Investigation of the Possible Pharmacologically Active Forms of the Nicotinic Acetylcholine Receptor Agonist Anabaseine

Three major forms of the nicotinic agonist toxin anabaseine (cyclic iminium, cyclic imine and the monocationic open-chain ammonium-ketone) co-exist in almost equal concentrations at physiological pH. We asked the question: Which of these forms is pharmacologically active? First, we investigated the pH dependence of anabaseine inhibition of [3H]-methylcarbamylcholine binding at rat brain α4β2 nicotinic acetylcholine receptors (nAChRs). These experiments indicated that one or both monocationic forms interact with the orthosteric binding site for ACh. However, since they occur at equal concentrations near physiological pH, we employed another approach, preparing a stable analog of each form and examining its agonist activities and binding affinities at several vertebrate brain and neuromuscular nAChRs. Only 2-(3-pyridyl)-1,4,5,6-tetrahydropyrimidine monohydrogen chloride (PTHP), the cyclic iminium analog, displayed nAChR potencies and binding affinities similar to anabaseine. The cyclic imine analog 2,3′-bipyridyl and the open-chain ammonium-ketone analog 5-methylamino-1-(3-pyridyl)-1-pentanone (MAPP), displayed ≤1% of the activity predicted if the one form was solely active. The lower potency of weakly basic 2,3′-bipyridyl can be explained by the presence of a small concentration of its monocationic form. Since the open chain ammonium-ketone monocationic form of anabaseine has some structural similarity to the neurotransmitter GABA, we also tested the ability of anabaseine and its 1,2-dehydropyrrolidinyl analog myosmine to activate a mammalian GABAA receptor, but no activity was detected. We conclude that the monocationic cyclic iminium is the form which avidly binds and activates vertebrate nAChRs.


Introduction
Over eighty years ago, the Belgian pharmacologist Z. M. Bacq discovered the presence of a nicotine-like substance in a marine worm, the Atlantic hoplonemertine Amphiporus lactifloreus [1,2].

Initially, we measured the pH-dependent binding of [ 3 H]-methylcarbamylcholine ([ 3 H]-MCC)
to rat brain α4β2 nAChRs. In the absence of anabaseine we found a large (approximately 3-fold) enhancement of [ 3 H]-MCC binding as pH increased from 6.0 to 7. 8

. While interesting in its own regard, since MCC is permanently ionized, this marked pH dependence of [ 3 H]-MCC nAChR affinity complicated the analysis of the effect of pH on anabaseine binding, which was indirectly measured by [ 3 H]-MCC displacement. Anabaseine inhibition of [ 3 H]-MCC binding at each pH in Figure 2 is expressed with respect to the binding of [ 3 H]-MCC alone at the same pH, so as to compensate for [ 3 H]-MCC binding pH dependence. If only one or both monocationic forms of anabaseine are active, one would anticipate that inhibition of [ 3 H]-MCC binding by the fixed concentration of anabaseine would be greatest at the low pHs and progressively decrease as the pH approaches and exceeds the pKa of its tertiary amine (iminium) group. However, if only the cyclic imine form binds with high affinity one would expect the opposite effect, an increase in the inhibition of [ 3 H]-MCC binding as the pH approaches and exceeds the tertiary amine pKa. Anabaseine displacement of [ 3 H]-MCC
binding did decrease with increased pH, as expected for an active monocationic form, but the decrease was quantitatively less than would be predicted. Going from 5.8 to 7.8, the unionized cyclic imine (I) concentration was predicted [31]to increase approximately 20-fold, but the concentration of each monocationic form (I + or AK + ) would decrease ~4-fold. The pH dependence of anabaseine inhibition of MCC binding was not as marked as predicted for the monocationic form, but was certainly more consistent with the hypothesis that one or more monocationic forms of anabaseine is binding to this brain receptor. One additional complication in the interpretation of the results in Figure 2 is that specific receptor binding (and its inhibition) is not necessarily proportional to the displacing ligand (anabaseine) concentration, as it depends on a saturable binding isotherm (For instance, a Michaelis-Menton equation relating competing ligand concentrations and their respective equilibrium dissociation constants, which may not have the same pH dependence. The concentration Figure 1. A: Structures of the three major forms of anabaseine that exist in dynamic equilibrium under physiological conditions; B: Anabaseine analogs synthesized to represent the particular form shown above. MAPP (also called N-methyl anabaseine) can still occur as a cyclic imine or iminium, but the equilibrium greatly (≥ 20-fold) favors the ammonium-ketone monocation form.

