Heparin-based blood purification attenuates organ injury in baboons with Streptococcus pneumoniae pneumonia

Abstract

Bacterial pneumonia is a major cause of morbidity and mortality worldwide despite the use of antibiotics, and novel therapies are urgently needed. Building on previous work, we aimed to 1) develop a baboon model of severe pneumococcal pneumonia and sepsis with organ dysfunction and 2) test the safety and efficacy of a novel extracorporeal blood filter to remove proinflammatory molecules and improve organ function. After a dose-finding pilot study, 12 animals were inoculated with Streptococcus pneumoniae [5 × 10⁹ colony-forming units (CFU)], given ceftriaxone at 24 h after inoculation, and randomized to extracorporeal blood purification using a filter coated with surface-immobilized heparin sulfate (n = 6) or sham treatment (n = 6) for 4 h at 30 h after inoculation. For safety analysis, four uninfected animals also underwent purification. At 48 h, necropsy was performed. Inoculated animals developed severe pneumonia and septic shock. Compared with sham-treated animals, septic animals treated with purification displayed significantly less kidney injury, metabolic acidosis, hypoglycemia, and shock (P < 0.05). Purification blocked the rise in peripheral blood S. pneumoniae DNA, attenuated bronchoalveolar lavage (BAL) CCL4, CCL2, and IL-18 levels, and reduced renal oxidative injury and classical NLRP3 inflammasome activation. Purification was safe in both uninfected and infected animals and produced no adverse effects. We demonstrate that heparin-based blood purification significantly attenuates levels of circulating S. pneumoniae DNA and BAL cytokines and is renal protective in baboons with severe pneumococcal pneumonia and septic shock. Purification was associated with less severe acute kidney injury, metabolic derangements, and shock. These results support future clinical studies in critically ill septic patients.

INTRODUCTION

Sepsis due to bacterial pneumonia is a leading cause of death in the United States and worldwide, particularly in adults over 65 yr of age (1–3). The most frequently identified pathogen in community-acquired bacterial pneumonia is Streptococcus pneumoniae, a virulent organism responsible for high rates of septic shock, bacteremia, respiratory failure, and ∼18% mortality rate among hospitalized patients (3–6). Although sepsis mortality is improved by early initiation of antibiotics (7, 8), antibiotic eradication of the infection is often insufficient for clinical recovery, and bactericidal antibiotics may even worsen inflammation temporarily by lysing bacterial cells, releasing pathogen-associated molecular patterns (PAMPs), and exaggerating the systemic inflammatory response syndrome (SIRS) (9–14). Moreover, previous clinical studies targeting SIRS, such as treatment with monoclonal antibodies against tumor necrosis factor (TNF)-α or indiscriminate removal of circulating cytokines, have shown either no benefit or harm (15–17).

Extracorporeal blood purification has been developed as an adjunctive sepsis therapy. A variety of strategies have been proposed and tested, including immunoadsorption (18), opsonin capture (19), mass and polarity-based capture (20, 21), and heparin/charge-based capture (22, 23). Heparin sulfate is a cell surface glycosaminoglycan for which bacteria have innate affinity (24). Studies have shown that heparin-coated beads can bind and remove common bacterial pathogens from whole blood (22, 23) and remove circulating TNF-α from peripheral blood of septic patients (25). One such filter, the Seraph 100 Microbind Affinity Blood Filter, was CE marked in 2019 and its efficacy in sepsis reported in two case reports (26, 27) and an observational study of 15 bacteremic dialysis patients (28). Although these reports showed modest improvements in several clinical end points, there is a need for more definitive and well-controlled preclinical data to determine which molecules to target for removal and how purification affects the host response and organ injury before this therapy can be effectively translated to humans.

We tested a novel heparin-based blood-purifying device, the Seraph 200 Microbind Affinity Blood Filter, in a nonhuman primate model of severe pneumococcal pneumonia and sepsis. Like the Seraph 100, the Seraph 200 filter contains site-immobilized heparin moieties, but it also contains cationic binding sites that may bind toxins such as endotoxin or peptidoglycans. Based on our baboon model of S. pneumoniae pneumonia (29), we developed a model of severe pneumococcal sepsis with organ dysfunction by presenting an inoculum of 5 × 10⁹ colony-forming units (CFU) to the lungs and providing intensive support to the animal starting at 24 h. Our hypothesis was that purifying blood with an extracorporeal heparin-impregnated filter would attenuate systemic inflammatory markers and organ damage in pneumonia and acute lung injury.

MATERIALS AND METHODS

Dose-Ranging Pilot Study

The study was approved by the Duke University Institutional Animal Care and Use Committee (Protocol No. A203-17-08) and the U.S. Department of Defense Animal Care and Use Review Office (Protocol No. DARPA-7832). Animals were purchased from Texas Biomedical Research Institute (San Antonio, TX), housed in the Duke University Vivarium (Durham, NC), determined to be tuberculosis free (3 negative purified protein derivative tests), and handled in accordance with American Association for the Accreditation of Laboratory Animal Care guidelines. The protocol was refined to ensure careful attention to animal welfare, including 1) provision of enrichment; 2) social housing (except after inoculation); 3) free access to monkey chow (PMI Nutrition International, St. Louis, MO) and water (except fasting overnight before sedation); 4) provision of fruit; 5) frequent welfare checks; 6) use of topical analgesia for procedures (when appropriate); and 7) use of humane end points rather than mortality.

We first performed a dose-finding pilot study to identify an optimal pneumonia severity for further study. Seven juvenile male colony-bred baboons were sedated with ketamine (20–25 mg/kg; Fort Dodge Animal Health, Fort Dodge, IL) and diazepam (5 mg; Hospira Inc., Lake Forest, IL), intubated, and mechanically ventilated with volume-controlled ventilation (Puritan Bennett 840, Minneapolis, MN) with fraction of inspired oxygen ($\text{F}{\text{I}}_{{\text{O}}_2}$) of 0.21, tidal volume of 10 mL/kg, and positive end-expiratory pressure (PEEP) of 2.5 cmH₂O. Respiratory rate was set at 16 breaths/min and adjusted as needed to maintain arterial carbon dioxide tension ($\mathrm{P}{\mathrm{a}}_{\mathrm{C}{\mathrm{O}}_2}$) of 35–45 mmHg. Mean arterial pressure (MAP), heart rate (HR), rectal temperature, pulse oximetry, and telemetry were monitored throughout the experiment. Animals received 1 L of 0.9% saline intravenously to prevent volume depletion and a warming blanket to prevent hypothermia. A baseline bronchoalveolar lavage (BAL) was performed in the lingula with 20 mL of 0.9% saline using a fiberoptic bronchoscope (Pentax, Montvale, NJ) on $\text{F}{\text{I}}_{{\text{O}}_2}$ 1.0. After the BAL, animals were inoculated with S. pneumoniae (Serotype 19 A-7; ATCC, Manassas, VA) at 1 × 10⁸ CFU (n = 1), 5 × 10⁹ CFU (n = 3), and 1 × 10¹⁰ CFU (n = 3). The inoculum (20 mL) was divided into quartiles and instilled equally among the right lower lobe, right middle lobe, left lower lobe, and lingula by bronchoscopy. Supplemental oxygen was weaned back to 0.21 over 1 h. After 6 h, the animals were extubated and placed in isolation.

