Preservation of Aqueous Outflow Facility after Second-Generation FIV Vector-Mediated Expression of Marker Genes in Anterior Segments of Human Eyes

Abstract

purpose. Feline immunodeficiency virus (FIV)–based lentiviral vectors produce effective genetic modification of the trabecular meshwork (TM) of human eyes in organ-perfusion culture, resulting in high-level expression of a β-galactosidase marker gene (lacZ) without loss of TM cellularity or architecture. However, effects on aqueous outflow physiology have not been determined, and the ability to monitor FIV vector transgene expression in living TM in situ has not been established. In the current study, transgene expression and outflow facility were evaluated in perfused human anterior segments after FIV vector transduction of lacZ or of a marker gene that can be monitored noninvasively, enhanced green fluorescent protein (eGFP).

methods. Second-generation FIV vectors were made with a protocol for scaled-up production that requires 10 times less input DNA and allows simplified concentration. One vector encodes β-galactosidase (vector CT26), and the other (bicistronic) encodes eGFP and neomycin phosphotransferase (vector GiNWF). Three pairs of eyes were injected with 1 × 10⁸ transducing units (TU) of CT26 in the right eye and with a control (mock lacZ) vector in the left eye. Three others were injected with 1 × 10⁸ TU GiNWF in the right eye only, with the left eye serving as an uninjected control. Intraocular pressure was recorded and transduction efficiency was determined.

results. The modified protocol produced high-titer FIV vectors, and coordinate expression of marker genes was observed with the bicistronic vector. In human eyes, the eGFP and lacZ vectors transduced 79% ± 15% and 82% ± 4% of TM cells, respectively, without cell loss compared with control eyes. Transduction and marker gene expression caused a transient decrease of outflow facility (30% ± 22%, P = 0.02), which resolved after 48 to 72 hours.

conclusions. FIV vectors produce high-level expression of eGFP in the TM of the cultured human eye, with transduction efficiency similar to that obtained with β-galactosidase vectors. Transduction and expression of these marker genes results in small and transient changes in outflow facility, suggesting suitability of this class of vectors for glaucoma gene therapy.

High intraocular pressure is the most common causal risk factor for glaucoma and is the key variable for clinical management. Intraocular pressure is primarily determined by aqueous humor production and outflow. The trabecular meshwork (TM) generates most of the resistance to outflow, which normally maintains intraocular pressure within the narrow range (10–21 mm Hg) that is essential for viability of the neuroretina. Impairment of TM-regulated outflow is believed to be the primary physiological derangement in the majority of primary open-angle glaucoma patients. ^{1 2}  

Strategies for gene therapy for glaucoma must consider the chronic nature of the disease and the generally nondividing nature of the target cells. Two general strategies have emerged. One focuses on the intraocular pressure problem, and the other on blocking its sequela, retinal ganglion cell death. ^{3 4} The accessibility of the anterior chamber and the restricted anatomic target are favorable for gene therapy directed at the TM. Gene therapy with retroviral vectors has appeal, because these vectors undergo reverse transcription of their single-stranded RNA genomes, generating a linear double-stranded DNA intermediate that that is subsequently integrated into the host genome in a reaction catalyzed by the retroviral integrase. Therefore, these vectors result in permanent transgenes and have potential to address the chronicity of glaucoma pathophysiology. However, an important consideration is that TM cells do not normally divide. Unlike conventional retroviral vectors based on, for example, murine leukemia viruses (MLVs), lentiviral vectors, such as those derived from HIV and FIV, integrate into the genomes of both dividing and nondividing cells. ^{5 6}  

FIV infects most feline species worldwide but is pathogenic only in the domestic cat. ⁷ This nonprimate lentivirus has been considered as a substrate for vector development for reasons that include its complete lack of serologic cross-reactivity with HIV-1 (for example in diagnostic HIV-1 antibody tests). ⁶ In addition, there is an extensive record showing no human infection or disease, despite widespread human exposure to FIV through its principal and very efficient natural mode of transmission (biting), and despite the virus’s ability to use a human chemokine receptor for entry. ^{6 8 9} Lentiviral vectors derived from FIV principally transduce TM after injection into the anterior chamber of the human eye. ¹⁰ Transduction was efficient (>70% of TM cells, exceeding 95% in some eyes), without affecting the fine structure of the TM or causing significant cellular loss. Moreover, this FIV vector, which transduces lacZ, has been shown by paired comparisons with normalized vector stocks to be equivalent to a lacZ vector, based on the human lentiviral pathogen HIV-1. ¹⁰ In contrast, an MLV lacZ vector failed to transduce the TM, ¹⁰ a result that is consistent with the need for target cells to passage through mitosis for MLV to integrate into chromatin. ¹¹  

