In Vivo Evolution of a Klebsiella pneumoniae Capsule Defect With wcaJ Mutation Promotes Complement-Mediated Opsonophagocytosis During Recurrent Infection

Abstract

Background

Klebsiella pneumoniae carbapenemase–producing K pneumoniae (KPC-Kp) bloodstream infections are associated with high mortality. We studied clinical bloodstream KPC-Kp isolates to investigate mechanisms of resistance to complement, a key host defense against bloodstream infection.

Methods

We tested growth of KPC-Kp isolates in human serum. In serial isolates from a single patient, we performed whole genome sequencing and tested for complement resistance and binding by mixing study, direct enzyme-linked immunosorbent assay, flow cytometry, and electron microscopy. We utilized an isogenic deletion mutant in phagocytosis assays and an acute lung infection model.

Results

We found serum resistance in 16 of 59 (27%) KPC-Kp clinical bloodstream isolates. In 5 genetically related bloodstream isolates from a single patient, we noted a loss-of-function mutation in the capsule biosynthesis gene, wcaJ. Disruption of wcaJ was associated with decreased polysaccharide capsule, resistance to complement-mediated killing, and surprisingly, increased binding of complement proteins. Furthermore, an isogenic wcaJ deletion mutant exhibited increased opsonophagocytosis in vitro and impaired in vivo control in the lung after airspace macrophage depletion in mice.

Conclusions

Loss of function in wcaJ led to increased complement resistance, complement binding, and opsonophagocytosis, which may promote KPC-Kp persistence by enabling coexistence of increased bloodstream fitness and reduced tissue virulence.

Graphical Abstract

Open in new tabDownload slide

The antibiotic resistance profile of Klebsiella pneumoniae is a public health concern [1, 2]. Klebsiella pneumoniae carbapenemase–producing K pneumoniae (KPC-Kp) demonstrate extensive antimicrobial resistance and are associated with high mortality rates [3, 4]. However, KPC-Kp clinical strains, including the epidemic sequence type 258 (ST258), generally lack classical virulence determinants such as hypermucoviscous capsules belonging to the K1 and K2 serotypes. Furthermore, ST258 strains generally do not induce lethality in immunocompetent mice due to rapid clearance of the bacteria [5–9]. This paradox of KPC-Kp infection, where a relatively avirulent strain causes high mortality in hospital settings, suggests deficiencies within the host defense program that may provide fertile ground for evolving KPC pathogenicity.

A major mechanism of Kp pathogenic potential is the expression of a polysaccharide capsule that enables the bacteria to evade complement-mediated killing and phagocytosis [10] and increases Kp pathogenicity in mice and humans [10–13]. There are widely varying reports of the susceptibility of KPC-Kp strains to serum-mediated killing [6, 11, 12, 14]. We sought to determine the mechanism of resistance of a series of clinical KPC-Kp isolates to serum-mediated killing to better understand interactions between KPC-Kp and the host. We found a series of genetically related bloodstream isolates from a single patient with recurrent KPC-Kp bacteremia that culminated in a loss-of-function mutation in the wcaJ gene, which encodes an enzyme that attaches the first glucose-1-phosphate sugar of the exopolysaccharide capsule onto undecaprenyl phosphate in the cytoplasm near the inner membrane [15]. This wcaJ loss-of-function mutation conferred resistance to complement-mediated serum killing but also increased complement C3 binding and susceptibility to opsonophagocytosis.

METHODS

Bacteria

Clinical KPC-Kp isolates were obtained from 3 academic medical referral centers in urban regions in the United States: University of Pittsburgh, University of Wisconsin, and University of Miami (Florida). Carbapenem and colistin resistance were confirmed by Etest (bioMérieux) and broth microdilution, respectively. Clonal complex 258 was confirmed by multiplex polymerase chain reaction (PCR). Multilocus sequence typing of serial isolates was determined by in silico typing based on whole genome sequencing (WGS). The reference K2 serotype K pneumoniae, ATCC (American Type Culture Collection) 43816, was utilized in select experiments. De-identified clinical data on the reference patient was obtained through an honest broker with approval of the University of Pittsburgh Institutional Review Board (Study: PRO12060302).

Serum

Blood was obtained by venipuncture after informed consent (University of Pittsburgh STUDY19040346) from healthy volunteers, processed for serum, and stored at −80°C. For assays, serum was thawed at 37°C and used immediately. Factor-depleted sera were obtained from Complement Technology.

