Characterization of a murine monoclonal antibody that mimics heparin-induced thrombocytopenia antibodies

The murine myeloma cell line P3 × 63Ag8U.1 (P3U1) was purchased from American Type Tissue Culture Collection (Rockville, MD). Tissue culture reagents were from Gibco BRL (Rockville, MD). Fetal bovine serum was from Hyclone (Logan, UT). Heparin solutions used in these studies were from Elkins-Sinn (Cherry Hill, NJ). Reagents purchased from Amersham Pharmacia Biotech (Piscataway, NJ) included ECL chemiluminescence detection kit, Hi Trap affinity columns, ProRPC 15 μm HR 10/10 chromatography column, 5-hydroxy [side chain-2-¹⁴C] tryptamine creatinine sulfate, and the Sequenase T₇ DNA polymerase kit. Immunochemicals used in the studies described below include anti-PF4, anti-NAP-2, and anti-IL-8 raised in rabbits from Preprotech (Rocky Hill, NJ), swine antirabbit (horseradish peroxidase conjugate) antibody from DAKO (Carpinteria, CA), and goat antihuman IgG, IgA, and IgM from ICN (Costa Mesa, CA). ImmunoPure Monoclonal Antibody Isotyping Kit and the BCA Protein Assay Reagent Kit were obtained from Pierce (Rockford, IL). Plates (6-well, 24-well, and 96-well) for tissue culture were products of Becton Dickinson (Franklin Lakes, NJ). Maxisorp microtiter plates used for the PF4/heparin ELISA were from Nunc Brand Products (Roskilde, Denmark). Colorimetric readings were measured using a Molecular Devices (Sunnyvale, CA) plate reader. Molecular biologic reagents included pT_7-7 vector, and Escherichia coli bacterial strain BL21(DE3)pLysS from Novagen (Madison, WI), VENT polymerase from New England Biolabs (Beverly, MA), the murine λFIX 129SV library from Stratagene (LaJolla, CA), TRIzol reagent and Superscript RT-PCR kit from Gibco/BRL, and AmpliTaq from Perkin-Elmer (Branchburg, NJ). MacVector (v. 6.0) software package was purchased from Oxford Molecular Group (Oxford, UK). All other chemicals and reagents were purchased from Sigma Chemical (St. Louis, MO).

Plasma was obtained from patients with a clinical diagnosis of HIT/HITT8 and from healthy volunteers. The clinical diagnosis was confirmed using the ¹⁴C-serotonin release assay (SRA)2 and by measuring antibodies to PF4/heparin by enzyme-linked immunosorbent assay (ELISA) (see below). Institutional approval was obtained for these studies.

Four 6- to 8-week-old female Balb/c mice were injected intraperitoneally on day 0 with a 50 μL sterile solution composed of 25 μL phosphate-buffered saline (PBS) containing recombinant hPF4 (50 μg), heparin (2 U), and 25 μL Freund's complete adjuvant. Subsequent injections containing 50 μg hPF4 and 2 U heparin in PBS were administered intraperitoneally or through the tail vein on days 12, 30, 41 and 48, and 62. Mice were given an intravenous boost of heparin and hPF4 on day 66, 3 days before they were sacrificed. Titers of anti-hPF4/heparin were monitored by ELISA (see below). The 2 mice expressing the highest serum titers (more than 1:100,000) of anti-hPF4/heparin antibodies were sacrificed, and their spleens were removed for fusion. Fusion and hybridoma selection were optimized using standard methodology.23 Hybridomas were cultured for 7 days, and their supernatants were screened for antibodies to hPF4/heparin and hPF4 by ELISA. Wells considered positive (A₄₀₅ > 0.8) were weaned from HAT supplement over 7 to 10 days, subcloned by limiting dilution, and grown in pristane-primed mice to generate ascites. Monoclonal antibodies to hPF4/heparin (KKO) and hPF4 alone (RTO) were isolated from ascitic fluid using the Hi Trap affinity columns according to the manufacturer's instructions. Isotyping was performed using the ImmunoPure Monoclonal Antibody Isotyping kit according to the manufacturer's instructions.