These results suggest that FIV vectors should be further investigated for glaucoma gene therapy. However, the previously studied β-galactosidase marker can be evaluated only after terminal tissue processing. The ability to monitor transgene expression in the living state is advantageous for establishing gene delivery approaches. To determine whether FIV vector-transduced enhanced green fluorescent protein (eGFP) could be used to quantify transgene expression in the human TM, we compared new vectors encoding eGFP with vectors encoding lacZ. In each of the anterior segments, we simultaneously determined the effects of transgene transfer and expression on aqueous outflow. Because methods used for glaucoma gene therapy must be capable of genetically modifying the human TM without counterproductively impairing its regulation of aqueous outflow, these experiments are a necessary step in ascertaining the clinical potential of these vectors. 

Methods

FIV Vector Plasmid Construction

Plasmids used to generate FIV vectors were as follows: packaging construct pCF1Δenv ^{6 8 10} ; pMD-G, which expresses vesicular stomatitis virus glycoprotein G (VSV-G) under control of the human cytomegalovirus immediate early promoter (CMVp); and two FIV transfer vectors: GiNWF, which encodes eGFP, and CT26, which encodes β-galactosidase (lacZ; Fig. 1 ). pGiNWF contains, from 5′ to 3′, a hybrid U3-substituted promoter derived from pCT5, ^{6 8} the FIV R repeat, U5 element, leader sequence, the first 311-bp of the gag gene, the Rev response element (RRE, nucleotides [nt] 8537–8952 of the FIV 34TF10 genome), a sequence (FIV nt 4904–5191) containing the FIV central polypurine tract (cPPT), and the central termination sequence (CTS), the CMVp, eGFP, an internal ribosomal entry site (IRES) neoR, the woodchuck hepatitis virus posttranscriptional regulatory element (WPRE), ¹² and the 3′ long terminal repeat (LTR). The cPPT-CTS combination is also referred to as the central DNA flap, because the strand initiations and terminations that occur at these loci result in a triple-stranded DNA flap structure at the completion of FIV reverse transcription. ¹³  

pCT25, a previously described lacZ transfer vector, ¹⁰ was parental to a sequentially constructed series of eGFP-containing plasmids eventuating in pGiNWF: pGiN (also called CT25.eGFP.ires.neo), pGiNW, pGiNWcPPT-CTS, and pGiNWF. To first construct pGiN, peGFP-1 (BDE Biosciences-Clontech, Palo Alto, CA) was cleaved with NotI, blunted with Klenow fragment, and then cleaved with BamHI. The resultant 0.74-kb fragment was inserted into an MLV retroviral vector between BamHI and HpaI, yielding the EcoRI-containing sequence GCGGCCAACGAATTC at the 3′ junction. This eGFP insert was then excised with BamHI and EcoRI and inserted into the BamHI-EcoRI–cleaved backbone of pCT25, thus replacing lacZ and generating pCT25.eGFP. A 1.49-kb Sal-Nhe fragment containing the internal ribosome entry site and neoR gene from pJZ308¹⁴ was then inserted by blunt-end ligation into the EcoRI site pf pCT25.eGFP, generating pGiN (pCT25.egfp.ires.neo). pGinW was constructed by inserting the WPRE by blunt insertion of an EcoRV-XhoI fragment of pBluescriptIISK+WPRE-B11 ¹² into the BspEI site of pGiN (yielding a regenerated BspEI site that is blocked by dam methylation). 