At 24 h, the animals were sedated, intubated, and ventilated as before. After collection of samples, 1 g of ceftriaxone (Hospira Inc., Lake Forest, IL) was administered intravenously once. Arterial and pulmonary artery catheters were inserted, and hemodynamic measurements were made every 6 h. For animals with hypotension (defined as mean arterial pressure < 65 mmHg), intravenous fluids (crystalloids) were administered as needed to maintain pulmonary capillary wedge pressure ≥12 cmH₂O. If mean arterial pressure remained <65 mmHg despite wedge pressure of 12 cmH₂O, dopamine (Hospira Inc.) was administered intravenously continuously and titrated every 5 min as needed to maintain mean arterial pressure ≥65 mmHg. At 48 h, animals were euthanized for necropsy. Criteria for early termination of the experiment (early euthanasia) were mean arterial pressure < 60 mmHg despite dopamine or arterial oxygen tension ($\mathrm{P}{\mathrm{a}}_{{\mathrm{O}}_2}$) < 55 mmHg despite $\text{F}{\text{I}}_{{\text{O}}_2}$ 50% and PEEP 10 cmH₂O.

Definitions

Animals were considered to have sepsis due to pneumonia if all three previously published criteria were met: 1) white blood cell (WBC) count >15,000/µL, <4,000/µL, greater than twofold change from baseline, or >90% neutrophilia; 2) isolation of S. pneumoniae from blood or BAL at 24 or 48 h; and 3) signs or symptoms of infection, including fever (>38.2°C), tachycardia (>100 beats/min), tachypnea (respiratory rate 25% above baseline), or infiltrate on chest radiograph (CXR) (29). Animals requiring dopamine to maintain mean arterial pressures ≥ 65 mmHg despite adequate preload (wedge pressure ≥12 cmH₂O) were considered to have septic shock. Acute lung injury (ALI) was defined as presence of histological tissue injury, protein leak (elevated bronchoalveolar lavage protein), acute inflammation (elevated bronchoalveolar lavage cell counts), and hypoxemia (30).

Extracorporeal Blood Purification

To test the safety and efficacy of a heparin-based extracorporeal blood purification device as a sepsis therapy, 16 baboons were allocated randomly to one of the following three experimental groups: S. pneumoniae/Purification (n = 6); S. pneumoniae/Sham (n = 6); and Uninfected/Purification (n = 4). These animals were inoculated with S. pneumoniae (5 × 10⁹ CFU) or vehicle and supported as before (Fig. 1).

Figure 1.

Experimental timeline. Schematic showing timeline (hours) of a single animal experiment, including time of inoculation (time 0); administration of ceftriaxone (at 24 h); necropsy (at 48 h); sample collections (blood, urine) at 0, 6, 24, 30, 36, 42, and 48 h; bronchoscopy and bronchoalveolar lavage (BAL) at 0, 24, and 48 h; timing and duration of mechanical ventilation (gray rectangles); and timing and duration of blood purification (BP, black bar).

The extracorporeal purification device was set up as follows: An acute care dialysis cartridge (CAR-125-B; NxStage Medical, Inc., Lawrence, MA) and an HF400 Renaflo filter (Medivators Inc., Minneapolis, MN) were connected to a modified NxStage System One continuous dialysis cycler (NxStage Medical, Inc.) and primed with 0.9% saline according to manufacturer instructions. For the Purification groups, the Seraph 200 Microbind Affinity Blood Filter (ExThera Medical, Martinez, CA) was primed with 2 L of 0.9% saline. After priming, the arterial and venous lines were disconnected from the Renaflo filter and reconnected to the experimental filter. The effluent (waste) line and the dialysate line remained connected to the HF400 filter, and rates for each were set to 0 mL/min.

At 30 h after inoculation, systemic heparin was administered (200 U/kg iv bolus, then 100 U/kg/h iv infusion), and the animals were connected to the continuous dialysis cycler via a double-lumen femoral vein dialysis catheter (Fig. 2A). Animals were treated for 4 h at a blood flow rate of 150 mL/min with the Seraph 200 Microbind Affinity Blood Filter (ExThera Medical, Martinez, CA) or a sham filter (Fig. 2B). The sham filter was hollow tubing designed to simulate the drop-in pressure across the arterial to venous limbs. Point-of-care activated clotting times (ACTs) were measured at baseline and every hour (Hemochron, Jr.; Instrumentation Laboratories), and heparin infusion was adjusted to keep ACTs between 280 and 400 s. After 4 h of hemofiltration, blood was returned to the animal, the dialysis circuit was disconnected, and heparin was stopped.

Figure 2.

A: extracorporeal blood purification setup. Schematic showing a simplified purification setup. Venous blood is drained from femoral venous dialysis catheter (red lines), through dialysis machine at 150 mL/min and through filter, and then returned to machine (blue line) and then animal via dialysis catheter. Dialysate (green triangles) and ultrafiltration (yellow triangles) rates were set to 0 and did not contribute to the treatment. B: pictures of filters. Top: Seraph 200 Microbind Affinity Blood Filter (ExThera Medical). Higher-resolution inset shows individual beads. Bottom: sham filter (hollow tubing) that simulates drop in pressure from venous to arterial limbs.

In an effort to conserve baboons and reduce costs without loss of statistical power, the S. pneumoniae/Sham group of six animals includes two animals in the dose-finding study that were inoculated with S. pneumoniae and heparinized for 4 h at 30 h after inoculation but did not receive sham purification. This decision was made before data analysis. No differences between these control subgroups were noted during the experiment.