Results
Initially, we measured the pH-dependent binding of [ 3 H]-methylcarbamylcholine ([ 3 H]-MCC) to rat brain α4β2 nAChRs. In the absence of anabaseine we found a large (approximately 3-fold) enhancement of [ 3 H]-MCC binding as pH increased from 6.0 to 7.8. While interesting in its own regard, since MCC is permanently ionized, this marked pH dependence of [ 3 H]-MCC nAChR affinity complicated the analysis of the effect of pH on anabaseine binding, which was indirectly measured by [ 3 H]-MCC displacement. Anabaseine inhibition of [ 3 H]-MCC binding at each pH in Figure 2 is expressed with respect to the binding of [ 3 H]-MCC alone at the same pH, so as to compensate for [ 3 H]-MCC binding pH dependence. If only one or both monocationic forms of anabaseine are active, one would anticipate that inhibition of [ 3 H]-MCC binding by the fixed concentration of anabaseine would be greatest at the low pHs and progressively decrease as the pH approaches and exceeds the pKa of its tertiary amine (iminium) group. However, if only the cyclic imine form binds with high affinity one would expect the opposite effect, an increase in the inhibition of [ 3 H]-MCC binding as the pH approaches and exceeds the tertiary amine pKa. Anabaseine displacement of [ 3 H]-MCC binding did decrease with increased pH, as expected for an active monocationic form, but the decrease was quantitatively less than would be predicted. Going from 5.8 to 7.8, the unionized cyclic imine (I) concentration was predicted [31] to increase approximately 20-fold, but the concentration of each monocationic form (I + or AK + ) would decrease~4-fold. The pH dependence of anabaseine inhibition of MCC binding was not as marked as predicted for the monocationic form, but was certainly more consistent with the hypothesis that one or more monocationic forms of anabaseine is binding to this brain receptor. One additional complication in the interpretation of the results in Figure 2 is that specific receptor binding (and its inhibition) is not necessarily proportional to the displacing ligand (anabaseine) concentration, as it depends on a saturable binding isotherm (For instance, a Michaelis-Menton equation relating competing ligand concentrations and their respective equilibrium dissociation constants, which may not have the same pH dependence. The concentration of [ 3 H]-MCC was 10 nM, which was almost the same as the K d (11 nM) of this radioligand with rat brain membranes, under our experimental conditions [5].
Since the two monocationic forms show the same pH dependence and the same concentration when expressed as % of total anabaseine, another approach was required to assess their activities separately. We resorted to synthesizing and testing stable analogs that resemble the three anabaseine forms. PTHP, the tetrahydropyrimidinyl analog of the cyclic iminium form, is predicted to be >99% ionized at pH 7.4 and does not undergo ring opening within the physiological pH range considered here, since the basic (pK a >10) amidinium group, unlike the imine group, is not susceptible to hydration and subsequent ring opening. 2,3 -bipyridyl (pK a 4.4) was selected as an analog that possesses the basic structure of the unionized cyclic imine; at pH 7.4 >99.9% of the molecules of 2,3 -bipyridyl will be unionized. The monocationic ammonium-ketone open-chain form analog selected was 5-methylamino-1-(3-pyridyl)-1-pentanone, MAPP, shown in Figure 1; it mainly (>99%) exists in their open-chain form in water at physiological pH; its cyclic imine or iminium form, N-methylanabaseine, is present in very small concentration (≤5%) [31][32][33]. An unionized open-chain amino-ketone form which occurs at high pHs [32] can be neglected due to its very small predicted concentrations over the pH range 5.8 to 7.8 that is of interest here.