Finally, the central DNA flap was inserted in several steps as follows. A 279-nt amplicon containing the FIV cPPT-CTS was synthesized by PCR, by using a sense primer tailed with a BstBI site (5′-ATATTTCGAATCAAATCAAACTAATAAAGTATGTATTGTGAAACAACCTCCTTGGATAATGCC-3′) and an antisense primer tailed with an Xba site (5′-ATATACCTCTTTTAGGTCTAGACTCTCATGTGTCTCCTAGG-3′). The sense primer fuses the cPPT with the 3′ end of the FIV RRE and deletes an unneeded splice acceptor. This BstB1-Xba amplicon was inserted into the corresponding sites of pGiNW, generating pGiNWcPPT-CTS. The latter maneuver removed the internal CMV promoter from pGinW and inserted the cPPT-CTS. To restore this promoter, the BamHI-XhoI fragment of pGiNW was inserted into the XhoI-BamHI backbone of pCR2.1 (Invitrogen, San Diego, CA). An XbaI linker (GCTCTAGAGC) was inserted into the Klenow-treated AflII site of this intermediate plasmid, and the 614-nt XbaI-XbaI fragment was then inserted into the Xba site of pGiNWcPPT-CTS, generating pGiNWF. 

CT26 contains, from 5′ to 3′, the hybrid promoter, R repeat, U5, leader, 311 bp of gag, RRE, central DNA flap, human CMV immediate early promoter lacZ, and 3′ LTR. To construct pCT26, the cPPT-CTS was PCR amplified from FIV 34 Tf10 with sense primer 5′-aaaaCCTTCAAGAGGctgcagaaacaacctccttggataatgcc-3′ and antisense primer 5-atataCCTTCAAGAGGtctagactctccttatgtgtctcctagg-3′. The amplicon was blunted into the EcoNI site of pCT25, inserting the cPPT downstream of the RRE. 

Large-Scale Vector Production and Concentration

A protocol was modified from that previously described ¹⁰ to enable scaled-up production of lentiviral vectors by calcium phosphate–mediated transfection of 293T cells. Ten-chamber Cell Factories (CF210; Nunc, Naperville, IL) were used for vector production, with ultracentrifugation in a large-volume (1.3-L) rotor. Briefly, 293T cells were maintained at a high frequency of passage before 2.52 × 10⁸ cells were seeded into one 10-chamber cell factory with a surface area of 6320 cm² and containing a total volume of 1 L medium (DMEM with 10% fetal calf serum [FCS] and antibiotics). After overnight incubation at 37°C in 5% CO₂, a calcium phosphate transfection mix containing 84 μg pMD-G, 252 μg pCF1Δenv, 252 μg transfer vector, 60 mL of 0.01 M Tris (pH 8.0), 6.72 mL of 2.5 M CaCl₂, and 67.2 mL 2× Hepes buffered saline (HBS; pH 7.0) was allowed to precipitate for 3 minutes. Precipitation was stopped by adding DMEM+10% FCS to a total volume of 1 L. This amount of DNA (0.093 μg/cm²) is equivalent to 7 μg DNA in a 75-cm² flask. The old medium was removed from the CF10s and the medium containing the transfection mix was added and allowed to settle on the cells for 18 hours before replacement with fresh medium. Vector supernatants were harvested 48 hours after this medium change, filtered through 0.22-μm filters, and ultracentrifuged in large-volume (250-mL) buckets (model 54477; Sorvall, Newtown, CT) in a fixed angle rotor (model A 621; Sorvall) at 19,000 rpm (49,000g) for 2 hours. Supernatants were decanted, and vector pellets were resuspended by pipetting in 10 mL PBS per bucket. Vector suspensions were then concentrated with a second round of ultracentrifugation at 49,000g in a swinging bucket rotor (Surespin; Sorvall), and pellets were reconstituted in 2 mL PBS and centrifuged for 5 minutes at 3000g to remove undispersible material. 

Mock vector, which was used as a control for pseudotransduction, ¹⁵ was prepared by transient transfection into 293T cells of the same amounts of pCT26 and pMD-G as for preparation of the CT26 vector, while omitting the packaging plasmid. Transfection efficiency for both the CT26 and the CT26 mock vector were greater than 80%, as evaluated by 5-bromo-4-chloro-3-indolyl-β-d-galactosidase (X-Gal) staining of the 293T producer cells. Mock vector supernatants underwent the same centrifugation and other processing steps as normal vectors. Vectors were titered by serial dilution onto CrFK cells plated in six-well dishes followed by either flow cytometry for eGFP (GiNWF) or by quantitating X-Gal–stained colony formation (CT26). 