Recombinant wild-type human and mouse PF4, mutant hPF4, hNAP-2, and hIL-8 were expressed in E coli as described.24Briefly, cDNA constructs for each chemokine (see below) were inserted in a pT7-7 vector, introduced into E coli BL21(DE3) pLysS, and grown in Luria broth containing 100 μg/mL ampicillin. Bacteria were grown to an A₆₀₀ of 1.0, followed by 3 hours of induction at 37°C with 1 mmol/L IPTG. Bacteria were lysed and sonicated, and the chemokine was purified at room temperature by affinity chromatography using heparin-agarose equilibrated with 50 mmol/L Tris HCl and 1 mmol/L EDTA, pH 8, and was eluted using a 0.2 to 2.0 mol/L NaCl gradient. Eluted proteins were further purified by reverse-phase chromatography using a ProRPC FPLC column.

Protein purity was assessed by 15% (wt/vol) sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis followed by Coomassie blue staining.24 Samples were also subjected to immunoblotting after electrotransfer to polyvinylidenedifluoride membranes using commercial rabbit anti-hPF4, anti-hNAP-2, or anti-hIL-8 polyclonal antibody, followed by swine antirabbit secondary antibody conjugated to horseradish peroxidase. Proteins were detected by ECL as described by the manufacturer. Protein concentrations were determined by the BCA assay using bovine serum albumin as the standard according to the manufacturer's instructions.

Recombinant hPF4 variants were generated using overlap polymerase chain reaction (PCR) as previously described.18 These constructs used wild-type hPF4 cDNA as templates and VENT polymerase enzyme. The sequence of each mutant construct was verified using the Sequenase T₇ DNA Polymerase Kit. Several new hPF4 constructs were generated, designated P37N, T38Q, A39V, L41V, and N47D PF4 that refer to specific amino acids in the 3rd domain of PF4 that were switched individually to those found in hNAP-2.

The cDNA corresponding to the coding region of mature murine PF4 (mPF4) was derived from sequencing an isolated murine λFIX 129 SV library clone using human PF4 cDNA as a probe. The coding primers for mPF4 were used to amplify mPF4 cDNA from murine platelet RNA by RT-PCR. Mature mPF4 cDNA was subcloned into pT7-7 vector with an ATG added in-frame at its N-terminus. cDNA encoding for mature NAP-2 and IL-8 were based on previously published sequences and isolated cDNA.18,25 

Binding of antibody to either PF4, PF4 variants, PF4 bound to various GAGs, or other chemokines complexed to heparin was measured using an ELISA-based method as previously described.2Briefly, to screen hybridoma culture supernatants, 96-well microtiter plates were coated overnight at room temperature with 50 μL/well PBS containing PF4 (final concentration, 10 μg/mL) in the presence or absence of heparin (0.2 U/mL). The plates were then washed 3 times with Tris-buffered saline/0.01% Tween-20, blocked with 10% fetal calf serum (FCS) (200 μL/well) in PBS for 2 hours at room temperature, and washed once more. Culture supernatant was added (50 μL/well) for 1 hour at room temperature. Unbound antibody was removed by washing, and 50 μL/well of alkaline-phosphatase-conjugated goat antimouse IgG, diluted 1:1000 in 10% FCS/PBS, was added for 1 hour at room temperature. After washing, 50 μL/well Sigma Fast p-nitrophenyl phosphate substrate was added, and the absorbance at 405 nm was measured. Binding of KKO to PF4 and related chemokines, or PF4 bound to various GAGs, was measured in the same manner except that an incubation volume of 100 μL/well was used.

The capacity of KKO to bind hPF4/heparin-coated wells in the presence of HIT sera was assessed using a concentration of KKO (diluted in 10% FCS/PBS) that produced 75% of maximal binding. The diluted KKO was added to wells in the presence of varying concentrations of HIT/HITT sera, and the binding of KKO to hPF4/heparin was measured by ELISA as described above.

Cultured human umbilical vein endothelial cells (HUVECs) were prepared as described.26 Cells were grown to confluence in complete media containing M199 supplemented with 20% FCS, 100 μg/mL penicillin, 100 μg/mL streptomycin, 5 μg/mL amphotericin B, endothelial cell growth supplement, and heparin (100 μg/mL) in 75 cm² flasks. The cells were then seeded onto 96-well microtiter plates26 at a density of 32,000 cells/well in the absence of heparin. After 48 hours in heparin-free medium, the cells were fixed with 0.05% glutaraldehyde, and binding of KKO or RTO was measured by ELISA as described above. In other experiments, binding of KKO to Chinese hamster ovary (CHO) cells lacking xylosyl transferase (provided by C. Esko, University of California and K. Williams, Thomas Jefferson University) was conducted in essentially the same manner.