Serum and Polymyxin B Resistance Assays

Bacterial strains were grown to log phase in tryptic soy broth (TSB) at 37°C and 250 RPM as previously described [11, 12]. Bacteria were then incubated at 37°C with 5% sterile TSB and 85% (vol/vol) serum or 83 µM polymyxin B (100 µg/mL in sterile phosphate-buffered saline) for 4 hours. In select experiments, bacteria were grown in sera depleted of complement C3, with or without 1:1 mixing with healthy serum, for 4 hours as previously described [11]. Bacterial growth was quantified by colony-forming units (CFU) via serial plating.

Whole Genome Sequencing

Genomic DNA from serial bloodstream isolates was extracted using a DNeasy Blood and Tissue Kit (Qiagen, Germantown, Maryland) from 1-mL bacterial cultures grown in brain heart infusion (BHI) media as previously described [16]. Next-generation sequencing libraries were prepared with a Nextera XT kit (Illumina, San Diego, California), and libraries were sequenced on an Illumina MiSeq using 300-bp paired-end reads. Genomes were assembled with SPAdes v3.13.0, annotated with prokka version 1.14.5, and compared to one another with breseq version 0.33.2 [17]. The reference isolate KLP1 was also long-read sequenced on the Pacific Biosciences platform and used as a reference for identifying genetic variants in serial isolates from the same patient. Multilocus sequence types and antimicrobial resistance genes were identified using online tools available through the Center for Genomic Epidemiology (http://www.genomicepidemiology.org/). The reference genome used, as well as Illumina read data for serial isolates newly sequenced in this study, are available as National Center for Biotechnology Information Bioproject PRJNA1023934.

Biofilm Assay

Biofilm activity of clinical isolates was determined as previously described [18]. Please see the Supplementary Methods for additional details.

Osmotic Tolerance Assay

Serial KPC-Kp isolates were grown overnight in TSB then diluted 1:40 in TSB with 50% (volume/volume) sterile de-ionized water and incubated at 37°C. Optical density at 600 nm (OD₆₀₀) was serially measured and osmotic tolerance was quantified by the ratio of OD₆₀₀ at 4 hours to 0 hours.

Congo Red Agar Plate Assay

Overnight cultures of bacteria were streaked on BHI agar supplemented with 5% (weight/volume) sucrose and 0.08% (weight/volume) Congo Red and grown at room temperature for 48 hours as previously described [19].

Electron Microscopy

For transmission electron microscopy, bacteria grown overnight on agar were stained with ruthenium red [20]. Please see the Supplementary Methods for additional details on electron microscopy.

wcaJ Allelic Replacement

Approximately 500 base pairs of upstream and downstream sequences of the wcaJ gene were synthesized into the pUC57 vector from Genescript. This vector was digested with BamHI and HindIII. The approximate 1000-base pair fragment containing the flanking wcaJ sequences was gel extracted and cloned into the pMQ297 vector [21]. Transformants were selected in Luria-Bertani (LB) broth plus hygromycin (140 µg/mL). To delete the chromosomal copy of wcaJ in the KLP1 isolate, the pMQ297 vector containing the upstream and downstream regions of wcaJ was introduced into the KLP1 isolate by electroporation. Transformants were selected for hygromycin resistance in LB broth plus hygromycin (140 µg/mL) grown at 42°C overnight. This allowed for integration of the plasmid into the bacterial genome. Merodiploids were temperature shifted to 30°C to promote removal of the plasmid from the chromosome. Individual colonies were tested for loss of hygromycin resistance, and hygromycin-susceptible strains were assessed for deletion of wcaJ with PCR using forward (5′-GGTCTATGTCTTCGCTACTGC-3′) and reverse (5′-GCAGACAGCTCACGATTACG-3′) primers.