Human Organ-Perfusion Culture and Vector Application

The Minnesota Lions Eye Bank (Minneapolis, MN) provided pairs of human eyes. Eyes were used in accordance with Mayo Clinic Institutional Review Board guidelines and according to the tenets of the Declaration of Helsinki. All were from donors (median age, 76.5 years, n = 6) without known eye disease and were placed in perfusion culture within 24 hours of death, as previously described. ^{16 17} After bisection at the equator, the retina, iris, and lens were removed, and the anterior segments were sealed into a custom-built culture vessel and perfused with DMEM and antibiotics at a normal rate (2.5 μL/min) with a microinfusion pump. The cultures were maintained at 37°C in 5% CO₂ in an incubator. Intraocular pressure was measured every 60 seconds for 5 days and recorded as averages per hour. One anterior segment of each pair (n = 3) was injected with a bolus of 1 × 10⁸ transduction units (TU) CT26 vector in a volume of 500 μL DMEM, and the fellow eyes received an equivalent volume of CT26 mock vector. Similarly, 1 × 10⁸ TU of GiNWF vector in 500 μL DMEM was injected into one anterior segment of a pair (n = 3), whereas the control fellow eye received the same volume of medium without vector. 

Assessment of Transduction

Transduction efficiency was determined as the mean percentage of transduced cells in four quadrants. Eyes in the eGFP group were divided into four quadrants, and 2-mm wedges were removed, rinsed in PBS, and placed into phenol red–free Dulbecco’s modified Eagle’s medium with 0.1% Hoechst 33342 (H-3570; Molecular Probes, Eugene, OR) for in vivo staining of nuclei. Expression of eGFP in freshly dissected, unfixed anterior segments was visualized by a frontal view toward the TM and a sagittal view of wedges mounted on their sides, by a microscope with a fluorescent light source (Eclipse E400; Nikon, Melville, NY) and by confocal microscopy (LSM 510; Carl Zeiss, Thornwood, NY). To measure transduction efficiency with GiNWF, eGFP-positive TM cell bodies and stained nuclei were manually counted at 400× magnification in microscopic fields comprising an area of 0.24 mm² in each of the four quadrants. Transduction efficiency is expressed as the mean percentage of eGFP-expressing cells versus stained nuclei ± SD. Medium-injected fellow eyes served as the control. Three-μm-thick, plastic-embedded sections (JB4; Polysciences, Warrington, PA) of two quadrants were used to compare cell loss and assess the morphology. 

Localization and extent of expression of eGFP were confirmed with specific antibody labeling. Briefly, 6-μm paraffin-embedded sections from two quadrants of each eye were deparaffinized (22-143975; Citrisolv; Fisherbrand, Fair Lawn, NJ) and rehydrated with decreasing concentrations of ethanol, and antigens were retrieved for 5 minutes in a steam chamber with piperazine-N-N′-bis(2-ethanesulfonic acid) (PIPES) buffer. Sections were incubated for 60 minutes with a primary rabbit anti-eGFP antibody (1:200 dilution, NB 600-303; Novus Biologicals, Littleton, CO) and for 30 minutes with a fluorescent phalloidin-labeled secondary goat anti-rabbit antibody (1:200 dilution, A-11088, Alexa 488; Molecular Probes). Nuclei were stained with a 1:1000 dilution of 4′,6′-diamino-2-phenylindole (DAPI; D-1306; Molecular Probes). 

For assessment of β-galactosidase expression in CT26 and CT26.mock vector-injected eyes, anterior segments were fixed for 15 minutes in 4% paraformaldehyde, rinsed in PBS, and incubated overnight in X-Gal staining solution. ¹⁰ β-galactosidase–positive cells and total TM cells were counted in random sections of each quadrant. Sections from two quadrants of each eye were embedded in plastic (JB4) for analysis of TM fine structure and morphology. Total number of cells in control eyes and transduced eyes were compared with Student’s paired t-test. 