Platelet activation by KKO in the presence of PF4 and heparin was assayed by the release of ¹⁴C-serotonin,2 with modification for a microtiter-well format. Briefly, citrated platelet-rich plasma obtained from aspirin-free healthy donors was labeled with ¹⁴C-serotonin (22.5 nCi/mL PRP) for 30 minutes at 37°C, after which platelet uptake of ¹⁴C-serotonin was blocked by the addition of excess imipramine (1 μmol/L final concentration). KKO or isotype control (30-320 μg/mL) in modified Tyrode's buffer (137 mmol/L NaCl, 3 mmol/L KCl, 0.4 mmol/L NaH₂PO₄, 12 mmol/L NaHCO₃, and 1 mmol/L MgCl₂·6H₂O, pH 7.0) was preincubated with either hPF4 (10 μg/mL) alone, heparin alone (0.2 U/mL), or hPF4 (10 μg/mL) plus heparin (0-100 U/mL). Labeled platelets (75 μL) were added in triplicate to wells containing antigen/antibody-containing mixture (20 μL) with 5 μL heparin or buffer to yield a 100 μL final volume. After a 1-hour incubation at room temperature, the reaction was terminated by the addition of 100 μL of 0.5% Na EDTA (pH 8.5), platelets were pelleted, and the release of¹⁴C-serotonin was measured by liquid scintillation. In other experiments, ¹⁴C-labeled platelets were incubated with the FcγRIIA blocking antibody mAb IV.3 (7.2 or 72 μg/mL)27 for 1 hour before the addition of KKO (250 μg/mL), PF4, and heparin. Controls for the assay included plasma from patients with HIT (SRA+, positive control), normal plasma, mIgG_2b (isotype control), RTO, and calcium ionophore A23187.

Total RNA was prepared from approximately 4 × 10⁶ hybridoma cells for clones KKO and RTO (TRIzol reagent) followed by cDNA synthesis primed with oligo dT following the manufacturer's instructions. Heavy- and light-chain immunoglobulin variable regions were amplified using the PCR as previously described28 according to the following framework 1 and constant region primers for murine γ and κ chains: heavy-chain forward 5′-GAGGTGAAGCTGGTGGAG(T/A)C(T/A)GG-3′, heavy-chain reverse 5′-GGGGCCAGTGGATAGAC-3′, light-chain forward 5′-CCAGTTCCGAGCTCCAGATGACCCAGACTCCA-3′, light-chain reverse 5′-GTTGGTGCAGCATCAGC-3′. The PCR products (350-400 bp) were gel purified by electroelution and directly sequenced using the above oligonucleotides and automated fluorescence sequencing at the Nucleic Acid/Protein Research Core Facility of The Children's Hospital of Philadelphia. Use of these “universal” variable region framework 1 primers provided heavy- and light-chain sequences that began at the 8th and 9th amino acid residues, respectively. To determine the authentic amino acid N-terminal residues for KKO chains, putative leader sequences were determined by searching Genbank for other murine monoclonal antibodies with close homology to KKO. For the heavy and light chains of KKO, sequence accession numbersAF025443 and M20830 provided candidate leader sequences. Sets of 5′ PCR primers beginning at the 5′ end of the leaders were synthesized (5′-ATGGGATGGAGCTATATCATCC-3′ for heavy chain and 5′-ATGATGAGTCCTGCCCAGTTCC-3′ for light chain). PCR amplifications of KKO heavy- and light-chain cDNA were performed using these primers paired with the original set of constant region reverse primers. PCR products of the appropriate size (400-450 bp) were obtained and served as templates to provide full-length variable region KKO sequences. Immunoglobulin gene family assignments for heavy and light chains were determined using the Kabat29 and Genbank databases. Alignments of the predicted amino acid sequences were performed using the MacVector software package.