Monitoring and Sample Collection

Respiratory rate (RR), the presence of cough and rhinorrhea, oral intake, and activity level were recorded before sedation in all animals at 0, 12, 15, 18, 21, and 24 h. Blood and urine samples were collected at 0, 6, 24, 30, 36, and 48 h. Urine output was measured every 6 h. Ventrodorsal chest radiographs (CXRs) were obtained at 0, 24, and 48 h. Ventilator physiology (peak inspiratory pressure, plateau pressure, and compliance) was measured every 6 h.

Microbiology

S. pneumoniae was cultured, counted, and prepared in 20 mL of 0.9% saline to achieve the final dose for instillation. The number of organisms was confirmed by serial dilutions both before and after instillation to verify preinoculation viability. Immediate postinoculation colony counts confirmed that actual doses administered were at goal.

Blood (24, 30, and 48 h) and bronchoalveolar lavage fluid (BALF) (0, 24, and 48 h) samples collected during the experiments were cultured for bacteria. Urine was tested for pneumococcal urinary antigen (BinaxNow S. pneumoniae; Inverness Medical, Cranfield, UK), and aliquots were banked at −80°C for future studies.

S. pneumoniae DNA was measured by polymerase chain reaction (PCR). Total DNA was extracted from whole blood with the DNeasy Blood and Tissue Kit (Qiagen). DNA purity was ascertained by spectrophotometry (A₂₆₀/A₂₈₀). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed on an ABI StepOnePlus using gene expression assay primers for S. pneumoniae DNA (Ba04230899 and Ba04230900; Applied Biosystems, Waltham, MA). 18S rRNA was used as an internal control. Quantification of gene expression was determined by the comparative threshold cycle and relative quantification method. Each sample was assayed in triplicate.

Laboratory Measurements

Complete blood counts and coagulation studies were performed at 0, 6, 24, 30, 36, and 48 h, and chemistries were performed at 0, 24, and 48 h. Arterial blood was collected in heparin for blood gas analysis (Instrumentation Laboratories model 682, Bedford, MA) at 0, 6, 24, 30, 36, 42, and 48 h. BALF cells were counted on a hemocytometer and concentrated onto a microscope slide (StatSpin Cytofuge; IRIS International, Westwood, MA) for Gram stain and differential. Whole BALF samples were stored at −80°C for quantitative PCR, and spun supernatant was assayed for protein and lactate dehydrogenase (LDH), as previously described (31). Myeloperoxidase (MPO) activity was measured in lung, heart, kidney, liver, and jejunum homogenates as previously described (31) and expressed as change in absorbance per minute per gram (wet wt) of tissue.

Lactate measurements were performed as follows: Plasma was isolated from peripheral blood (acid-citrate-dextrose anticoagulant) after Ficoll density centrifugation and was clarified immediately by centrifugation at 1,200 g for 10 min at 8°C. After clarification, plasma was immediately stored at −80°C until the time of testing. Plasma lactate determinations were made with the MxP Quant 500 kit (Biocrates AG, Innsbruck, Austria) in strict accordance with their detailed protocol. Kits are supplied with internal standard applied to each well of the 96-well extraction plate. Ten microliters of each blank, calibration standard, Biocrates QC, Global Reference QC, plasma sample, and SPQC were added to the appropriate wells. The plate was then dried under a gentle stream of nitrogen for 10 min. The samples were derivatized with phenyl isothiocyanate and then eluted with 5 mM ammonium acetate in methanol. Samples were diluted with either 1:1 methanol-water for the UPLC analysis (1:1) or running solvent (a proprietary mixture provided by Biocrates) for flow injection analysis (50:1). Ultrahigh-performance liquid chromatography (UHPLC) separation was performed with an Exion AD liquid chromatograph (Sciex, Framingham, MA) with a proprietary analytical column and guard column provided with the MxP Quant 500 kit (Biocrates AG, Innsbruck, Austria). Analytes were separated with a gradient from 0.2% formic acid in water to 0.2% formic acid in acetonitrile. Total UPLC analysis time was ∼6 min per sample. With two injections, one in electrospray ionization positive mode (ESI⁺) and one in negative mode (ESI⁻), samples for UPLC analysis were introduced directly into a QTRAP 6500⁺ (Sciex) operating in the Multiple Reaction Monitoring (MRM) mode. MRM transitions (compound-specific precursor to product ion transitions) for each analyte and internal standard were collected over the appropriate retention time. The UPLC-tandem mass spectrometry (MS/MS) data were imported into Sciex application MultiQuant for peak integration, calibration, and concentration calculations. Acylcarnitines, monosaccharides (hexose), diglycerides, triglycerides, lysophosphatidylcholines, phosphatidylcholines, sphingomyelins, ceramides, and cholesteryl esters were analyzed in three separate flow injection analysis-tandem mass spectrometry (FIA-MS/MS) methods with total analysis time of ∼3.8 min per injection using electrospray ionization in positive mode (ESI⁺) on a Xevo TQ-S triple quadrupole mass spectrometer (Waters). Sample introduction was performed with Acquity UPLC (Waters). Compound-specific precursor-to-product ion transitions for each analyte and internal standard were collected. The FIA-MS/MS data were analyzed with Biocrates MetIDQ software.

Cytokine and Chemokine Measurements

Cytokines and chemokines [CCL2, CCL3, CCL4, CXCL8, interleukin (IL)-1β, IL-6, and IL-10] were measured in plasma (0, 24, 30, 36, and 48 h after inoculation) and BALF supernatant samples (0, 24, and 48 h after inoculation) in duplicate with a V-PLEX Custom Assay for nonhuman primates (Meso Scale Discovery, Rockville, MD). IL-18 was measured in the same samples with a Meso Scale Discovery U-PLEX assay. For samples returning cytokine/chemokine values below the lower limit of quantification (LLOQ), ½ LLOQ was used for the statistical analyses. Data for TNF-α were considered unreliable and not reported.

Tissue Collection and Preparation

At 48 h, the animals were deeply sedated with ketamine and diazepam and euthanized with 20 mL of intravenous saturated KCl solution. Immediate necropsy was performed to harvest and prepare the tissues. After the thorax was opened, the left bronchus was ligated and the left lung was removed. Ex vivo BAL of the left lower lobe was performed with 60 mL of ice-cold 0.9% NaCl. Lung tissue samples from grossly involved (e.g., lower lobes) and uninvolved (e.g., upper lobes) sites along with heart, kidney, adrenal, diaphragm, skeletal muscle, liver, and spleen tissue samples were fixed with 10% formalin (Azer Scientific, Morgantown, PA) and snap frozen and stored at −80°C. Formalin-fixed tissues were embedded in paraffin, sectioned in 1-mm slices, and stained with hematoxylin and eosin (H&E) unless otherwise specified. Three randomly selected slides from sites of involved and uninvolved lung tissue of each animal were evaluated under light microscopy by a blinded pathologist. Five intra-alveolar findings (edema, leukocytes, alveolar filling, fibrin, and necrosis) were scored 0 (absent) to 3 (severe) for each slide to generate the modified lung injury score, as described previously (32).

