Challenging the role of a NOX4-Piezo1 axis in neutrophil bactericidal function

We are writing to express our concerns over some of the content in “trans-Endothelial neutrophil migration activates bactericidal function via Piezo1 mechanosensing” by Mukhopadhyay et al.¹ The authors claim that mechanical force required for neutrophil migration through endothelial junctions induces Piezo1 mechanosensor activation that potentiates microbial killing via NADPH oxidase NOX4 activity at the neutrophil phagolysosomal membrane. We take issue with the claims related to NOX4 and have considerable concerns about the experimental methods employed, the reagents used, and the primary conclusion that NOX4 is “crucial” for bacterial killing by transmigrated neutrophils (PMNs) (depicted in Mukhopadhyay et al.¹ Figures 1E, 5E-F, 6, S6).

Understanding the shortcomings of the report requires familiarity with decades-long work that characterized a multimeric NADPH oxidase complex, termed NADPH oxidase 2 (NOX2), in PMNs.² This phagocyte oxidase is highly expressed in PMNs and other innate immune cells, and its activation requires subunit phosphorylation, regulatory membrane lipids, and active GTP-bound RAC2. During phagocytosis, the active NOX2 complex assembles on phagolysosomal membranes, producing superoxide anion and, by rapid dismutation, H₂O₂ in the phagosomal lumen, thereby promoting microbial killing by myeloperoxidase-catalyzed generation of hypochlorous acid (HOCl) and other derivative oxidants. Inactivating mutations in CYBB (encoding the catalytic subunit NOX2), NCF1, or other NOX2 complex-associated genes result in recurrent, life-threatening infections in patients with chronic granulomatous disease. In contrast, NOX4 is highly expressed in renal tubular cells with lower expression in myofibroblasts, endothelial cells, and epithelial cells. While NOX2 generates superoxide in large quantities on activation,² NOX4 constitutively produces H₂O₂ at nanomolar concentrations.³ With these principles in mind, we have the following concerns with the publication by Mukhopadhyay et al.¹

The statement that “PMNs express several NADPH oxidase isoforms, NOX1, NOX2 and NOX4…” is not supported by published data. Expression of NOX2, but not of NOX1 or NOX4, has been reported in PMNs (reference 20 does not support the authors’ claim). Cited references 30–32, 51, and 20 lack any information supporting the statements: “NOX4 generates H₂O₂ in the phago-lysosomal compartment…”, “…upregulate the phagolysosomal protein NOX4…”, “NOX4 mediates bacterial killing via ROS generation…”, and “NOX4 is localized in the cup-shaped invaginations of the plasma membrane, specifically in phagosomes and phagolysosomes of PMNs…”, respectively.

The authors’ focus on NOX4 stems from their RNA-seq analysis of tissue-resident subpopulations of transmigrating PMNs after intratracheal LPS challenge (depicted in Mukhopadhyay et al.¹ Figures 5E and 5F; Table S1), showing Nox4 upregulation in wild-type (WT) PMNs whereas Cybb (NOX2) expression is only slightly changed. Inspection of raw RNA-seq read counts before normalization shows much lower counts for Nox4 (average 156) relative to Cybb (average 9,353) across all conditions. Data from Table S1 in the article are used in Figures 6A and 6B to show normalized expression of each NOX relative to transmigrated neutrophils (TM-PMNs) versus non-transmigrated PMNs. ¹ This normalization removes the large relative differences in expression of NOX isoforms. Detailed qPCR data comparing Nox4 versus Cybb expression in WT versus Piezo^−/− TM-PMNs, depicted as ΔCq values, are not presented. Considering the absence of Nox4 expression at baseline (human and murine PMNs, differentiated HL60 cells, and PLB-985 cells at http://collinslab.ucdavis.edu/neutrophilgeneexpression/,⁴ scRNA-seq data in C57BL/6 mice, Broad Institute, https://singlecell.broadinstitute.org/single_cell/study/SCP978), a 2.6-fold upregulation of 8 Nox4 variants (only one full-length protein coding) suggests that NOX4 expression is very low and highly unlikely to contribute to increased bactericidal PMN function, especially in the presence of abundant endogenous NOX2.

Upregulation of NOX4 (combined 7 variants, one full-length) is shown in quiescent human blood PMNs after 30 min treatment with the Piezo1 agonist Yoda1 (Figure 6C, normalized with fold-change indicated; ΔCq values of NOX4 and CYBB not reported¹). NOX4 protein is detected after 2 h of Yoda1 treatment (quiescent human blood PMNs) (Figure 6D¹) and at 2 h postmigration through 5-μm pores (differentiated HL-60 cells) (Figure 6E¹). Since quiescent human PMNs and differentiated HL-60 cells do not express NOX4 at baseline,⁴ it is unclear how much NOX4 protein can be induced within 2 h of treatment. Likewise, the 2-fold increased bactericidal activity (Figure 1E¹) in transmigrated murine PMNs 15 min after intratracheal LPS application is attributed by the authors to upregulated NOX4 protein in a cell type without detectable Nox4 mRNA in bone marrow.