Seven days after fusion, 128 of 1152 wells contained antibodies to either hPF4/heparin or hPF4 by ELISA using A₄₀₅ ≥ 0.8 as an arbitrary cut-off value. Cells were subcultured when their supernatants generated A₄₀₅ ratios to hPF4/heparin versus hPF4 > 1.5. Cell populations showing the greatest relative specificity for hPF4/heparin versus hPF4, and those with high reactivity to hPF4 alone, underwent 3 additional rounds of subcloning. Two monoclonal antibodies, an IgG_2bκ hPF4/heparin-dependent antibody designated KKO and an IgG_2bκ anti-hPF4 designated RTO, were ultimately isolated and subjected to further characterization. KKO exhibited an A₄₀₅ ratio of >30 to hPF4/heparin versus hPF4, whereas RTO demonstrated an A₄₀₅ ratio of ≤1. Monoclonal antibodies purified from ascites were used for all subsequent studies.

KKO bound to hPF4/heparin complexes in a dose-dependent manner (Figure 1A; half-maximal binding 0.036 μg/mL). The relative specificity of KKO was determined by adding various dilutions of KKO to wells coated with hPF4/heparin or hPF4 alone. Specificity for the complex was evident at all antibody concentrations tested (0.007-36 μg/mL; Figure 1A). At concentrations less than 0.14 μg/mL, the ratio of binding of KKO to PF4/heparin versus PF4 determined by A₄₀₅ was more than 400. No binding of KKO to PF4 alone was evident at concentrations less than or equal to 0.072 μg/mL. KKO did not bind to immobilized heparin at any concentration tested (0.007-36 μg/mL). Although KKO demonstrated heparin-dependent binding to hPF4, binding of RTO to hPF4 was unaffected by the presence of heparin at all concentrations of antibody tested (Figure 1B).

Naturally occurring HIT antibodies recognize hPF4/heparin complexes formed over a narrow range of molar ratios between the reactants.30 To determine whether KKO showed similar characteristics, antibody binding was measured at a fixed concentration of hPF4 (10 μg/mL) and varying concentrations of heparin (0.01-50 U/mL). Optimal binding of KKO occurred at a molar ratio of hPF4-to-heparin of 3:1. Reduced KKO binding was observed in the presence of higher and lower concentrations of heparin (Figure1C).

Another feature of naturally occurring HIT antibodies is their cross-reactivity with complexes composed of PF4 and other sulfated GAGs.31,32 Similarly, KKO bound to complexes formed between hPF4 (10 μg/mL) and 0-500 μg/mL chondroitin sulfate A, chondroitin sulfate B, or dermatan sulfate, chondroitin sulfate C, heparan sulfate, and dextran sulfate (M_r 8000) following a pattern similar to that reported previously for HIT antibodies33 (Figure2).

Similar to HIT-IgG,34 KKO bound to cultured endothelial cells (Figure 3) and CHO cells (not shown) in the presence of exogenous hPF4 but not to CHO cells lacking heparan sulfate- or chondroitin sulfate-containing proteoglycans under the same conditions (data not shown). Binding of KKO to HUVECs was inhibited by heparin at concentrations (0.2 U/mL) known to dissociate PF4 from the cell surface (Figure 3). In comparison to KKO, equimolar concentrations of RTO bound one-third as well to HUVEC in the presence of either PF4 or PF4 and heparin (data not shown), consistent with ELISA data shown in Figures 1A and 1B. Whereas binding of KKO to PF4 was enhanced in the presence of heparin (Figure 1A), binding of RTO was diminished (Figure1B), suggesting that RTO recognized an epitope on PF4 that was masked or altered by heparin or heparinlike molecules.

HIT antibodies27 and immune complexes containing murine IgG₂ antibodies activate human platelets through a process that requires FcγRIIA.35,36 To determine whether KKO activates platelets through a similar pathway,¹⁴C-serotonin-labeled platelets were incubated with KKO (30-320 μg/mL) in the presence of either hPF4, heparin, or hPF4/heparin. KKO (80 and 160 μg/mL) stimulated¹⁴C-serotonin release in a heparin-dependent manner (Figure4) when preincubated with hPF4 (10 μg/mL). Somewhat higher concentrations of antibody (KKO > 180 μg/mL) were required to initiate serotonin release when the antibody was preincubated with hPF4 (10 μg/mL) complexed to heparin (1 U/mL) (data not shown). Neither KKO nor isotype control activated platelets in the presence of buffer alone or heparin alone (data not shown).¹⁴C-serotonin release induced by KKO was almost completely inhibited by the FcγRIIA-specific mAb IV.3 (less than 5% release at 0.5, 1, and 5 U/mL heparin) (data not shown). RTO did not effect serotonin release in the presence or absence of hPF4 or hPF4/heparin at all concentrations tested (31-250 μg/mL).