Protein Studies

Protein was extracted from kidney tissue and quantified with the bicinchoninic acid (BCA) method. Protein samples (5 µg) were pipetted onto polyvinylidene difluoride membranes (Millipore), dried for 15 min, blocked with 5% nonfat dry milk in Tris-buffered saline with Tween 20 (TBST) for 10 min, washed with TBST for 5 min, and probed overnight at 4°C with primary antibodies against neutrophil gelatinase-associated lipocalin (NGAL), cleaved caspase-1, cleaved interleukin (IL)-1β, IL-18, IL-10, NLRP3, and tubulin (to confirm protein loading). Membranes were washed with TBST (3×), treated with the appropriate secondary antibody conjugated to horseradish peroxidase (60 min), and developed with enhanced chemiluminescence (Luminol, Santa Cruz Biotechnology). Bands were digitized and protein density quantified in the middynamic range (Quantity One, Bio-Rad).

Immunofluorescence Studies

Formalin-fixed kidney samples were paraffin embedded, cut into 5-μm sections, and mounted on slides. Immunofluorescence staining was performed for mitochondrial citrate synthase (green) and 8-hydroxydeoxyguanosine (8-OHdG) (red) at 1:200 dilution. Nuclei were stained with DAPI (Molecular Probes). Alexa Fluor-coupled secondary antibody (Invitrogen) conjugation was performed at 1:400 dilution. Stained sections were imaged on a Nikon Eclipse 50i fluorescence microscope with a Nikon digital camera attachment.

Statistical Analysis

Grouped data are expressed as means ± SE unless otherwise specified. All statistical analysis planning is a priori. The primary outcome was change in pathogen-associated molecular targets. We calculated n = 6 per group to achieve 80% power at α of 0.05. Serial time points are analyzed by one-way or two-way repeated-measures ANOVA with Fisher’s least significant difference post hoc testing (Prism; GraphPad Software, La Jolla, CA). Protein measurements were analyzed with Student’s t test. P < 0.05 is considered statistically significant. Trends are noted for P < 0.1.

RESULTS

Dose-Ranging Pilot Study

The baseline characteristics and physiological data for all experimental animals are shown in Table 1. The dose-ranging pilot study was used to determine the optimal dose of S. pneumoniae that produces severe pneumococcal pneumonia and organ dysfunction without excess mortality. Animals were given escalating doses of S. pneumoniae ranging from 1 × 10⁸ to 1 × 10¹⁰ CFU. All animals met established criteria for pneumonia (29), exhibiting a change in WBC count (Supplemental Fig. S1A; see https://doi.org/10.6084/m9.figshare.12678560), isolation of S. pneumoniae in blood and/or bronchoalveolar lavage fluid (BALF) (Supplemental Fig. S1, B and C), and signs or symptoms of pneumonia, including fever or hypothermia, tachycardia, tachypnea, and/or infiltrates on chest radiograph (Supplemental Fig. S2; see https://doi.org/10.6084/m9.figshare.12678656). Animals displayed signs of acute organ failure and metabolic derangements, with metabolic acidosis (Supplemental Fig. S3, A–C; see https://doi.org/10.6084/m9.figshare.12678674), hypoglycemia, acute kidney injury, hypoxemia, and septic shock requiring dopamine (Supplemental Fig. S4; see https://doi.org/10.6084/m9.figshare.12678707). Septic shock in the 10¹⁰ CFU group was particularly refractory, leading to early euthanasia in two of the three 10¹⁰ CFU animals.

Table 1.

Baseline characteristics and physiology

	Animal No.	Age, yr	Sex	Weight, kg	RR	Temp, °C	SBP	DBP	HR, beats/min	Group
1	29564	7.5	Male	28.0	24	36.9	112	60	87	Dose finding
2	31202	5.0	Male	22.2	12	36.5	94	61	96	Dose finding
3	31210	5.0	Male	22.6	21	37.0	104	58	81	Dose finding
4	31032	5.4	Male	17.5	26	37.6	98	41	102	Dose finding
5	31022	5.5	Male	17.9	21	37.3	85	50	120	Dose finding
6	31207	5.2	Male	21.2	25	37.4	115	57	102	Dose finding and S. pneumoniae/Sham
7	31785	4.9	Male	24.5	17	37.4	113	64	90	Dose finding and S. pneumoniae/Sham
8	32307	4.9	Male	16.5	30	36.6	85	50	104	S. pneumoniae/Sham
9	31033	5.7	Male	20.4	40	37.4	93	49	91	S. pneumoniae/Sham
10	31065	5.9	Male	26.1	28	38.0	126	93	95	S. pneumoniae/Sham
11	31846	5.2	Male	22.0	28	38.3	84	41	104	S. pneumoniae/Sham
12	31135	5.6	Male	25.5	24	36.8	88	41	80	S. pneumoniae/Purification
13	31553	5.4	Male	22.2	24	37.4	107	47	77	S. pneumoniae/Purification
14	31113	5.6	Male	23.6	28	37.3	78	44	91	S. pneumoniae/Purification
15	31836	5.3	Male	19.6	20	36.9	95	62	81	S. pneumoniae/Purification
16	32539	4.5	Male	19.1	18	36.8	113	74	118	S. pneumoniae/Purification
17	31805	5.2	Male	22.7	21	37.7	96	48	89	S. pneumoniae/Purification
18	32552	4.6	Male	17.3	25	36.8	95	52	101	Uninfected/Purification
19	32298	5.0	Male	21.1	35	37.1	94	48	103	Uninfected/Purification
20	32215	5.2	Male	21.6	16	37.1	112	50	73	Uninfected/Purification
21	31322	5.1	Male	21.2	34	37.0	77	39	75	Uninfected/Purification

Two of the animals (31207 and 31785) were used for both the dose-finding study and the S. pneumoniae/Sham experiments. They were given heparin infusions for 4 h at the same time as the other S. pneumoniae/Sham animals but did not receive sham purification. DBP, diastolic blood pressure; HR, heart rate; RR, respiratory rate; SBP, systolic blood pressure; Temp, temperature.