Mar. Drugs 2019, 17, x FOR PEER REVIEW 4 of 13
of [ 3 H]-MCC was 10 nM, which was almost the same as the Kd (11 nM) of this radioligand with rat brain membranes, under our experimental conditions [5].
Since the two monocationic forms show the same pH dependence and the same concentration when expressed as % of total anabaseine, another approach was required to assess their activities separately. We resorted to synthesizing and testing stable analogs that resemble the three anabaseine forms. PTHP, the tetrahydropyrimidinyl analog of the cyclic iminium form, is predicted to be >99% ionized at pH 7.4 and does not undergo ring opening within the physiological pH range considered here, since the basic (pKa >10) amidinium group, unlike the imine group, is not susceptible to hydration and subsequent ring opening.

. Right ordinate: pH dependence of anabaseine inhibition of [ 3 H]-MCC binding to rat brain α4β2 receptors (Standard error bars included). Anabaseine inhibition (mean ± SEM) at each pH was the average of six replicate measurements.
PTHP was the only anabaseine analog that approached anabaseine in potency and binding affinity. Its potency and affinity for both the frog and human neuromuscular receptors was slightly inferior to that of anabaseine, but its agonist activity on vertebrate nAChRs had not been reported (Please see Figures 3 and 4). The other analogs displayed less than 1% of anabaseine's potency and binding affinity. The small potency of 2,3'-bipyridyl can be interpreted as being entirely due to its monocationic form, since the pKa of the most basic nitrogen on the 2'pyridyl ring is 4.4 [34].

Figure 2.
Concentrations of the three major forms of anabaseine as a function of pH as determined by UV spectrophotometry [32]. Left Ordinate: Concentration of each form expressed as % of the total anabaseine concentration. The red curve is the estimated cyclic imine concentration and the blue curve is the concentration of the cyclic iminium or the monocationic ammonium-ketone form of anabaseine (Assuming that K H = 1.0, the concentrations of these ionized forms are equal). Right ordinate: pH dependence of anabaseine inhibition of [ 3 H]-MCC binding to rat brain α4β2 receptors (Standard error bars included). Anabaseine inhibition (mean ± SEM) at each pH was the average of six replicate measurements.
PTHP was the only anabaseine analog that approached anabaseine in potency and binding affinity. Its potency and affinity for both the frog and human neuromuscular receptors was slightly inferior to that of anabaseine, but its agonist activity on vertebrate nAChRs had not been reported (Please see Figures 3 and 4). The other analogs displayed less than 1% of anabaseine's potency and binding affinity. The small potency of 2,3 -bipyridyl can be interpreted as being entirely due to its monocationic form, since the pKa of the most basic nitrogen on the 2 pyridyl ring is 4.4 [34]. Figure 3. Activation of the human α4β2 neuronal nAChR expressed in tsA201 cells as measured by membrane depolarization using the FlexStation assay. Each point generally is the mean of four separate measurements. All responses were normalized with respect to the response of the cells to 0.5 μM epibatidine. Potency and efficacy estimates from these curves are found in Table 1.  . Activation of the human fetal (TE671 cells) skeletal muscle nAChR by the various stable anabaseine analogs measured by membrane depolarization using the FlexStation assay. Each point is generally the mean of four separate measurements. Responses were normalized to the 5 μM epibatidine response. Potency and efficacy estimates from these curves are found in Table 1. . Activation of the human fetal (TE671 cells) skeletal muscle nAChR by the various stable anabaseine analogs measured by membrane depolarization using the FlexStation assay. Each point is generally the mean of four separate measurements. Responses were normalized to the 5 µM epibatidine response. Potency and efficacy estimates from these curves are found in Table 1.
GABA A receptors are homologous structures that have a common ancestry with nAChRs. Structure-activity studies on GABA A receptors have shown that both the ammonium and carboxy groups are not absolutely required for agonist activity and the intermediate structure between these two ionizable groups can vary considerably as well-in some cases flexibility is minimized by a ring structure, such as in the agonist muscimol, and in other analogs the hydrocarbon chain is shortened or lengthened one methylene unit without loss of agonist activity. Since the open-chain cationic form of anabaseine resembles GABA by having a primary ammonium group that is separated a similar distance from a carbonyl group, we tested anabaseine and its pyrroline homolog myosmine on a human GABA A receptor heterologously expressed in a HEK cell line [35]. Figure 5 shows that anabaseine and myosmine are not GABA A agonists, at least on this particular GABA A receptor subtype.