wcaJ Complementation

The capsule operon from KLP1 was cloned into the pBBR1-based cloning vector pMQ300 [21] using yeast in vivo recombination [22] and placed under control of the Escherichia coli lac promoter. The approximately 21.3 kb cps operon from galF through gnd was amplified in 2 pieces of overlapping amplicons with primers that also directed recombination with pMQ300 using PrimeSTAR GXL high-fidelity polymerase (Takara Bio). Primers to amplify the DNA were 5470 (5′-ttgtgagcggataacaatttcacacaggaaacagctGTGAAGATGAATATGGCGAATTTG-3′, lowercase letters direct recombination and uppercase priming) and 5429 (5′-gcatgatggaaggccctttcag-3′) for the galF containing amplicon and 5469 (5′-ccagtgccaagcttgcatgcctgcaggtcgactctagcaTTATTCCAGCCACTCGGTATG-3′) and 5474 (5′− GGTAATGATGCCAATTTGTTG-3′) for the gnd containing amplicon. The 2 amplicons and pMQ300 that had been linearized by SmaI (New England Biolabs) were used to transform Saccharomyces cerevisiae strain InvSc1 (Invitrogen). Uracil prototrophic transformants were pooled and plasmids were obtained and used to transform E coli strain S17-1 lambda pir [23]. Candidates were tested by PCR, and the chosen plasmid was validated by whole plasmid sequencing (SNPsaurus, LLC) and dubbed pMQ787. The plasmid pMQ787 and vector control pMQ300 were moved into KLP1 and KLP1ΔwcaJ by conjugation with selection on LB agar with hygromycin (140 µg/mL) and kanamycin (50 µg/mL).

Opsonized Phagocytosis Assay

Phagocytosis of bacteria, with or without serum opsonization, was performed in RAW 264.7 macrophage cells (ATCC TIB-71) as previously described and further detailed in the Supplementary Methods [24].

Mouse Infection Model

Wild-type (WT) C57BL/6J mice were purchased from Jackson Laboratory and maintained using a protocol approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Eight- to 16-week-old, age- and sex-matched mice were utilized in all experiments. Alveolar macrophages were depleted with intratracheal inoculation of 0.5 mg room temperature clodronate (Encapsula #CLD-8901) in 0.1 mL volume or equivalent volume of empty liposome control 24 hours prior to infection. After isoflurane anesthesia, mice were intratracheally inoculated with 10⁴ CFU of strain KLP1 (n = 4, all received liposome vehicle) or KLP1ΔwcaJ in 0.1 mL (n = 7 liposome vehicle, n = 6 clodronate) [25]. Necropsy was performed 24 hours postinfection with collection of bronchoalveolar lavage fluid (BALF) and CFU counts as previously described [11, 26, 27]. BALF cells were pelleted at 600g for 10 minutes at 4°C and the cell-free BALF supernatant was separated and stored at −80°C for subsequent assays including a modified immunoassay [28] to measure complement C3 deposition on bacteria by BALF, further described in the Supplementary Methods. BALF cells were processed for flow cytometry as described in the Supplementary Methods.

Statistical Analysis

Mann–Whitney test, or Kruskal–Wallis test with Dunn post hoc test for multiple comparisons when comparing >2 groups, was used unless otherwise indicated. P < .05 was considered significant. All statistics were performed using GraphPad Prism V9 software (GraphPad, La Jolla, California).

RESULTS

ST258 KPC-Kp Clinical Isolates Demonstrate Variable Growth in Serum

Although ST258 KPC-Kp clinical isolates are thought to be sensitive to serum killing [6, 12, 29], others have suggested that many KPC-Kp clinical isolates persist in healthy serum [14]. We screened 59 ST258 clinical isolates from 3 medical centers for growth in serum (Figure 1A). Forty-three of the 59 isolates (73%) demonstrated little or no growth, indicating serum sensitivity (Figure 1B, Supplementary Table 1). In contrast, 10 of 59 (17%) isolates demonstrated moderate growth, defined as between 2-fold to 5-fold increase in OD₆₀₀ at 4 hours compared to baseline. An additional 6 of 59 (10%) isolates demonstrated high growth in serum, with a >5-fold increase in OD₆₀₀. We identified serial KPC-Kp blood culture isolates with varying serum resistance from a single patient.

Patient Case History

The patient presented to the hospital in 2012 for abdominal pain and jaundice. Medical history was notable for a liver transplant 10 years prior. The patient had an extended hospitalization (115 days) that was notable for persistent biloma and recurrent bloodstream infections with KPC-Kp. The initial bloodstream KPC-Kp isolate was collected on hospital day 7. The patient was subsequently treated with intravenous colistin, which was continued through discharge, as well as doripenem and, later, gentamicin, which was the contemporaneous standard of care [30]. The patient died 12 days after discharge, although the cause of death was not available.

