Pulmonary FABP4 is an inverse biomarker of pneumonia in critically ill children and adults

Manuscript

Lower respiratory tract infection (LRTI) is a leading cause of death worldwide, yet the diagnosis is often challenging since non-infectious respiratory processes can present similarly. Furthermore, existing culture-based microbiologic testing is time-consuming, can return falsely negative with preceding antibiotics, and may identify microbes unrelated to a pulmonary infection. Consequently, clinicians often turn to rapid host response-based peripheral blood biomarkers, such as procalcitonin and C-reactive protein (CRP), for gauging LRTI suspicion and deciding to initiate empiric antibiotics. Procalcitonin has some utility in distinguishing bacterial and viral etiologies of LRTI, but diagnostic performance is limited in critically ill patients, and CRP performs worse.(1,2) Presently, no clinically available host biomarkers accurately distinguish all-cause LRTI from non-infectious respiratory illness.

To address the need for better LRTI diagnostics, we previously utilized respiratory metatranscriptomics to build multi-gene host classifiers that accurately differentiate LRTI from non-infectious acute respiratory syndromes in critically ill adults and children. These classifiers achieved remarkable accuracy, both when used on their own and when integrated with microbial detection.(3,4) While promising, the practical complexities of RNA sequencing precludes immediate implementation in many clinical settings. We therefore sought to identify whether a single gene could accurately diagnose LRTI in patients with acute respiratory failure.

We first evaluated a recently described(4) prospective multicenter cohort of 261 critically ill children with acute respiratory failure (Table 1). Transcriptional profiling was performed on tracheal aspirate (TA) samples collected <24 hours after intubation using previously described methods.(4) Retrospective clinical adjudication classified subjects into four LRTI groups: Definite (clinical LRTI diagnosis with compatible microbiology, n=117), Suspected (clinical LRTI diagnosis with negative microbiology, n=57), Indeterminate (clinical uncertainty regarding LRTI, n=37), and No Evidence (clear non-infectious etiology with negative microbiology, n=50). Most Definite LRTI cases were viral, reflecting the typical microbiology of pediatric community-acquired LRTI.(5) Etiologies of non-infectious respiratory failure included seizures, trauma, airway abnormalities, and ingestion. Differential gene expression analysis between Definite and No Evidence subjects identified 4,718 genes at a Benjamini-Hochberg adjusted p-value (P_adj)<0.05, with the gene fatty acid-binding protein 4 (FABP4) emerging as the most promising candidate biomarker (P_adj<0.0001) (Figure 1A). Five-fold cross-validation on normalized FABP4 gene expression in the Definite and No Evidence groups yielded an area under the receiver operator curve (AUC) of 0.90±0.07 for diagnosing LRTI, equating to a sensitivity of 86±7% and specificity of 90±7% at Youden’s index (Figure 1B). FABP4 classified 73.7% of the Suspected group and 54.1% of the Indeterminate group as LRTI, aligning with stronger suspicion of LRTI in the former.

• Download figure
• Open in new tab

Figure 1: FABP4 performance in adult and pediatric cohorts.

(A) Volcano plot displaying differential expression of all genes quantified by RNA sequencing between Definite LRTI and No Evidence subjects within the pediatric cohort. Log2(fold change) is plotted against -log10(P_adj) for each gene. Statistically significant genes (P_adj<0.05) are shown in black and non-significant genes are shown in grey. FABP4 gene is highlighted in orange as highly downregulated in LRTI patients. (B) Receiver operating characteristic (ROC) curve demonstrating FABP4 performance in diagnosing LRTI in the pediatric (orange) and adult (green) cohorts. The faded lines indicate results from each of the 5-fold validations on normalized FABP4 gene counts. The area under the curve (AUC) is presented as mean ± standard deviation.

