) in BSIs (Diekema et al., 2019; European Antimicrobial Resistance Collaborators, 2022) underscores the urgent need for diagnostic approaches that not only identify pathogens but also detect key AMR markers. This is particularly relevant for Gram-negative (GN) bacterial species, such as third-generation cephalosporin-resistant and carbapenem-resistant Enterobacterales (GBD 2021 Antimicrobial Resistance Collaborators, 2024; Ikhimiukor et al., 2024), where AMR significantly limits therapeutic options and emphasizes the importance of its early detection.

Although the clinical impact of carbapenemase production is well established, the prevalence and significance of non-carbapenemase-producing extended-spectrum β-lactamase (ESBL)-producing Enterobacterales remain less defined. The Centers for Disease Control and Prevention (CDC) reported that in 2017, there were an estimated 197,400 cases of ESBL-producing Enterobacterales among hospitalized patients in the United States, resulting in approximately 9,100 deaths (CDC, 2025). Additionally, a study from Finland observed that the annual proportion of ESBL-producing E. coli among blood isolates increased from 2.4% to 8.6% in males and from 1.6% to 6.4% in females over a 12-year period (Ilmavirta et al., 2023). Regarding ESBL types, Castanheira et al. (2021) noted that the global dominance of CTX-M-type β-lactamases has largely supplanted SHV-type ESBLs, reflecting a shift in epidemiology. However, SHV-ESBLs are often encoded by self-transmissible plasmids that frequently harbor resistance genes to other antibiotic classes, suggesting that their contribution to β-lactam resistance might still be significant, especially in specific geographical contexts (Liakopoulos et al., 2016).

Conventional culture-based methods remain the cornerstone of BSI diagnosis (Lamy et al., 2020), but their inherent slowness—requiring 24–72 hours for pathogen identification and antimicrobial susceptibility testing (AST) (Miller et al., 2018)—poses challenges in time-sensitive clinical settings. To address these limitations, molecular diagnostic assays such as the BioFire Blood Culture Identification 2 (BCID2) panel (bioMérieux, Marcy l’Étoile, France) (Berinson et al., 2021) have been introduced to accelerate pathogen identification and resistance gene detection directly from positive blood cultures (PBCs) (Peker et al., 2018). These assays provide actionable results within hours, allowing clinicians to initiate targeted therapy earlier than culture-based methods. However, their limitations, including high costs and restricted coverage of detectable pathogens and AMR markers, highlight the ongoing need for innovative solutions in molecular diagnostics.

This study reports on the evaluation of a novel molecular assay, the Gram-negative resistance (GNR) microchip (Alifax S.r.l., Polverara, PD, Italy), which received the Conformité Européene (CE)-in vitro device (IVD) certification in 2022 for the detection of clinically relevant AMR genes in GN bacterial species from PBCs. The assay specifically targets β-lactam resistance genes, including those encoding ESBLs and carbapenemases, and was evaluated using both simulated and clinical PBCs.

We studied GN bacterial organisms from simulated (n=146) and clinical (n=106) PBC samples, corresponding to the technical and clinical evaluations of the GNR microchip assay ( Figure 1 ). All organisms exhibited phenotypic antimicrobial profiles suggestive of acquired ESBL/AmpC and/or carbapenemase-encoding genes, based on MIC values exceeding the EUCAST-established ECOFFs or resistant breakpoints for key β-lactams, including third-generation cephalosporins (cefotaxime and ceftazidime) and carbapenems (imipenem and meropenem).

As shown in Table 1 , using WGS as the reference assay, the GNR microchip assay correctly detected at least one β-lactam resistance gene in all 146 (100%) organisms from simulated samples. This corresponded to a total of 203 β-lactam resistance genes, including 84 encoding ESBL/AmpC-type enzymes (39 CMY-2-like, 38 CTX-M-1/9 group, and 7 SHV-ESBL) and 119 encoding carbapenemase-type enzymes (37 KPC, 27 NDM, 16 VIM, 16 OXA-48-like, 13 OXA-23-like, and 10 IMP). Among these, 49 organisms (47 Enterobacterales and 2 Acinetobacter baumannii) carried more than one β-lactam resistance gene. Notably, one K. pneumoniae isolate was positive for an SHV-ESBL gene (bla _SHV-31) by WGS but was also positive for a KPC-encoding gene, which was correctly detected by the GNR microchip assay.