At necropsy, the lungs displayed gross hepatization (Supplemental Fig. S5A; see https://doi.org/10.6084/m9.figshare.12678719). To understand how inoculum affected ALI severity, we performed blinded histopathological ALI scoring. Randomly selected lung tissue blocks were scored 0 to 3+ for inflammation, alveolar edema, consolidation, necrosis, and fibrin deposition (29). Lung histopathology showed alveolar filling by neutrophils, edema, and fibrin (Supplemental Fig. S5, C–E). The difference between 10⁹ and 10¹⁰ CFU was not statistically significant (Supplemental Fig. S5F).

To determine the time course and kinetics of PAMP release during pneumococcal infection, we measured S. pneumoniae DNA in peripheral whole blood. Despite sterilization of blood cultures 30 h after inoculation (6 h after antibiotics) (Supplemental Fig. S1C), S. pneumoniae DNA remained detectable in peripheral blood and peaked at 36 h after inoculation (12 h after antibiotics) in the 10⁸ CFU and 10⁹ CFU animals and was still rising at necropsy in the single surviving 10¹⁰ CFU animal (Supplemental Fig. S6; see https://doi.org/10.6084/m9.figshare.12678722).

Extracorporeal Blood Purification

Based on the pilot study, we chose 5 × 10⁹ CFU as the optimal inoculum that produced severe S. pneumoniae pneumonia and sepsis without causing experiment-limiting mortality. Twelve animals were inoculated with 5 × 10⁹ CFU of S. pneumoniae and randomized to either blood purification or sham therapy, delivered at 30 h after inoculation for 4 h. Both groups developed comparable tachycardia, tachypnea, and fever (Fig. 3, A–C). However, shock (dopamine requirements) increased significantly in the Sham group (P < 0.05) but not in the Purification group (Fig. 3D). There was also a statistical trend for fewer animals with shock at necropsy in the Purification group (P = 0.079) and significantly more dopamine-hours in the Sham group, but not the Purification group, compared to the uninfected control animals (Supplemental Fig. S7; see https://doi.org/10.6084/m9.figshare.14417558). One animal in the Sham group met euthanasia criteria before 48 h because of refractory shock, whereas all animals in the Purification group survived to 48 h. Fluid administration was nearly identical between the Sham and Purification groups (Supplemental Fig. S8; see https://doi.org/10.6084/m9.figshare.12678728), and there were no statistically significant differences in urine output or 24-h fluid balance.

Figure 3.

Blood purification vs. sham: hemodynamics. A–C: heart rate (A), respiratory rate (B), and temperature (C) are shown over time for each animal group. The gray bars in A and C reflect the duration of blood purification therapy. D: peak dopamine doses for each S. pneumoniae-infected animal shown before and after the sham or purification intervention. There was a significant increase in dopamine dose in the sham-treated (pre- vs. posttreatment) but not in the purification-treated animals. *P < 0.05 by repeated-measures 2-way ANOVA with Fisher’s least significant difference post hoc test. Error bars show standard error.

Inoculated animals in both groups developed lobar consolidation on chest radiograph by 24 h, and there was no significant difference in radiographs between the two groups at 48 h. Moreover, ventilator characteristics such as compliance and PEEP were not different between groups. Histopathological examination of inoculated lobes showed neutrophils, edema, and fibrin in the alveolar spaces without necrosis. Acute lung injury scores were not significantly different between the Sham and Purification groups (7.3 ± 0.73 vs. 6.3 ± 0.66, respectively; P = 0.31) (Fig. 4A). However, myeloperoxidase activity, a measure of neutrophilic infiltration, was significantly higher in the Purification group compared with Sham (85.4 ± 16.8 vs. 49.8 ± 15.7; P < 0.001) (Fig. 4B).

Figure 4.

Blood purification vs. sham: pulmonary pathology. A: blinded lung histopathology scores for each animal on lung tissue taken at necropsy. B: myeloperoxidase enzyme activity measured in each animal for the tissues listed plotted on a logarithmic scale. OD, optical density. Values of 0.1 reflect no measurable activity. B shows a significant difference in between S. pneumoniae/Sham and S. pneumoniae/Purification groups. C–F: bronchoalveolar lavage fluid (BALF) lactate dehydrogenase (LDH) activity (C), total protein concentration (D), total cell counts (E), and cell differentials (F) of each animal over time. C and E are plotted on logarithmic scales. F shows a significant difference in % neutrophils and % lymphocytes in BALF of S. pneumoniae/Sham animals vs. S. pneumoniae/Purification animals. G: arterial oxygen tension ($\mathrm{P}{\mathrm{a}}_{{\mathrm{O}}_2}$)-to-fraction of inspired oxygen ($\text{F}{\text{I}}_{{\text{O}}_2}$) ratios are shown at 0 h (baseline), 24 h after inoculation, and necropsy. *P < 0.05 by 2-way ANOVA with Fisher’s least significant difference post hoc test. Black horizontal lines represent group means.

Bronchoalveolar lavage fluid (BALF) studies including lactate dehydrogenase activity, total protein concentration, and total cell counts were similar between the Sham and Purification groups (Fig. 4, C–E). However, the percentage of BALF neutrophils was significantly higher and the percentage of lymphocytes was significantly lower in the hemofiltration group (both P < 0.05) (Fig. 4F). $\mathrm{P}{\mathrm{a}}_{{\mathrm{O}}_2}$-to-$\text{F}{\text{I}}_{{\text{O}}_2}$ ratios at 24 h after inoculation and at necropsy were also similar between the Sham and Purification groups (Fig. 4G).

All inoculated animals developed leukocytosis at 24 h; however, WBC counts, platelet counts, and coagulation profiles were not significantly different between the two infected groups. Animals in the Sham group, however, did develop significantly more hypoglycemia (serum glucose 42 ± 9 vs. 68 ± 11 mg/dL; P < 0.01) (Fig. 5A), acute kidney injury (creatinine 1.7 ± 0.32 vs. 1.25 ± 0.17 mg/dL; P < 0.05) (Fig. 5B), and acidosis (pH 7.25 ± 0.06 vs. 7.35 ± 0.03; P < 0.05) (Fig. 5C) compared with animals in the Purification group. The acidosis was metabolic, as we adjusted the respiratory rates to maintain $\mathrm{P}{\mathrm{a}}_{\mathrm{C}{\mathrm{O}}_2}$ levels within a certain range (materials and methods and Fig. 5D), and the bicarbonate levels were significantly lower in the Sham filtration group compared with the Purification group (Fig. 5E). Furthermore, the plasma lactate levels were significantly higher in the Sham group but not different between the infected and the uninfected Purification groups (Fig. 5F).