In Vivo Development of Serum and Polymyxin Resistance

We characterized serial KPC-Kp blood culture isolates from this patient by WGS, which revealed 5 genetically related isolates: KLP0, KLP1, KLP2, KLP6, and KLP7, all of which were meropenem resistant by clinical microbiology testing (Figure 2A). KLP1 was used as the reference isolate. All 5 isolates were predicted to belong to ST258 clade I and encoded KL106/wzi29-type capsule loci. We identified 3 mutations with functional consequences in KLP2, which was isolated on hospital day 12: (1) a T insertion into a polyT tract leading to a frame-shift mutation in the 3-oxosteroid 1-dehydrogenase gene ksdD (involved in lipid metabolism), (2) a 725-bp deletion (truncation) in the cadC transcriptional regulator gene of the cadAB operon [31], and (3) a nonsynonymous single-nucleotide polymorphism, S132P, in the peptidyl-dipeptidase dpc [32]. On hospital day 39, KLP6 was isolated and additionally contains an IS5 family transposase insertion disrupting mgrB, which is the most common cause of colistin resistance in K pneumoniae [33–36]. KLP7, which demonstrates a capsule locus disruption with an IS5 transposase insertion in the wcaJ gene, was isolated from the patient's bloodstream on hospital day 95.

To assess the functional consequences of the accumulated genetic mutations, we repeated serum resistance testing of the 5 serial isolates (Figure 2B). We found increased serum resistance in KLP7 compared to KLP1 as quantified by CFU after growth in serum for 4 hours, which was comparable to serum resistance of the ATCC 43816 reference strain [12, 37]. We further documented growth of KLP6 and KLP7 in 85% (v/v) polymyxin B (Figure 2C), which is consistent with disruption of mgrB in both isolates [33]. We then sought to understand how KLP7 escapes serum-mediated killing.

wcaJ Mutation Is Associated With Impaired Osmotic Tolerance and Capsule Alteration

We found significantly increased biofilm production in the KLP2 isolate (Figure 3A) using a crystal violet assay. KLP2 was isolated 8 days after placement of a central venous catheter. Notably, all venous and arterial catheters were later removed on hospital day 18 due to persistent K pneumoniae bacteremia and were not replaced until hospital day 23. We found significantly decreased growth of KLP7 compared to KLP1 in TSB with 50% vol/vol de-ionized water (Figure 3B), suggesting decreased osmotic tolerance with wcaJ mutation. We also noted gross differences when comparing KLP1 with the KLP6 and KLP7 isolates (Figure 3C). These gross changes were further explored by transmission electron microscopy with ruthenium red staining (Figure 3D), which demonstrated marked reduction in polysaccharide capsule by KLP7. Together, these data suggest that progressive genetic alterations in serial bloodstream isolates may have conferred serum resistance through capsular polysaccharide modifications.

wcaJ Mutant Clinical KPC-Kp Isolate Exhibits Resistance to Complement-Mediated Serum Killing and Increased Complement Binding

We next examined whether the wcaJ mutation in KLP7 conferred serum resistance through evasion of complement-mediated killing, which can occur with K pneumoniae capsule alterations. Normal healthy serum killed the KLP1 isolate (Figure 4A), which was impaired in serum depleted of complement component C3. Mixing healthy serum 1:1 with C3-depleted serum restored effective killing, confirming that complement mediates serum killing of KLP1. In contrast, the KLP7 isolate was resistant to serum killing in normal healthy serum, C3-depleted serum, and mixed serum. We hypothesized that KLP7 resists serum killing by limiting complement binding [38]. Surprisingly, we found significantly increased binding of complement C3 (Figure 4B) and membrane attack complex (Figure 4C) to the KLP7 isolate by direct enzyme-linked immunosorbent assay (ELISA), which was confirmed by flow cytometry (Figure 4D–E). Immuno-electron microscopy further validated increased complement C3 binding to KLP7 that was largely absent in KLP1 (Figure 4E). We sought to determine whether the wcaJ mutation conferred the surprising phenotype of increased complement binding and complement resistance.