The target-binding selectivity and sensitivity of the anti-NOX4 antibody used (rabbit polyclonal, synthetic peptide to residues 100–200 of human NOX4, Novus NB-110-58849) is not properly validated by the company or the authors. Guidelines for detection of NOX by immunoblot and RT-qPCR⁵ demonstrate that this Novus anti-NOX4 antibody detects a 64kDa non-specific band across all conditions (Figure 5 in Diebold et al.⁵). Consistent with the caveats outlined,⁵ this NOX4 antibody detected in Figure S6E¹ a strong “NOX4” protein band in differentiated HL60 cells lentivirally transduced with control shRNA, although NOX4 is absent from this cell type.⁴ Diebold et al. also addressed published, flawed data where siRNA silencing of NADPH oxidases was shown in cells with very low endogenous or absent NOX expression (Cq values ≥ 30) using defined and validated antibodies.⁵ Given these concerns, we believe that the evidence for NOX4 expression in PMNs reflects the use of non-specific reagents.

NOX2 protein is detected in human PMNs as a single band at 65–70 kDa (Figure 6D, NOX2 antibody information missing¹). Human neutrophil NOX2 is heavily glycosylated, presenting as multiple bands around 91 kDa.² Hence, the results of NOX2 immunoblotting in Figure 6D¹ are difficult to reconcile with the well-characterized post-translational processing of NOX2.

The authors use the drug Setanaxib (GKT137831) as a NOX4 inhibitor (Figure 6F, S6F, and S6G¹). GKT137831 is considered selective for NOX1 and NOX4 at low nanomolar concentrations in cell-free assays,⁶ but at the 10 μM concentration used in this cell-based study, it certainly inhibits other NOX isoforms, most notably NOX2. The compound also interferes with Amplex Red-based H₂O₂ detection used here.⁷ Data validating GKT137831 as a selective inhibitor targeting solely NOX4 activity, if present, but not the highly expressed NOX2 are not presented.

The unresolved questions regarding sufficiently high de novo NOX4 expression in TM-PMNs that can act in a bactericidal manner independently of NOX2 activation requires more stringent experimental design and controls for Figures 6H and 6I.¹ This includes immunoblots performed with validated antibodies to ascertain Nox4 ablation in TM-PMNs derived from infected Nox4^fl/flLy6Gcre+ mice but not from infected Nox4^fl/flLy6G-cre-mice and using mice with NOX2 complex inactivation (Cybb^−/y, Ncf1^−/−, Ncf2^−/−) and another NOX4 inactivation model (Nox4^−/−, Nox4^fl/flMRP8cre+) for sufficiently controlled P. aeruginosa infection experiments. To propose an additional bactericidal mechanism of PMNs without addressing NOX2, the key antimicrobial NOX in PMNs, is highly unusual. Controls for cell-based assays and mice are readily available (JAX, PLB-985 CYBB^−/− cells, PLB-985 NCF1ΔGT cells).

H₂O₂ measurements in Figure 6J¹ do not include calculations of H₂O₂ production by using H₂O₂ standard curves or controls (superoxide dismutase, catalase, diphenyleneiodonium), leaving the reader uninformed as to the identity of oxidants produced or the magnitude of activity. According to the STAR Methods section, the authors stimulate H₂O₂ generation by using phorbol ester (PMA), an agent that potently activates NOX2. Control conditions such as H₂O₂ production by differentiated HL60 cells not subjected to in vitro transmigration or without PMA stimulation are missing. In contrast to the data presented, NOX4 generates H₂O₂ constitutively and is not stimulated by PMA.³ H₂O₂ generation by NOX4 ranges from 30 to 100 nmol H₂O₂/mg protein/h when highly overexpressed,³ while modeling of phagosomal NOX2-derived ROS indicates superoxide flux at 5.2 mM/s, with 2 μM H₂O₂ present that will be immediately utilized for HOCl generation at a rate of 134 mM/min.⁸ Thus, the rate of H₂O₂ production by endogenous NOX2 far exceeds the amount of H₂O₂ generated by NOX4, even in overexpressing cell systems.

The proposed phagolysosomal localization of NOX4 rests on overexpression of inactive NOX4 (NOX4-mGFP, OriGene RC208007L4V) in differentiated HL60 cells without co-localizing phagocytosed bacteria with NOX4 and NOX2, and without measuring phagosomal NOX4-derived H₂O₂ (Figures S6A and S6B¹), both critical determinants to support any claim that a given entity (here, endogenous NOX4 induced by Yoda1 or transmigration) is active at the phagosome. The net contribution of NOX4 to H₂O₂ generation after 30 min of Yoda1 stimulation or after 2 h transmigration could be determined by using readily available tools, for example comparison of PMNs derived from Cybb^−/y and Nox4^−/− mice or use of neutrophil-differentiated CYBB^−/− PLB-985 cells.

Taken together, the use of non-specific reagents (antibodies, inhibitors), lack of rigor in experimental design and interpretation, and failure to take into account the literature on NOX2 and NOX4 lead us to conclude that the claims in this publication—namely, that the molecular mechanism responsible for enhanced bacterial killing by transmigrated neutrophils is due to a Piezo1-NOX4 axis, culminating in neutrophil NOX4 protein expression and NOX4 H₂O₂-mediated bacterial killing, as reported by Mukhopadhyay et al.¹—are not supported by their data and seem implausible from the data provided. There are multiple published research reports, consensus guidelines, and book chapters with up-to-date information about all mammalian NOX, including validation of antibodies, inhibitors, and oxidant measurements.^9,10 Utilizing these readily accessible resources will benefit the validity and reproducibility of research studies on NADPH oxidases.