Figure 5.

Purification vs. sham: laboratory tests. Laboratory tests shown for each animal over time. A shows statistically significant differences in glucose levels between all 3 groups. B, C, and E show significant differences in creatinine, pH, and bicarbonate levels, respectively, between S. pneumoniae/Sham and S. pneumoniae/Purification groups. D: arterial carbon dioxide tension ($\mathrm{P}{\mathrm{a}}_{\mathrm{C}{\mathrm{O}}_2}$). F shows that lactate is significantly higher in the S. pneumoniae/Sham group vs. the Uninfected/Purification group but not in the S. pneumoniae/Purification group. The gray bars reflect the duration of the purification treatment. Error bars show standard error. *P < 0.05 by 2-way ANOVA with Fisher’s least significant difference post hoc test.

S. pneumoniae was isolated from BALF at 24 and 48 h, but colony counts were not significantly different between the two groups (Fig. 6A). Peripheral blood S. pneumoniae DNA was detectable via PCR at 24, 30, 36, and 48 h. Compared with sham treatment, blood purification significantly attenuated the postantibiotic rise in S. pneumoniae antigen at 36 h after inoculation (2.80-fold vs. 0.74-fold change; P < 0.05) (Fig. 6B). Animals in both groups were bacteremic at 24 h, but by necropsy blood had sterilized (Fig. 6C). Blood and BALF cultures in the uninfected control animals were negative for all time points.

Figure 6.

Purification vs. sham: microbiology. A: bronchoalveolar cultures shown for each animal at each time point plotted on a logarithmic scale. BALF, bronchoalveolar lavage fluid; CFU, colony-forming units. Black horizontal lines represent group means. B: S. pneumoniae DNA (fold change) measured by qPCR over time. S. pneumoniae DNA is significantly higher at 36 h in the S. pneumoniae/Sham group compared with the S. pneumoniae/Purification group. C: blood cultures shown for each animal at each time point plotted on a logarithmic scale. Black horizontal lines represent group means. Binary culture results (positive or negative) are shown at bottom of A and C to denote cultures performed in each group that were negative. The gray bar reflects the duration of purification therapy. Error bars show standard error. P = 0.01 calculated by 2-way ANOVA with Fisher’s least significant difference post hoc test.

Selected cytokines and chemokines in plasma and BALF samples were measured from each group at 0, 24, and 48 h. Blood purification did not significantly alter plasma cytokine or chemokine levels (Fig. 7). However, compared with sham treatment, S. pneumoniae-infected animals treated with purification displayed significantly reduced BALF CCL4 (16,612 ± 9,871 vs. 39,434 ± 6,718 pg/mL; P < 0.05), BALF CCL2 (16,664 ± 8,280 vs. 40,860 ± 6,538 pg/mL; P < 0.01), and BALF interleukin-18 (25.4 ± 9 vs. 63.8 ± 16.2 pg/mL; P < 0.05) (Fig. 8).

Figure 7.

Purification vs. sham: plasma cytokines. Median cytokine and chemokine levels measured in each animal over time plotted on logarithmic scales. Error bars show interquartile range.

Figure 8.

Purification vs. sham: bronchoalveolar fluid cytokines. Cytokine and chemokine levels measured in each animal over time plotted on logarithmic scales. There were statistically significant differences in CCL2, CCL4, and IL-18 levels in bronchoalveolar lavage fluid of S. pneumoniae/Sham vs. S. pneumoniae/Purification groups. Error bars show standard error. *P < 0.05 by 2-way ANOVA with Fisher’s least significant difference post hoc test.

To further investigate a potential mechanism for the observed mitigation in acute kidney injury in the blood purification group, we first measured renal expression of neutrophil gelatinase-associated lipocalin (NGAL), a renal tubular protein and biomarker of acute kidney injury (33), and found that renal NGAL was significantly higher in the Sham group compared with the Purification group (Supplemental Fig. S9A; see https://doi.org/10.6084/m9.figshare.14418479). We next measured renal protein markers of classical NLRP3 inflammasome activation, including cleaved caspase-1, IL-18, cleaved IL-1β, and NLRP3, and found that these were significantly reduced in the Purification group compared with Sham (Supplemental Fig. S9, B–E). We also found that renal IL-10, a counterregulatory cytokine, was significantly elevated in the Purification group compared with Sham (Supplemental Fig. S9F). Taken together, these data support the renal protection seen with blood purification and link it to mitigation of renal inflammasome activation.

Because mitochondrial injury is evident in sepsis-induced organ failure (33–35), we next performed immunofluorescence microscopy on formalin-fixed kidney tissues for mitochondrial citrate synthase (Supplemental Fig. S10; see https://doi.org/10.6084/m9.figshare.14418482). There was increased distribution of citrate synthase (green) in the Purification group compared with Sham, indicative of mitochondrial injury and loss in the sham-treated animals. Staining for 8-hydroxydeoxyguanosine (8-OHdG), a marker of oxidative DNA damage, showed increased distribution in the Sham group compared with the Purification group. Taken together, blood purification was associated with reduced renal mitochondrial injury and oxidative DNA damage.

To determine the safety of blood purification therapy, we treated four uninfected animals (inoculated with sterile 0.9% NaCl solution) with the heparin-based filter at the same time points and measured physiology, laboratory, and organ injury end points. These animals maintained normal vital signs during and after purification (Fig. 3). Chest radiographs were unremarkable, as were oxygenation, lung histopathology, myeloperoxidase activity, and BALF studies (Fig. 4). Aside from a transient leukocytosis seen at 6 h (before purification), hematologic studies were unremarkable. Creatinine, glucose, and pH were comparable between uninfected animals and the infected animals that underwent the purification procedure (Fig. 5). All of the measured BALF chemokines/cytokines, except IL-10, increased slightly above baseline by 24 h after postinoculation, before extracorporeal treatment, likely reflecting host response to intubation and mechanical ventilation (Fig. 8).