Deletion of wcaJ Increases Complement Binding and Opsonophagocytosis

We deleted wcaJ in the KLP1 isolate background to generate an isogenic pair of wild-type and ΔwcaJ strains (Supplementary Figure 1). WGS confirmed that no other genes were disrupted. Scanning electron microscopy confirmed marked changes in capsule abundance and architecture in KLP1ΔwcaJ that resembled KLP7 (Figure 5A). KLP1ΔwcaJ was also resistant to serum killing in normal healthy serum, C3-depleted serum, and mixed serum (Figure 5B). We performed complementation of wcaJ into the KLP1 ΔwcaJ strain, which partially restored serum sensitivity (Figure 5C), confirming that deletion of wcaJ mediates serum resistance. If loss of wcaJ function yielded resistance to complement-mediated serum killing, we wondered why serum resistance did not lead to overwhelming infection in this patient. Given increased complement deposition with wcaJ mutation, we hypothesized that the KLP1ΔwcaJ mutant would be more susceptible to opsonophagocytosis. In the absence of preincubation with serum, there was no significant difference in phagocytosis of KLP1ΔwcaJ by RAW264.7 macrophages compared to KLP1 (Figure 5D). In contrast, opsonization with 20% serum increased phagocytic uptake of KLP1ΔwcaJ compared to opsonized KLP1 and nonopsonized KLP1ΔwcaJ. These results suggest that wcaJ loss-of-function mutation increases resistance to complement-mediated killing, but also increases complement binding and susceptibility to opsonophagocytosis, which we sought to confirm in vivo.

Disabling Opsonophagocytosis Through Depletion of Alveolar Macrophages Impairs Control of KPC-Kp ΔwcaJ in an Acute Lung Infection Model

We tested the in vivo impact of disabling opsonophagocytosis on KLP1ΔwcaJ by depleting alveolar macrophages in C57BL/6J mice via intratracheal administration of clodronate liposomes 24 hours prior to intratracheal infection and performed necropsy 24 hours postinfection (Figure 6A). Flow cytometry of BALF (Supplementary Figure 2) collected at necropsy confirmed a nearly 50-fold decrease in alveolar macrophages with clodronate (Figure 6B) but no significant difference in airspace neutrophils compared to liposome vehicle (Figure 6C). BALF from mice inoculated with KLP1ΔwcaJ was sufficient to enable C3 binding as determined by direct ELISA (Figure 6D), which supports the concept that pathogens can be opsonized in the alveolar space [39]. We observed increased BALF protein in mice treated with clodronate (Figure 6E), which is a nonspecific marker of increased lung microvascular permeability. Most importantly, we found that mice depleted of alveolar macrophages by clodronate exhibited impaired control of KLP1ΔwcaJ as determined by significantly increased CFU counts in lung homogenate (Figure 6F). These data suggest that opsonophagocytosis contributes to host defense against KPC-Kp and that wcaJ deletion may limit pathogenicity by increasing complement-mediated opsonophagocytosis.

DISCUSSION

We provide a unique assessment of host–pathogen interaction and bacterial evolution within a single critically ill patient. We note that the patient survived numerous episodes of KPC-Kp bloodstream infection, with an infecting strain of KPC-Kp evolving several mutations over time culminating in a loss-of-function wcaJ mutation that limited polysaccharide capsule biosynthesis and conferred resistance to complement-mediated serum killing. While this loss-of-function wcaJ mutation increased fitness of the bacterium in the bloodstream, the same mutation also increased complement deposition on the microbial surface, which increased susceptibility to opsonophagocytosis by the host. These findings suggest that the functional consequences of wcaJ mutation were both enhanced bloodstream fitness and limited tissue pathogenicity, which may ultimately promote bacterial persistence within the host.

Klebsiella pneumoniae extracellular capsule is a key virulence factor and modulation of polysaccharide capsule genes contributes to KPC-Kp evolution and fitness in the human host [5, 40, 41]. The wcaJ gene resides within the cps gene cluster, which is reviewed in [42]. Others have demonstrated that deletion of wcaJ in the K1 serotype SGH10 strain can promote gut persistence [41]. Furthermore, capsule-deficient KPC-Kp isolates exhibit increased persistence in the host with the urinary tract as a common reservoir [5]. Notably, insertion sequences conferring loss-of-function in wbaP, a homolog of wcaJ, were the most commonly identified genetic modifications [5]. Finally, genetic deletion of wcaJ from the K3 serotype reference strain ATCC 13883 increased biofilm formation and modulation of phagocytosis that may support persistence [43]. Despite similar phenotypes with the original isolate and wcaJ deletion mutant, including confirmation by allelic replacement, it remains possible that mutation of a single gene such as wcaJ in the capsule operon may have unknown effects such as production of an aberrant capsule. However, the potential clinical importance of wcaJ mutation is supported by a prior clinical report [44]. Notably, in vivo development of wcaJ mutation during 4.5 years of abdominal carriage of KPC-Kp was documented in a bloodstream isolate obtained during periods of cholangitis [44]. In fact, the wcaJ mutant was the only bloodstream isolate out of 18 clinical isolates [44]. We can only speculate as to the site of origin that led to KPC-Kp persistence in the patient in our report, but we note the presence of a biloma throughout their hospitalization that required a percutaneous transhepatic biliary drain. Future work should consider whether mutation of wcaJ impacts other K pneumoniae capsule types and whether complement resistance influences KPC-Kp pathogenesis and persistence at tissue sites other than the lung and blood.