Discussion
While most nAChR agonists and antagonists binding at ACh orthosteric sites are basic molecules with at least one ionizable N, there are exceptions, including lophotoxin [36], the neonicotinoids [37,38] and some recently synthesized pyrimidine agonists [39]. To optimize a molecule to serve as a drug or selective molecular probe of some receptor it is important to identify the form in which it interacts optimally with its target. In the case of anabaseine, our identification of the cyclic iminium form as the most active (and possibly the only) form suggests that structural modifications that enhance the basicity of its tetrahydropyridyl nitrogen will enhance potency and possibly selectivity for the intended receptor target. In fact, a major improvement in stability occurs with addition at the 3-position of the tetrahydropyridyl ring of an electron-conjugated system containing an aromatic ring, as in 3-(4-dimethylaminobenzylidene)-anabaseine [40] or 3-(2,4-dimethoxybenzylidene)-anabaseine, also called GTS-21 [41]. These anabaseine derivatives do not undergo ring opening (hydrolysis of the imine bond) under physiological conditions, due to their extended π electron conjugation [42].
Previously, it was assumed (but not experimentally tested) that the cyclic iminium form of anabaseine is the one which binds avidly to the ACh binding site in nAChRs [8]. This form was modeled to fit into a muscle type homology model [43]. The crystal structure of the molluscan acetylcholine binding protein binding anabaseine [44] actually was derived from crystals containing both the cyclic iminium and the open-chain monocationic form of anabaseine occupying some of the five identical binding sites. Unlike the cyclic iminium, the ammonium-ketone form ammonium group did not insert into the "aromatic box." In the preparation of the AChBP-anabaseine crystals very high anabaseine concentrations were used so that most of the five sites would be occupied; therefore the presence of the ammonium-ketone form in the crystal structure does not necessarily indicate that this form of anabaseine would bind to AChBPs and nAChRs at the low concentrations of anabaseine that occur in our in vitro functional and binding studies.
The ability of a variety of 2-(aryl)-1,4,5,6-tetrahydropyrmidines to block neuromuscular nAChRs and produce teratogenic effects in chick embryos was reported long ago [45,46]. Some of these compounds were reported to be nAChR agonists. The only study of PTHP on neurons was that of Upshall et al. [47] on leech Retzius cells. Our results also indicate that PTHP is a potent agonist at neuronal as well as neuromuscular nAChRs. The 2-aryl compounds reported in the Brimblecombe et al. study did not include PTHP or other compounds having a 2-(3-pyridyl)-substituent, which is known to be essential for the agonist activity of nicotine and of anabaseine. For PTHP to be equipotent with anabaseine, based on the I + form being the only active form, its EC 50 would need to be~1/3 of the EC 50 of anabaseine, since the anabaseine (I + ) is~1/3 of its total concentration. The data for PTHP and anabaseine in Figures 3 and 4 Table 1, indicate that PTHP has approximately 30% of the potency predicted for the anabaseine cyclic iminium form. One possible explanation for this slightly inferior potency is that the additional N on the tetrahydropyrmidinyl ring must form an H bond with some other atom within the aromatic box of the nAChR orthosteric binding site and this may be energetically unfavorable.  Table 1, indicate that PTHP has approximately 30% of the potency predicted for the anabaseine cyclic iminium form. One possible explanation for this slightly inferior potency is that the additional N on the tetrahydropyrmidinyl ring must form an H bond with some other atom within the aromatic box of the nAChR orthosteric binding site and this may be energetically unfavorable. While 2,3'-bipyridyl displayed only a marginal potency at vertebrate receptors that could be predicted from its pKa, it has been found to be quite paralytic to insects [48] and crustaceans [49,50]. At least some arthropod nAChRs are known to not require a cationic ligand for activation, as shown by many studies with neonicotinoid insecticides such as imidachloprid [37]. These neonicotinoids also display a limited activity at vertebrate nAChRs [38]. It seems likely that 2,3'-bipyridyl affects some arthropod nAChRs through its unionized form.