DISCUSSION

To test the efficacy and safety of a novel blood purification therapy, a novel nonhuman primate model of severe pneumococcal sepsis with organ dysfunction that closely mimics the human disease was established. In this model, the severity of pneumonia-induced acute lung injury was similar in treated and untreated animals, but blood purification with a heparin-based filter significantly improved the systemic impacts of pneumonia, attenuating hypoglycemia, acute kidney injury, metabolic acidosis, and shock, the cause of early death in animals inoculated with the highest pathogen dose. These improvements were associated with lower circulating S. pneumoniae DNA, lower BALF cytokines, and attenuated renal inflammasome activation and oxidative damage. We also found that purification using a novel heparin-based filter is safe in both uninfected control animals and animals with sepsis, suggesting that it is a promising approach to treating septic patients.

We chose nonhuman primates for this study for several reasons. First, unlike rodents, baboons display hemodynamic and immunologic host responses to sepsis very similar to humans (29, 36). Second, their larger size allows for invasive monitoring and serial blood, urine, and respiratory sampling. Third, baboons have pulmonary anatomy (37) and upright posture similar to humans and develop lobar pneumonia with dependent alveolar edema that closely mirrors human disease (29). These factors increase the likelihood that a beneficial therapy is successfully developed and a harmful one is abandoned (38).

Building on a standard nonhuman primate pneumonia model (29), we performed a dose-titration study to develop a novel model of severe S. pneumoniae pneumonia and sepsis with local (lung) and systemic (e.g., kidney) organ dysfunction. This model closely mirrors the human host response to sepsis and incorporates standard-of-care treatments such as antibiotics, fluids, and intensive care support, and is therefore highly clinically relevant (38). Infection severity ranged from mild illness at 10⁸ CFU to cardiopulmonary collapse at 10¹⁰ CFU. Refractory septic shock in these animals was associated with hypothermia, leukopenia, metabolic acidosis, and profound hypoglycemia, features common in fatal human sepsis (39–43). However, because of the experiment-limiting mortality at 10¹⁰ CFU, we chose the 5 × 10⁹ CFU dose for further study.

We found a significant rise in S. pneumoniae DNA peaking 12 h after ceftriaxone administration. Beta-lactams are bactericidal antibiotics that disrupt bacterial cell wall integrity and cause release of cell wall fragments, such as endotoxin or lipoteichoic acid, into the circulation (14), which can rise severalfold (9–11, 14, 44–47). Such PAMP release further activates the inflammatory cascade (12–14), is associated with more cytokine production, lactic acidosis, and shock (9, 10), and is a plausible mechanism for the shock and acute organ dysfunction observed in our study. In fact, these findings raise questions about antibiotic selection in sepsis, such as whether initial treatment with bacteriostatic rather than bactericidal antibiotics would cause less PAMP release.

Although prior efforts to attenuate host inflammatory responses including cytokine inactivation or removal have been ineffective or harmful (15–17), newer blood purification technologies have been developed that can bind and remove circulating host- and pathogen-derived proinflammatory products (18, 19, 21–23). However, clinical testing to date has been limited. In a small clinical study of blood purification in septic patients, Schefold et al. (18) showed that immunoadsorption-capture of endotoxin, IL-6, and C5a significantly improved APACHE II scores, suggesting that PAMPs contribute to acute organ injury and their removal may improve outcomes. High-volume hemofiltration therapy, where ultrafiltration is performed well above 35 mL/kg/h, can also eliminate inflammatory mediators from circulation (48, 49), but a large multicenter study showed no effect on mortality (50). CytoSorb employs porous beads to filter inflammatory molecules by mass and polarity (51). In vitro and human studies have found that CytoSorb effectively removes proinflammatory cytokines and PAMPs (51, 52). In patients with sepsis and respiratory failure on mechanical ventilation, CytoSorb was safe and removed IL-6 but did not significantly affect plasma IL-6 levels, improve acute organ injury, or decrease time on mechanical ventilation (20). Although a retrospective case-control sepsis study found that CytoSorb was associated with reduced mortality, there was no benefit seen in the subgroup of patients with pneumonia (21), highlighting the need for additional pneumosepsis therapies. Plasmapheresis is another extracorporeal sepsis therapeutic to remove circulating inflammogens and may improve hemodynamics, organ injury, and/or mortality (53–56), although the literature is mixed and at risk of bias (17). Our study sought to address this gap by investigating the effects of a novel filter, the Seraph 200 Microbind Affinity Blood Filter, in a translational model of severe pneumonia. The Seraph 200 Microbind Affinity Blood Filter contains surface-immobilized heparin on beads analogous to mammalian cells, exploiting bacteria-cell interactions. Heparin has a known binding affinity for common bacteria and bacterial DNA, and based on prior work (22, 23) we hypothesized that heparin-based blood purification in our nonhuman primate model would reduce PAMP burden and attenuate pneumonia- and sepsis-mediated acute organ injury. We found that the heparin filter significantly reduced levels of circulating whole blood S. pneumoniae DNA as measured by qPCR, which was associated with less acute organ failure, metabolic acidosis, and shock. Because both Sham and Purification groups received intravenous heparin during the experiment, the heparin ligands in the filter, not the intravenous heparin, are responsible for the observed PAMP reduction; however, we cannot exclude a possible PAMP-heparin interaction in the bloodstream, which if present would have mitigated differences in PAMPs between groups. Additionally, heparin-based blood purification in our study did not alter any of the measured serum cytokine levels, suggesting that the improvements we found in organ injury and metabolic derangements were independent of circulating cytokines.

PAMPs are a known trigger for inflammasome assembly and maturation of proinflammatory cytokines (57). Our data suggest that removal of circulating PAMPs via heparin-based blood purification altered proinflammatory and oxidative host responses. Specifically, we found significantly less renal caspase-1-dependent NLRP3 inflammasome activation. This was linked to renal protection based on lower renal NGAL expression and plasma creatinine levels. The Purification group also displayed significantly lower BALF IL-18 levels at necropsy, suggesting mitigation of inflammasome activation in the lung as well. Furthermore, we found increased distribution of renal citrate synthase, indicative of increased mitochondrial mass, in the Purification group compared with Sham. Mitochondrial oxidative injury is well described in sepsis (33, 34, 58–61) and is a stimulus for mitochondrial biogenesis and counterinflammatory signaling such as IL-10 production (62), which we observed in our study as well. Taken together, the blood purification group displayed less acute kidney injury linked to suppressed inflammasome activation and increased mitochondrial IL-10 signaling. Future animal studies should incorporate inflammasome end points into the experimental design to understand more fully the effects of the heparin-based filter on host response to sepsis.

