A Rapidly Fatal Infection With Haemophilus influenzae Serotype E Harboring bla_ROB-1: The Dilemma of Safe De-escalation in the Setting of Potential Extended-Spectrum β-Lactamase Production

CASE

A 66-year-old female with asthma-chronic obstructive lung disease overlap syndrome (ACOS) and rheumatoid arthritis (RA) presented with respiratory distress, altered mental status, and sepsis. Her family reported no sick contacts, recent travel, pets, or other epidemiological exposures, and she had been feeling well, with no fevers, chills, or respiratory symptoms. Her RA did not require treatment with biologic response modifiers, prednisone, or methotrexate, and there was no evidence of pulmonary involvement. She was not continuously oxygen dependent, but she was using 2L intermittently immediately before admission. In the preceding 5 years, she visited the emergency department once and was admitted once for her ACOS.

Upon admission, she required urgent endotracheal intubation and was empirically treated with ampicillin, cefepime, vancomycin, azithromycin, and oseltamivir. Her Charlson Comorbidity Index (CCI), Pitt Bacteremia (PBS), and Simplified Acute Physiology Scores were 4, 3, and 77, respectively. Chest x-ray showed consolidation of her right lung. Gram-stain of her sputum revealed pleomorphic Gram-negative coccobacilli. A multiplexed polymerase chain reaction respiratory viral panel was negative. Blood and sputum cultures grew nonhemolytic, slightly opaque, dome-shaped colonies on chocolate agar. However, there was no growth on sheep’s blood or MacConkey agars. Over the next 48 hours, her temperature and PBS rose to 108.7°F and 8, respectively. She developed shock and multiorgan failure. The organism was identified as a β-lactamase-producing Haemophilus influenzae (HI), and de-escalation to ampicillin-sulbactam, or ceftriaxone, or piperacillin-tazobactam was debated. Due to the fulminant nature of her presentation, the possibility of higher-level resistance to β-lactams, and the fear of possibly inducing further extended spectrum β-lactamase resistance by using ceftriaxone [1, 2], cefepime was continued until susceptibilities could be confirmed. Four days after admission, numerous septic brain emboli occurred, and she died before susceptibility information became available.

Using microbroth dilution and/or disk diffusion tests prepared according to Clinical Laboratory Standards Institute methods and interpretative breakpoints, the isolate was resistant to penicillin and ampicillin, intermediate to amoxicillin-clavulanate and piperacillin-tazobactam, and susceptible to azithromycin, cefepime, and ceftriaxone. Phylogenetic analysis and antibiotic resistance gene content, based on its whole genome sequence (WGS), revealed that the isolate was serotype E and sequence type 966. It was most closely related to isolates from the Netherlands and Italy, but it was distinct from a large outbreak at a military base in Fort Benning, Georgia (Figure 1). The isolate contained bla_ROB-1, a β-lactamase-encoding gene, and 66 putative virulence-associated genes. Amino acid mutations S130G and R244K, which have been shown to confer resistance to some extended spectrum cephalosporins and β-lactamase inhibitors [1], were also present.

Figure 1.

Dendrogram of Haemophilus influenzae from fatal infection. Core genome dendrogram H influenza from this report (red text) and related H influenzae genomes from GenBank (black text). Branch lengths are indicative of strain relationships.

DISCUSSION

We describe an interesting isolate and a noteworthy clinical outcome. The isolate is notable for its resistance and virulence gene content, phenotypic susceptibility, and the challenges encountered during antibiotic susceptibility testing (AST). The patient’s course is remarkable for its fulminance in someone not substantially immunocompromised, nor at the extremes of age. Indeed, her initial CCI between 3 and 4 predicted a 10-year survival of 50%–70%. However, before the special test media could be acquired by the hospital laboratory, the patient had died. This emphasizes the value of having local antibiograms available to guide empiric treatment. In addition, having the AST for the specific isolate available would have fostered stewardship by informing us that it would have been safe to de-escalate to ceftriaxone. Given the presence of β-lactamase production, and a gene with mutations that encode resistance to extended-spectrum β-lactams and some β-lactamase inhibitors, we were extremely hesitant to de-escalate. Although HI infections have increased 50% in some US locations over the last decade [3], and atypical presentations such as pyogenic infections of the genitourinary system have emerged [4], descriptions of type E strains using comparative genomics, and the dilemmas HI can present to clinicians and nonreference laboratories, remain scarce.

Although matrix-assisted laser desorption ionization time-of-flight (MALDI-ToF) has simplified the identification of HI, AST remains complicated [2]. Most hospital laboratories do not provide AST results on individual isolates or antibiograms for HI due to its fastidious nature and the specialized testing media required. In addition, variation of inoculum concentration may lead to falsely resistant results with some β-lactams, particularly for β-lactamase-producing strains such as the one in this report [5]. Furthermore, no interpretative breakpoints exist for many relevant antibiotics. In fact, one of the major AST agencies is only now establishing breakpoints for piperacillin-tazobactam. The other major AST agency states that “results of susceptibility testing using breakpoints provided for oral β-lactams, macrolides, and ketolide agents may not be useful for the management of individual patients” [2, 5]. Even with advances such as MALDI-ToF and WGS, susceptibility testing and preliminary results still can vex clinical decision making. Clinical laboratories continue to struggle generating timely and actionable AST reports, due to regulatory and technical barriers [6, 7]. In addition, although WGS has become the “standard-of-care” for epidemiology, evidence for using WGS-inferred AST to guide clinical decision making is lacking [8].

For HI, susceptibility to most β-lactams is usually predicted by susceptibility to ampicillin [2]. Ampicillin resistance often occurs via β-lactamases production or altered penicillin-binding proteins (PBPs). Strains with altered PBPs are referred to as β-lactamase-negative, ampicillin resistant (BLNAR), but AST agencies debate what constitutes a BLNAR strain and whether BLNAR strains should also be considered resistant to amoxicillin/clavulanate, ampicillin/sulbactam, cefaclor, cefetamet, cefonicid, cefprozil, cefuroxime, loracarbef, and piperacillin/tazobactam despite apparent in vitro susceptibility to these agents [2, 5].

Of the 2 main β-lactamase-encoding genes in HI, bla_TEM and bla_ROB-1, the former is far more prevalent, with only approximately 10% of strains worldwide harboring bla_ROB-1 [1, 2]. Both genes are plasmid-mediated and encode class A serine enzymes that are typically inhibited by clavulanate and tazobactam [1]. ROB-1 degrades cefprozil but not the oxyimino-cephalosporins, ceftriaxone, and cefotaxime. Cefaclor exposure may select for bla_ROB-1-carrying strains, because approximately 70% of cefaclor-resistant strains harbor bla_ROB-1. bla_ROB-1 is more prevalent in the United States and Mexico, where cefaclor use is relatively high [1, 2]. The above suggests that piperacillin-tazobactam, ampicillin-sulbactam, or ceftriaxone are reasonable empiric choices; for BLNAR strains, cefepime should be considered [2].

CONCLUSIONS

Finally, this report demonstrates the ongoing need for further investigations of bacterial virulence correlates and mechanisms [9, 10]. Toward that end, we make the isolate and its WGS available to other researchers.