Results

Outcome of Modified Vector Production Procedures

Several modifications were made to the original ⁶ vector system. gag sequences in transfer vectors were reduced to the most 5′ 311 nt. We have found that these 311 nt contain a necessary encapsidation determinant; vectors with this truncated gag segment are in fact packaged into virions at higher efficiency than those with the longer gag segment contained in the original vectors¹⁸ and, in the case of the lacZ vector CT25, have been shown to transduce human TM effectively. ¹⁰ New transfer vectors used in the present study also contained new transgenes (eGFP and neoR) and additional viral elements that are involved in reverse transcription and viral nuclear import ¹³ or that enhance transgene mRNA levels ¹² (Fig. 1) . To enhance scalability of vector production, we also developed modified procedures that use much smaller amounts of plasmid DNA than previously described protocols (10 times less in micrograms of DNA per square centimeter of 293T cell monolayer), as well as high-surface-area vessels (Cell Factory; Nunc) and high volume ultracentrifugation (see the Methods and Materials section). Large-scale vector supernatants produced with these methods in cell factories averaged 2.8 ± 1.5 × 10⁶ TU/mL (n = 8). The first round of ultracentrifugation resulted in vector recoveries of 60% to 80%, the second round in recovery of 40% to 60%, and maximum titers of 2.5 × 10⁹ TU/mL. Vector preparations used for the organ perfusion experiments contained 6.83 × 10⁸ TU/mL (GiNWF) and 2.5 × 10⁹ TU/mL (CT26) and were diluted for injection of 1 × 10⁸ TU in a volume of 500 μL. For bicistronic vector preparations, eGFP and neoR titers correlated within 50% of each other and more than 95% of G418-stable colonies expressed eGFP, indicating good function of the internal ribosome entry site. 

Expression of eGFP and β-Galactosidase in Human TM

GiNWF vector injection into cultured anterior segments produced efficient TM transduction, as determined by visualization of expression of eGFP (Fig. 2) . Transduction was limited to the TM, with the exception of occasional eGFP-positive cells elsewhere within the outflow pathway (corneal endothelium in proximity to the TM, Schlemm’s canal, collector channels). Cell counts performed with nuclear counterstaining revealed that a mean of 82% ± 4% of TM cells were eGFP positive and no cell loss was apparent (P = 0.94 for comparison with control eyes, n = 3). Staining of tissue sections with anti-eGFP antibodies confirmed the extent and location of expression of eGFP (data not shown). Plastic-embedded sections showed preserved TM morphology and cellularity. Transduction of two unpaired eyes with a version of GiNWF without the central DNA flap or WPRE (pGiN) showed less efficient transduction (data not shown). 

Transduction with the lacZ-encoding CT26 vector also resulted in high-level transduction (79% ± 15%) of TM cells (Fig. 3) . The fellow eyes injected with the CT26.mock vector did not stain positively for β-galactosidase, confirming absence of pseudotransduction. ¹⁵ Comparison of total cell numbers showed no significant cell loss (P = 0.18 for comparison with control eyes, n = 3) and the morphology was well preserved (data not shown). Expression of β-galactosidase was mainly observed in TM, although a few corneal endothelial cells in proximity to the TM and occasional cells in the endothelia of Schlemm’s canal and the collector channels were also transduced. 

Effects on Outflow Facility

Despite the extensive transduction and high levels of marker gene expression in both sets of eye pairs, only slight and transient changes in outflow facility were observed (Fig. 4) . In six eyes with high (79% or greater) transduction of TM cells with lacZ or eGFP, a mean 30% ± 22% (P = 0.02) peak reduction in outflow facility was observed. This minimal decrease in outflow facility was followed by stable return to preinjection levels by approximately 48 to 72 hours. The experiments were terminated at 5 days. The mean peak decreases were not significantly different between the lacZ and eGFP groups (25% ± 30% and 36% ± 17%, respectively, P = 0.6, n = 3 per group). CT26.mock vector–injected eyes did not show a significant change in outflow facility compared with medium-injected control eyes. 

Discussion

The goal of glaucoma gene therapy directed to the anterior chamber is to restore and preserve aqueous outflow through the TM. For therapeutic transgenes to be validated however, and for eventual clinical application to proceed, a method for stable and nontoxic genetic modification of the postmitotic cells involved in glaucoma pathogenesis is needed. This capability is necessary for experimental evaluation of candidate therapeutic genes and for eventual use of those genes in therapy. In particular, the method of gene transfer must not cause counterproductive disruption of aqueous outflow through the TM. We have shown that lentiviral vectors derived from either HIV or FIV can be used to express β-galactosidase in most human TM cells in the organ-perfusion model. ¹⁰ Our histologic analyses in those experiments suggest that cellularity is well preserved and that lentiviral vector transduction is minimally toxic. In the current study, we extended the results of this model by demonstrating that eGFP can also be used as an effective marker gene in FIV vectors, by testing second-generation vectors that incorporate the FIV central DNA flap and WPRE, by using more scaleable production methods, and by showing that the extensive transduction observed with these vectors does not significantly impair outflow facility. The use of eGFP will facilitate noninvasive, sequential monitoring of transgene expression in animal experiments. 