Despite the complement resistance conferred by wcaJ loss of function, we were surprised to also find increased complement binding. Importantly, we note increased binding of the membrane attack complex (MAC), which can lyse gram-negative bacteria. Others have suggested that capsule-deficient K pneumoniae strains resist killing by sequestering complement deposition at a distance from the bacterial membrane to limit membrane damage [45]. MAC binding in the absence of bacterial lysis has also been demonstrated in extended-spectrum β-lactamase–producing K pneumoniae clinical bloodstream isolates, although the authors did not describe remote binding of MAC, suggesting other potential mechanisms of resistance to MAC lysis [46]. Notably, local action of complement C5 convertases at the bacterial membrane as well as direct anchoring of MAC precursors such as C5b–C7 may be crucial for execution of MAC pore formation and gram-negative bacterial lysis [47, 48], although this has not been investigated in the ST258 lineage. Finally, it is possible that increased complement binding may lead to saturation of early MAC proteins, leading to steric hindrance that impairs final assembly of the large multiprotein C9 complex β-barrel structured pore [49].

Although resistance to complement-mediated killing is thought to confer virulence in K pneumoniae, we noted limited tissue virulence with wcaJ loss of function. We propose that although the wcaJ mutation increased fitness in the bloodstream, the associated accentuation of complement binding also increased susceptibility to opsonophagocytosis and limited tissue virulence. This finding may provide another perspective to understand how interaction between the host and KPC-Kp strains determines their pathogenic potential. We have previously shown that functional deficiency of the alternative complement pathway can increase host susceptibility to a KPC-Kp clinical isolate in serum-killing assays and mouse models of disseminated pneumonia [11]. Yet, KPC-Kp isolates have a wide range of serum susceptibility and KPC-Kp infections are not uniformly fatal. In fact, the patient in our case was discharged to home 20 days after recovery of the complement-resistant wcaJ mutant isolate. Therefore, phagocytosis may serve as an additional layer of host defense to clear KPC-Kp bloodstream isolates that are complement-resistant, and supports the importance of both complement and phagocytosis in defending the lung against gram-negative pathogens [11, 13, 50].

In summary, we describe the in vivo development of complement resistance in a patient with recurrent KPC-Kp bloodstream infection via an IS5 transposon insertion conferring loss-of-function mutation in the wcaJ gene. Despite conferring resistance to complement-mediated serum killing, the wcaJ mutation also increased complement binding and susceptibility to opsonophagocytosis. We speculate that acquisition of these 2 traits led to persistence within the host by increasing bloodstream fitness and limiting tissue pathogenicity.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Disclaimer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, Department of Veterans Affairs, or any other sponsoring agency.

Financial support. This work was supported by the National Institutes of Health (award numbers T32 HL007563, F32 HL142172, and L30 HL143734 to W. B.; R38 HL150207 to N. T.; T32 AI138954 to N. C.; R01EY032517 to R. Q. S.; R01HL112937 to V. P. F.; R01AI104895 to Y. D.; and R01 HL136143, R01HL114453, R01 HL142084, and K24 HL143285 to J. S. L.); Career Development Award from the US Department of Veterans Affairs (award number IK2 BX004886 to W. B.); and the National Center for Advancing Translational Sciences of the National Institutes of Health (award number KL2TR001856 to A. I.). Electron microscopy at the University of Pittsburgh Center for Biologic Imaging was supported by National Institutes of Health Office of the Director (award numbers S10OD010625 and S10OD019973); the American Heart Association (Pre-Doctoral Fellowship 18PRE33960033 to T. F. O.); the Vascular Medicine Institute; the Hemophilia Center of Western Pennsylvania; and the Institute for Transfusion Medicine (Vitalant Foundation) (to W. B.).

References

Author notes