As shown in Table 2 , the GNR microchip assay detected at least one β-lactam resistance gene in 104 (98.1%) of 106 organisms from clinical samples. Excluding 13 genes not detectable by the BCID2 panel assay (6 encoding CMY-2-like, 2 encoding SHV-ESBL, and 5 encoding OXA-23-like; all identified by WGS), this corresponded to a total of 113 β-lactam resistance genes correctly detected by the GNR microchip assay. These genes included 65 encoding ESBL-type enzymes (all CTX-M-1/9 group) and 48 encoding carbapenemase-type enzymes (32 KPC, 7 VIM, 6 NDM, and 3 OXA-48-like). Additionally, two organisms (P. aeruginosa, 1.9%) tested negative for ESBL/AmpC- or carbapenemase-encoding genes by the GNR microchip assay. WGS analysis confirmed the absence of acquired β-lactam resistance genes in these isolates, suggesting that alternative mechanisms, such as increased expression of efflux pumps, may explain their β-lactam-resistant phenotype. Six organisms (2 Escherichia coli, 1 Klebsiella pneumoniae, 1 A. baumannii, 1 Enterobacter cloacae, and 1 Citrobacter freundii) carried more than one β-lactam resistance gene. These genes were concordantly detected by both the GNR microchip assay and the WGS reference assay and, as expected, only partially by the BCID2 panel assay. Details on these genes are provided as Supplementary Material ( Supplementary Table S2 ).

Table 3 provides an overview of β-lactam resistance gene variants identified in this study by WGS analysis. Among organisms from simulated PBC samples, the most frequently detected β-lactamase types were CMY-2-like (n=39), CTX-M-1/9 group (n=38), KPC (n=37), and NDM (n=27), collectively accounting for 141 (69.1%) of the 204 β-lactamase-encoding genes identified. The predominant variants within each β-lactamase type were CMY-16 (21/39), CTX-M-15 (35/38), KPC-3 (25/37), and NDM-1 (24/27). For CTX-M variants (Naas et al., 2017), CTX-M-15 was the most prevalent and belongs to the CTX-M-1 group, which also includes CTX-M-32 (one organism) and CTX-M-55 (one organism), while CTX-M-27 was the only representative of the CTX-M-9 group (one organism). For KPC variants (Naas et al., 2017), KPC-3 was the most frequent (25/37) and, along with KPC-2 (three organisms), belongs to the carbapenemase class. In contrast, KPC-31 (four organisms), KPC-66 (two organisms), KPC-49 (one organism), and KPC-50 (one organism) are classified as inhibitor-resistant (IR) extended-spectrum β-lactamases (ESBLs), while KPC-29 (one organism) is classified as an IR carbapenemase, based on the established β-lactamase classification system (Naas et al., 2017).

This study provides a comprehensive evaluation of the Alifax GNR microchip assay for the detection of β-lactam resistance genes directly from PBC samples. To our knowledge, this is the first study assessing its performance on a large set of PBC samples, including both simulated samples—derived from WGS-characterized isolates—and clinical samples. The findings demonstrate that the GNR assay reliably detects a wide spectrum of β-lactam resistance genes, with a broader target range compared to the BCID2 panel assay, which served as a comparator for clinical samples.

In simulated samples, the GNR assay detected all but one of the β-lactam resistance genes identified by WGS (99.5% concordance). The single discordant case involved an SHV-ESBL gene (bla _SHV-31) that was not detected by the GNR assay, though a coexisting KPC gene was correctly identified. In clinical samples, 98.1% of organisms carried at least one β-lactam resistance gene detected by the GNR assay, and for genes targeted by both assays, results were fully concordant with those of the BCID2 assay. WGS analysis confirmed 13 additional β-lactam resistance genes not included in BCID2, further supporting the broader detection capability of the GNR assay.

The inclusion of bla _CMY-2-like and bla _OXA-23-like, which are absent from the BCID2 assay, enhances the GNR assay’s diagnostic utility. CMY-2 is the most common plasmid-mediated AmpC β-lactamase in Enterobacterales (Pires et al., 2015), while OXA-23 is a key carbapenemase in A. baumannii (Koirala et al., 2020), highlighting the clinical relevance of these targets. However, the GNR assay does not include some β-lactam resistance genes relevant for Acinetobacter species such as OXA-58 and OXA-24/40 (Evans and Amyes, 2014), which could further enhance its coverage.