The ability of a variety of 2-(aryl)-1,4,5,6-tetrahydropyrmidines to block neuromuscular nAChRs and produce teratogenic effects in chick embryos was reported long ago [45,46]. Some of these compounds were reported to be nAChR agonists. The only study of PTHP on neurons was that of Upshall et al [47] on leech Retzius cells. Our results also indicate that PTHP is a potent agonist at neuronal as well as neuromuscular nAChRs. The 2-aryl compounds reported in the Brimblecombe et al study did not include PTHP or other compounds having a 2-(3-pyridyl)-substituent, which is known to be essential for the agonist activity of nicotine and of anabaseine. For PTHP to be equipotent with anabaseine, based on the I + form being the only active form, its EC50 would need to be ~1/3 of the EC50 of anabaseine, since the anabaseine (I + ) is ~1/3 of its total concentration. The data for PTHP and anabaseine in Figures 3 and 4 and in
One can speculate as to what advantage(s) might accrue to an organism (in this case, a hoplonemertine or ant) to produce and secrete an imine-bond containing toxin like anabaseine. The entire body wall integument as well as the contiguous epithelium of the anterior proboscis of P. peregrina contains very high concentrations of anabaseine [5]. The proboscis everts and wraps around the prey during its capture allowing extensive contact of this epithelium with its annelid prey. The mineralized stylet of the proboscis can be observed to produce multiple punctures of the prey integument, which should further facilitate envenomation. In addition, potential predators will also be exposed in these ways to anabaseine and related alkaloids. Perhaps a major advantage of possessing an imine bond, besides an enhanced affinity for the nAChR orthosteric site [8], is its low Figure 5. Investigation of possible activation of the human GABA A receptor measured by membrane depolarization using a FlexStation assay. Each point is the mean of three separate measurements. All responses were normalized with respect to the response of the cells to 1 mM µM GABA. The EC 50 value for GABA was 3.0 µM (95% confidence interval = 1.8-5.1 µM). While 2,3 -bipyridyl displayed only a marginal potency at vertebrate receptors that could be predicted from its pK a , it has been found to be quite paralytic to insects [48] and crustaceans [49,50]. At least some arthropod nAChRs are known to not require a cationic ligand for activation, as shown by many studies with neonicotinoid insecticides such as imidachloprid [37]. These neonicotinoids also display a limited activity at vertebrate nAChRs [38]. It seems likely that 2,3 -bipyridyl affects some arthropod nAChRs through its unionized form.
One can speculate as to what advantage(s) might accrue to an organism (in this case, a hoplonemertine or ant) to produce and secrete an imine-bond containing toxin like anabaseine. The entire body wall integument as well as the contiguous epithelium of the anterior proboscis of P. peregrina contains very high concentrations of anabaseine [5]. The proboscis everts and wraps around the prey during its capture allowing extensive contact of this epithelium with its annelid prey. The mineralized stylet of the proboscis can be observed to produce multiple punctures of the prey integument, which should further facilitate envenomation. In addition, potential predators will also be exposed in these ways to anabaseine and related alkaloids. Perhaps a major advantage of possessing an imine bond, besides an enhanced affinity for the nAChR orthosteric site [8], is its low basicity, which allows the unionized form to reach a relatively high concentration in the neutral pH range of the marine environment and thereby enhance passive diffusion across the integument of the prey or predator. The integument of a marine organism exposed to sea water is likely to have a pH near 8, well above the pH range in which anabaseine is primarily ionized. Thus, the toxin has a much better chance of entering the prey/predator than would anabaseine, the more basic secondary amine analog of anabaseine.
Another consideration is how anabaseine is packaged within the venom gland cells [51]. It seems most likely that it is accumulated within secretory vesicles, which are generally acidic (pH 5-7). The ionized forms, especially the very polar open-chain forms, would be much less likely to diffuse out of the vesicles than would the cyclic imine form, so their dominance under these acidic conditions should enhance the energy efficiency of anabaseine storage.
Thus, it is possible that each of the coexisting forms of anabaseine have functional significance, not only for understanding how this potent nAChR agonist interacts with its receptors, but they also may play different roles in storage and release of the toxin by the nemertine.
We have synthesized other anabaseine compounds whose properties will be reported in the future (Kem et al., in preparation). The current paper predicts that their relative pharmacological potencies will be affected by the extent that the cyclic iminium form dominates the equilibrium between these different forms. 3-Pyridylamidine hydrochloride (PTHP Precursor): Our procedure was similar to that described by Schaefer and Peters [52] and Brown and Evans [53]  I]-α-BTX) were used in binding experiments with whole rat brain and TE671 cell membranes, respectively, according to Kem et al. [41]. Rat brain membranes (200 µg of protein) on TsA201 cell membranes (100 µg protein) expressing human α4β2 nAChRs were incubated with 0.5 nM [ 3 H]-cytisine in a final volume of 500 µL of binding saline for 4 h at 5 • C. The experiments on α7 nAChRs rat brain membranes involved incubation with 0.5 nM [ 125 I]-α-Btx for 3 h at 37 • C to assure that equilibrium was reached. Eight different concentrations of the experimental compound were usually tested in triplicate. Nonspecific binding was measured in the presence of 1 mM (S)-nicotine hydrogen tartrate (Sigma-Aldrich). Data were fitted using GraphPad Prism software (Version 4 GraphPad Software, San Diego, CA, USA) by nonlinear regression analyses to a sigmoidal one-site model with variable slope. Compound affinity for Torpedo electric organ membrane nAChRs was assessed by inhibition of 0.5 nM [ 125 I]-α-BTX binding over a 30 min incubation period at 5 • C, since α-Btx binds irreversibly to skeletal muscle nAChRs. The IC 50 was also estimated by the same Prism software.