Given that LRTI microbiology and immune responses vary with age, it was uncertain whether FABP4 would perform as well in diagnosing adult LRTI. We therefore assessed FABP4 expression in an independent prospective cohort of 202 adults requiring intubation for acute respiratory failure (Table 1).(3) A similar retrospective clinical LRTI adjudication approach identified 44 Definite, 36 Suspected, 66 Indeterminate, and 56 No Evidence subjects. In contrast to the pediatric cohort, the adult Definite subjects had predominantly bacterial LRTI and included subjects with both hospital-acquired and community-acquired pneumonia. Etiologies of non-infectious respiratory failure included cardiac disease, neurologic syndromes, and primary pulmonary processes (e.g., COPD, ILD, pulmonary embolism). Utilizing transcriptional profiling of TA samples collected <72 hours after intubation, we compared FABP4 gene expression across LRTI status. FABP4 exhibited a similarly striking decrease in expression in adults with Definite LRTI compared to No Evidence (P_adj<0.0001) and achieved an AUC of 0.85±0.12 when used as a diagnostic classifier in these two groups (Figure 1). Similar to the pediatric cohort, FABP4 predicted LRTI in 66.7% of the Suspected group and 31.8% of the Indeterminate group.

So what is FABP4, and why might it be downregulated in LRTI? Fatty-acid binding proteins are a family of intracellular proteins that regulate fatty acid trafficking and signaling.(6) In the lung, FABP4 is considered a marker of tissue-resident alveolar macrophages, and severe SARS-CoV-2 infection has recently been shown to drastically deplete the FABP4-expressing alveolar macrophage population.(7,8) Prior studies have also linked increased FABP4 levels to non-infectious chronic pulmonary conditions including bronchopulmonary dysplasia, sarcoidosis, and asthma.(6) The precise mechanism for FABP4 downregulation in infection – or, alternatively, upregulation in non-infectious lung disease – remains unclear. From a diagnostic lens, a study of ventilator-associated pneumonia identified FABP4 as one of the most downregulated proteins in TA, and a study of COPD patients noted decreased sputum FABP4 protein levels in bacterial infection, supporting our transcriptomic findings at the protein level.(9,10)

Taken together, these results demonstrate that FABP4 is a promising age-agnostic, inverse-biomarker for all-cause LRTI in two independent cohorts of critically ill, intubated patients. FABP4 achieved high accuracy for LRTI diagnosis in both cohorts, with performance exceeding those of existing clinical biomarkers(1,2) despite differing ages, microbiology, comorbidity burden, and preceding healthcare exposures. As host immune responses are more dynamic and informative at the site of infection, the identification of an accurate LRTI biomarker from respiratory fluid rather than peripheral blood, while novel, is perhaps not surprising. Diagnostically, FABP4 could add value to standard TA culture, both in discriminating LRTI pathogens from colonizing microbes and in diagnosing LRTI when cultures are hindered by preceding antibiotics. In addition to improving LRTI diagnosis for the individual patient, a pan-pathogen host biomarker of LRTI could have utility for pandemic preparedness and infection control, enabling rapid screening, early isolation, and other prevention measures.

These preliminary findings require independent validation in additional cohorts with both transcriptomic and proteomic approaches. It remains unknown whether FABP4 has diagnostic utility in non-critically ill patients, or when measured in more broadly accessible sample types, such as sputum, nasal swabs, or blood. To better understand the mechanism of FABP4 modulation and to gauge whether particular cell types may mediate or confound FABP4 expression, cytometry and single-cell RNA sequencing on lower respiratory tract samples from patients with LRTI, non-infectious respiratory syndromes, and healthy controls will be valuable. Lastly, an important step towards clinical translation will be development of RT-qPCR and/or lateral-flow protein assays. In sum, we find that pulmonary FABP4 is a novel, downregulated biomarker of severe LRTI across the age spectrum, with promising utility in critically ill patients.

Data and Code Availability

All data, code, and results are available at: https://github.com/infectiousdisease-langelier-lab/LRTI_FABP4_classifier for reviewing and reproducing the analyses.

Ethics Statement

The pediatric cohort study was approved by the Collaborative Pediatric Critical Care Research IRB at the University of Utah (protocol #00088656). Informed consent was obtained from parents or other legal guardians. The adult cohort was approved by the University of California Institutional Review Board (protocol #10-02701) and informed consent was obtained from patients or surrogate decision makers.