WGS analysis in simulated samples provided insights into the distribution of β-lactamase variants, emphasizing the relevance of distinguishing functionally different groups. The classification of KPC enzymes, for instance, is increasingly recognized as clinically significant (Ding et al., 2023). While KPC-3 and KPC-2 function as carbapenemases, other variants such as those found in this study (KPC-31, KPC-66, KPC-49, and KPC-50) exhibit IR-ESBL activity, whereas KPC-29 is classified as an IR-carbapenemase. Notably, KPC-31 and KPC-66 have been linked to resistance to ceftazidime-avibactam (CZA) (Ding et al., 2023), underscoring the need for precise differentiation of β-lactamase variants to inform antimicrobial therapy. Similarly, while the GNR microchip assay effectively detects CTX-M-producing Enterobacterales, its grouping of CTX-M targets into CTX-M-1/9 (and CTX-M-2/8) does not fully account for the clinically relevant distinction between CTX-M-1 and CTX-M-9 subgroups. This differentiation is particularly important given that certain CTX-M variants have been associated with distinct epidemiological trends and β-lactam resistance profiles (Castanheira et al., 2021; Bush and Bradford, 2020). A more refined classification of CTX-M and KPC subtypes in molecular assays could improve the clinical utility of β-lactam resistance detection, enhancing treatment decisions and antimicrobial stewardship efforts.

While this study provides strong technical evaluation of the GNR microchip assay, several limitations should be acknowledged. First, the study design did not include a consecutive clinical sample set but instead enriched for β-lactam-resistant organisms based on phenotypic criteria, limiting the generalizability of findings to routine clinical workflows. Second, WGS was not performed for all clinical PBC samples, leaving some resistance gene profiles incompletely characterized. Third, the study does not assess the potential clinical impact of the assay. Rapid molecular assays, including the GNR microchip assay, are most valuable when they facilitate early targeted therapy adjustments (Mauri et al., 2024), particularly in BSIs where timely intervention is crucial (World Health Organization, 2025). Evaluating whether the assay enables faster clinical decisions, reduces time to appropriate therapy, or improves outcomes would be critical to justify its integration into diagnostic workflows.

Fourth, unlike BCID2, the GNR microchip assay requires an additional Gram stain step to select the appropriate cartridge, which may delay processing and reduce suitability in time-critical settings. Fifth, although the MM system also provides species identification (ID) when combined with the GNID cartridge—designed to detect 15 key GN pathogens including E. coli, K. pneumoniae, Proteus spp., P. aeruginosa, A. baumannii, Stenotrophomonas maltophilia—species-level ID data were not reported. Nonetheless, all isolates, including K. oxytoca and P. aeruginosa, were correctly identified, contrasting with prior reports of occasional failures in polymicrobial samples (Mauri et al., 2024).

Sixth, we did not report results for an additional 46 samples with β-lactam-susceptible organisms, in which the GNR microchip assay did not detect any β-lactam resistance genes. This is because no WGS analysis was performed on these isolates that would have been necessary to confirm the GNR microchip assay-negative results. Seventh, the assay’s performance in polymicrobial samples was not evaluated, and its ability to resolve mixed resistance profiles (e.g., E. coli + K. pneumoniae) remains to be determined. Eighth, while potential interference from sample-related inhibitors (e.g., heparin, hemolysis) was not assessed in our simulated PBCs, internal validation by the manufacturer did not reveal significant effects from common inhibitors.

Ninth, although no formal cost-effectiveness analysis was conducted, the combined cost of the GNR and GNID cartridges is below €100, which is lower than the BCID2 panel assay (>€140). In settings where MALDI-TOF MS-based identification is already implemented, pairing it with the GNR microchip assay may offer a focused and cost-effective diagnostic alternative for GN-PBCs. However, this approach may require specific workflow adaptations and staff training. Tenth, while the GNR microchip assay received CE-IVD certification in 2022, the process of certification under the new European IVDR framework is ongoing, and the assay is not currently FDA-cleared. In contrast, the BCID2 panel is FDA-cleared and widely adopted in clinical practice. These regulatory differences may influence the assay’s adoption in different healthcare systems.

Finally, the GNR microchip assay focuses specifically on β-lactam resistance and does not detect resistance markers for other antibiotic classes. Future versions could benefit from expanded gene targets, particularly for non-fermenting GN bacteria (Huang et al., 2024), and the inclusion of point mutations affecting β-lactamase activity (Benhadid-Brahmi et al., 2025). Integration with phenotypic AST systems—such as the Vitek Reveal rapid AST by bioMérieux (Menchinelli et al., 2024)—could enhance clinical utility by offering a broader phenotypic-genotypic resistance overview (Cortes-Lara et al., 2025). Several of these aspects should be carefully evaluated before adopting the assay in specific diagnostic scenarios.

In conclusion, the GNR microchip assay represents a reliable molecular tool for detecting β-lactam resistance genes from PBC samples, with potential applications in routine diagnostics. Its expanded gene coverage compared to existing panels is a major strength, particularly for AMR determinants not targeted by other assays. However, further validation in larger, prospective clinical cohorts is warranted to establish its full diagnostic value. Future studies should also assess hands-on time, turnaround time, and real-world implementation, including clinical impact and antimicrobial stewardship outcomes.