We previously reported that a subset of HIT/HITT antibodies require an epitope in the third domain of hPF4 to bind in the presence of heparin. This region was defined using chimeras between hPF4 and the structurally related chemokine, NAP-2, which is not recognized by HIT antibodies.18 We now have investigated both the specificity of this subset of HIT antibodies in greater detail and the involvement of this domain in the binding of KKO. To do so, single amino acid substitutions were introduced into the third domain of PF4, between Cys36 and Cys⁵². Each PF4 variant was incubated with an optimal concentration of heparin18 and the binding of 23 HIT sera and KKO was measured.

Binding of HIT antibodies to heparin complexed with the PF4 variants P37N, T38Q, A39V, L41V, and N47D was moderately impaired compared with wild-type PF4 (see Table 1), whereas binding to NAP-2 was minimal. Binding of antibody to the hPF4 variants was more variable than to wild-type protein, suggesting differences in the proportion of antibodies in individual sera sensitive to changes in the third domain of the protein. HIT sera could be divided arbitrarily into subgroups showing marked (A₄₀₅ < 1.0), intermediate (A₄₀₅ ≥ 1.0 - < 2.5), or little (A₄₀₅  >2.5), sensitivity to the P37N mutation. Binding of HIT antibodies to the other mutations in the third domain (T38Q, A39V, L41V, and N47D) generally showed the same pattern of reactivity (ie, sera containing antibodies that recognized P37N also recognized T38Q, and so on). These data suggest that there are at least 2 antigenic sites recognized by HIT sera. Sera from most patients have antibodies that recognize both sites, but a subgroup is composed of antibodies that predominantly recognize an antigenic site (or sites) involving the third domain of hPF4.

We then characterized the antigenic site recognized by KKO. We reported previously that more than 95% of HIT antibodies do not recognize IL-8 or NAP-2 complexed to heparin,18,19 though they share approximately 30% and approximately 60% amino acid homology, respectively, with hPF4. KKO shared this characteristic and did not bind either chemokine complexed to heparin (Figure5A). In addition, none of the 23 HIT sera tested (data not shown) or KKO reacted with murine PF4 complexed to heparin (Figure 5A), in spite of mPF4 displaying approximately 80% amino acid sequence identity with its human homologue. Finally, KKO bound strongly to the PF4 variants, P37N, T38Q, A39V, and L41V in the presence of heparin (Figure 5A), similar to the behavior of the subgroup of HIT sera characterized as insensitive to mutations in the third domain (see Table 1).

In light of these findings, cross-competition experiments were then performed to determine whether KKO and this subset of HIT antibodies recognized an overlapping site in PF4. To do so, binding of KKO to hPF4/heparin was measured in the presence of increasing amounts of 4 HIT sera insensitive to mutations in the third domain (data shown in black, Figure 5B) and 5 sensitive sera (data shown in gray, Figure 5B). Three of 4 insensitive serum samples inhibited the binding of KKO to hPF4/heparin by more than 50%. In contrast, none of the 5 sera designated sensitive inhibited binding of KKO to a similar extent. These data are consistent with the pattern of KKO binding to the third domain mutants shown in Figure 5A, and they suggest further that KKO recognizes an epitope that overlaps with 1 recognized by a subset of HIT antibodies.

The predicted amino acid sequence of KKO, a PF4/heparin complex-specific mAb, and RTO, a nonheparin dependent anti-PF4 mAb, were compared (Figures 6A and 6B). Sequence analysis revealed the use of very disparate V_H families and J_H gene segments and V_L families and J_L gene segments for KKO and RTO heavy and light chains, respectively. Although similarities in the idiotopes expressed by the 2 antibodies based on their primary heavy and light chain sequences cannot be ruled out, it is clear that KKO and RTO are not genetically (or clonally) related nor do they bear any obvious predicted structural homology to each other.