The results validate the protocol for large-scale production of lentiviral vectors. Transient transfection of 10 times less DNA in 293T cells within high surface area slides (Cell Factory; Nunc) and high volume, fixed-angle ultracentrifugation resulted in high titer vectors that were effective in the eye. Standard protocols use 5 to 10 times as much transfection DNA, which is expensive and time consuming to produce, and generally concentrate vectors in smaller volumes. ^{5 6 10}  

Aqueous humor outflow is a complex process. The unique architecture of the TM, the phagocytic biology of the cells, their arrangement within a collagenous lattice, and the extracellular matrix they elaborate are all believed to play important roles. A final regulatory step may occur at the interface of the TM with Schlemm’s canal, where bulk flow occurs through large outpouchings (giant vacuoles) in the endothelium. ¹⁹ A potential concern for gene therapy is that gene transfer methods might disrupt the physiology of this structure. Despite a high level of transgene expression, FIV vector–mediated gene transfer caused only transient, slight (mean of 30%) declines in outflow facility, with stable return to normal baseline levels from 48 to 72 hours after transduction until the end of the experiment at 5 days. These results compare favorably with those obtained by Borrás et al. ¹⁹ in this model after injection of adenoviral vectors. After injection of 1 × 10⁸ TU of adenoviral vectors, outflow facility declined 13% compared with control eyes within the first 4 hours, and was reduced by 54% after 12 hours. Subsequently, baseline outflow facility was reached at 36 hours and continued to increase to approximately 20% higher than baseline until the end of the experiment at 48 hours. ²⁰ As in our previous experiments, ¹⁰ we observed preferential transduction of TM. Targeted transduction prevents unwanted effects of transgene expression on neighboring structures. This is of particular importance in ocular gene therapy, in which anatomic structures in close proximity serve highly specialized functions. 

 Supported by National Eye Institute Grant NIH EY07065-15 (DHJ) and National Institute for Allergy and Infections Diseases Grant AI47536 (EMP).

 Submitted for publication April 29, 2002; revised June 20, 2002; accepted July 17, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

FIV transfer vectors GiNWF and CT26. RRE: Rev response element. Central DNA flap: FIV segment containing the cPPT and CTS, which are used in reverse transcription of FIV genomic mRNAs and result in an 88-nt triple-stranded overlap structure in the preintegration complex. ¹³ CMVp, human cytomegalovirus immediate early promoter; IRES, internal ribosome entry site; Neo, neomycin phosphotransferase; WPRE, woodchuck hepatitis virus posttranscriptional regulatory element; R, repeat element; U5, 5′ unique element. Both full-length vector mRNAs are transcribed from a hybrid promoter (CMV/FIV) that permits high-level expression in human cells. Reverse transcription results in its replacement with a transcriptionally silent feline 3′ unique (U3) element in the human target cell. ⁶

Figure 2.

 

(a) Confocal microscopy view with frontal and sagittal view of TM 5 days after transduction with GiNWF and the TM in the fellow control eye. Imaging depth of 20 μm with wide pinhole. (b) Conventional fluorescence microscopic view. Magnification, ×400.

Figure 3.

 

Chamber angles 5 days after injection of CT26 (a), and CT26.mock vector (b). Sections are stained with X-Gal to detect expression of β-galactosidase. SC, Schlemm’s canal.

Figure 4.

 

Intraocular pressure. (a) Recording of eye injected with CT26 and CT26.mock vector control. (b) Ratio of outflow facility (1/R) before transduction (C₀) and at recorded minimum (C_min) and C₀ and 72 hours after transduction (C_72h). Outflow facility ratio of transduced eyes was significantly different from the control at the recorded minimum (*P = 0.02), but returned to stable baseline at 72 hours.

The authors thank Manuel Llano for helpful suggestions, and Dennis Mrdjenovich for technical advice about large-volume ultracentrifugation.