Our study has several limitations. First, we used a shorter duration of blood purification (a single 4-h treatment) relative to prior human clinical studies that treated intermittently for 6–7 h daily for 5–7 days (18, 20) or continuously for 2 days (21), although this was the dose used in the CE Marking study (28). This shorter treatment time may have been insufficient to remove significant quantities of circulating cytokines and may explain why peripheral cytokine levels were not significantly different between groups. Second, our model employed a single dose of ceftriaxone, less than would be given in a clinical study, where each additional dose of bactericidal antibiotic may further release PAMPs. Third, our study ended at 48 h, which may precede the development or progression of acute organ injuries (63, 64). Fourth, the use of baboons to conduct a study with clinical end points, such as mortality, is neither feasible nor ethical and necessitates the use of surrogate end points (38). Each of these limitations should be addressable in a future clinical trial where septic human subjects would receive daily or continuous blood purification therapy for several days. Fifth, as with other nonhuman primate studies, and unlike rodent experiments, our preclinical study cannot determine mechanism (38). However, our data are associational and identify known PAMPs as a potential therapeutic biomarker and target for removal. Circulating PAMPs should be used in future clinical studies as an end point. Sixth, the sham purification group included two septic animals that were treated with intravenous heparin rather than the sham extracorporeal procedure, which may have introduced bias in the control group, although we found no differences between these animals. Seventh, dopamine as a vasopressor for sepsis is not first-line therapy in patients; however, our use in this study was based on prior experience, and both groups were treated equally. Eighth, we did not use the Sepsis-3 definition of sepsis in our study, although if applied the results would not change, and we caution the application of clinical definitions to animal models. Finally, studies have shown that sepsis outcomes can vary by sex, with females tending toward lower organ failure and mortality compared with males (65, 66). Septic females also have lower TNF-α and IL-6 levels and higher IL-10 levels (67, 68). These differences in host immune response may be tied directly to sex hormone signaling (66). As our study used only male baboons, differences in host response by sex to heparin-based blood purification are not clear but may be less dramatic in females and are deserving of further study.

In general, preclinical studies are safety checks that prevent harmful therapies from reaching patients and causing injury or death (38, 69). Preclinical testing of this blood purification therapy was necessary to ensure safety. Anticipated adverse effects included bleeding due to systemic heparin administration and hypotension due to intravascular fluid shifts, but neither occurred in either the uninfected animals or the septic animals that underwent blood purification. No unanticipated adverse effects or laboratory abnormalities were identified either, but the duration of blood purification was limited. This safety profile supports the advancement of heparin-based blood purification therapy for pneumonia and sepsis into the clinical arena. Such clinical testing should be rigorous, to ensure not only that the purification treatment removes specific molecules but also that it significantly improves patient-centered outcomes (70).

In conclusion, the release of pathogen-associated molecular patterns (PAMPs) after antibiotic administration in patients with sepsis may be more clinically relevant than is currently recognized and may worsen the systemic inflammatory response syndrome, septic shock, and acute organ failure. Our data suggest that these complications are mitigated by PAMP capture therapies, such as heparin-based blood purification. Our results support introduction of this new therapy into the clinical arena.

SUPPLEMENTAL DATA

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.12678560

Supplemental Fig. S2: https://doi.org/10.6084/m9.figshare.12678656

Supplemental Fig. S3: https://doi.org/10.6084/m9.figshare.12678674

Supplemental Fig. S4: https://doi.org/10.6084/m9.figshare.12678707

Supplemental Fig. S5: https://doi.org/10.6084/m9.figshare.12678719

Supplemental Fig. S6: https://doi.org/10.6084/m9.figshare.12678722

Supplemental Fig. S7: https://doi.org/10.6084/m9.figshare.14417558

Supplemental Fig. S8: https://doi.org/10.6084/m9.figshare.12678728

Supplemental Fig. S9: https://doi.org/10.6084/m9.figshare.14418479

Supplemental Fig. S10: https://doi.org/10.6084/m9.figshare.14418482

GRANTS

This work was funded by DARPA Grant HR0011-15-2-0057 and National Heart, Lung, and Blood Institute Grant K08 HL-130557.

DISCLAIMERS

DARPA did not have a role in study design, data collection, analysis, interpretation, or in writing the manuscript.

DISCLOSURES

B.D.K. has received honoraria from La Jolla Pharmaceutical Company, Shionogi, Paratek Pharmaceuticals, and Boehringer Ingelheim and grant funding from NIH (K08 HL-130557), DARPA, the Marcus Foundation, and the U.S. Department of Defense and has a patent pending on the use of mesenchymal stromal cell products to treat viral infections. C.W.W. has received grant funding from NIH and DARPA and is a cofounder of Predigen, Inc. E.L.T. has received grant funding from DARPA, NIH, and DTRA; is a cofounder of and has equity in Predigen, Inc.; has received honoraria from bioMerieux; has a patent issued for biomarkers for the molecular classification of bacterial infection; and has a patent pending for methods to diagnose and treat acute respiratory infections. G.S.G. has received grants from DARPA, NIH, and DTRA; is a cofounder of and has equity in Predigen, Inc.; has a patent issued for biomarkers for the molecular classification of bacterial infection; and has a patent pending for methods to diagnose and treat acute respiratory infections. D.M.M. has received grant funding from NIH. H.B.S. has received grant funding from the U.S. Department of Defense, NIH, and the U.S. Office of Naval Research. C.A.P. has received grant funding from NIH, the Department of Veterans’ Affairs, the U.S. Department of Defense, and the U.S. Office of Naval Research. K.E.W.-W. has received grant funding from the U.S. Department of Defense. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

L.C., B.D.K., H.B.S., C.A.P., and K.E.W.-W. conceived and designed research; L.C., B.D.K., Z.R.H., D.M.M., H.B.S., C.A.P., and K.E.W.-W. performed experiments; L.C., B.D.K., V.L.R., Z.R.H., D.M.M., H.B.S., C.A.P., and K.E.W.-W. analyzed data; L.C., B.D.K., V.L.R., Z.R.H., C.W.W., E.L.T., G.S.G., D.M.M., H.B.S., C.A.P., and K.E.W.-W. interpreted results of experiments; L.C., B.D.K., and H.B.S. prepared figures; L.C. and B.D.K. drafted manuscript; L.C., B.D.K., V.L.R., Z.R.H., C.W.W., E.L.T., G.S.G., D.M.M., H.B.S., C.A.P., and K.E.W.-W. edited and revised manuscript; L.C., B.D.K., V.L.R., Z.R.H., C.W.W., E.L.T., G.S.G., D.M.M., H.B.S., C.A.P., and K.E.W.-W. approved final version of manuscript.