We generated an mAb, KKO, that shares important serologic and functional properties with naturally occurring anti-PF4/heparin antibodies found in patients with heparin-induced thrombocytopenia/thrombosis. KKO recognizes PF4 in complex with heparin over a narrow range of molar ratios approaching 1:1, similar to the behavior of naturally occurring HIT antibodies.37 KKO also recognizes complexes between PF4 and other GAGs, but not heparin itself or heparin complexed with either mPF4, NAP-2, or IL-8, all features shared with more than 95% of HIT antibodies. KKO binds to cells that express GAGs, such as endothelial cells, only when PF4 is provided exogenously,17,34,38 but it does not bind to cells that lack GAGs and that are therefore unable to form the requisite antigenic complex. KKO also activates human platelets through a heparin- and PF4-dependent mechanism that is mediated through FcγRIIA.27 

Binding of KKO to PF4/heparin exceeds binding to PF4 alone by more than 400 times under optimal conditions. However, the finding that KKO binds to PF4 at high concentrations is consistent with the hypothesis that heparin and other GAGs induce a conformational change in the protein that generates neoepitopes within the PF4 molecule that are responsible for antibody formation.18,20,39 The fact that mice injected with heparin and PF4 generate complex-specific antibodies speaks to the immunogenicity of these putative neoepitopes and may have implications for the mechanism by which antibody formation is stimulated in humans.

The results of several studies indicate that naturally occurring anti-PF4/heparin antibodies are polyspecific.12,18 KKO competes with a subset of these antibodies for binding to PF4/heparin. We previously identified a region in the third domain of PF4 that is required for recognition by a subset of HIT/HITT antibodies.18Additional mutations in this region affirm its contribution to the immunodominant epitope recognized by some, but not other, naturally occurring antibodies. The results of direct binding studies to these variant PF4 molecules, as well as competition studies using HIT/HITT sera, subdivided by their capacity to recognize these variants, suggest the epitope for KKO lies outside this domain. Sequence comparison of KKO, a heparin-dependent antibody to PF4, with RTO, a PF4-specific antibody, does not suggest any obvious genetic relatedness or structural similarity between the combining regions of antibodies themselves or any relationship between their binding sites on PF4.

The likelihood for thrombosis to develop in patients with HIT has been attributed to antibody-mediated platelet activation in vivo.27,40 Of interest, the concentrations of KKO required to activate platelets (≥ 80 μg/mL) greatly exceed the minimal amount of KKO needed to detect PF4/heparin by ELISA (0.036 μg/mL). In addition to reflecting the difference in sensitivity of the 2 assays,6,41 this finding may indicate a threshold of Fc receptor occupancy that must be exceeded to initiate platelet activation. Naturally occurring differences either in the expression of FcγRIIA,42,43 the extent to which receptor expression is up-regulated when platelets are activated,43,44 the presence of higher titers of IgG antibodies in patients with thrombosis,11 or the poorly understood role of the platelet FcγRIIA-H/R¹³¹ polymorphism may each contribute to the tendency for overt disease to develop in some patients with sensitization to PF4/heparin. Platelet activation by KKO may be limited by the number and the conformation of the PF4/heparin complexes that can bind to the platelet surface and the affinity of murine IgG_2b for human FcγRIIA.

Although the relationship between plasma PF4/heparin-specific antibodies and HIT is well established, there is no formal proof that these antibodies, or a subset of these antibodies, cause thrombocytopenia or thrombosis. We have reported previously that mice immunized with HIT-IgG develop murine antibodies to human PF4/heparin complexes and PF4-dependent endothelial cell-reactive antibodies as part of the idiotype/anti-idiotype network.45 Immunized mice developed thrombocytopenia when exposed to heparin. However, because these mice likely developed antibodies with additional specificities as a result of epitope spread, the contribution of PF4/heparin antibodies per se remains unproved. Generation of KKO and other PF4/heparin-specific mAbs will facilitate efforts to identify the pathophysiological role of anti-PF4/heparin antibodies in vivo. It is also anticipated that KKO and related antibodies will help to elucidate the physiologic role of PF4 in hemostasis, serve as a reagent standard for diagnostic assays, and provide a platform to identify therapeutic alternatives for patients with HIT/HITT.

We are indebted to Joyce Lehner and Charles Hart of ZymoGenetics, Inc, Seattle, WA, for their technical advice on monoclonal antibody preparations. We thank Tara Hendry-Hofer for her excellent technical assistance.