Cell Culture
TsA201 cells expressing human α4β2 were maintained in media consisting of Dulbecco's Modified Eagle medium supplemented with 10% FBS, 100 units/mL penicillin and 100 µg/mL streptomycin, 2 mM l-glutamine, 0.5 mg/mL zeocin and 0.6 mg/mL geneticin [55]. Cells were grown in 75 cm 2 culture flasks, which were housed in a humidified incubator (Fisher Scientific, Atlanta, GA, USA at 37 • C in an atmosphere of 5% CO 2 . They were grown to around 80-90% confluence after harvesting with 0.25% trypsin and being split weekly at a subcultivation ratio of between 1:6 and 1:10.

Nicotinic Receptor FlexStation Functional Assays
Our experimental protocol was based on the initial study of Fitch et al. [56]. Cells were seeded at a density of roughly 5 × 10 4 to 10 5 cells/well in 96-well flat-bottom black wall culture plates coated with 50 µg /ml poly-d-lysine hydrobromide (Sigma-Aldrich, 70-150 kDa) and grown overnight in 100 µL culture medium. A proprietary membrane potential dye obtained from Molecular Devices (San Diego, CA, USA) was prepared by dissolving one bottle of dye into 30 mL of Hanks Saline (pH = 7.4) containing 20 mM HEPES buffer. The cells were incubated with 100 µL of dye for 30 min at 37 • C prior to the robotically controlled concentration-response experiment. Serial dilutions of a compound for dose-response analysis were prepared in 96-well plates by evaporation of a methanolic stock solution and then reconstituted in the appropriate volume of Hanks saline. Fluid transfer and readings were performed by a FlexStation fluorimeter (Molecular Devices). Excitation and emission wavelengths were set to 530 nm and 565 nm with a cutoff of 550 nm. The first 17 s were used as a basal reading. At 18 s, a test compound was added to determine the EC 50 , followed by addition of 25 µL KCl (40 mM final concentration) at 160 s to serve as a fluorescence calibrant. Compounds that had no measurable depolarizing activity on the TsA201 cells were then tested for their ability to inhibit a 5 µM control ACh response. Data were fitted and graphed with Prism (GraphPad) to determine the EC 50 .

GABA Receptor Flexstation Funcional Assays
WSS1 cells from ATCC (Manassas, VA, USA) which express functional GABA A receptors were cultured as previously described by Wong et al. [57]. The cells were grown to near confluence and then transferred to black well, clear bottom 96-well assay plates (Corning Incorporated, Corning, NY, USA) 12-18 h prior to the assay. Molecular Devices blue dye (Sunnyvale, CA, USA) was reconstituted as previously described [58]. The WSS1 cells were allowed to come to room temperature, the culture medium was then aspirated, replaced with the blue dye and equilibrated for 30 min before being placed on a FlexStation 3 plate reader (Molecular Devices, Sunnyvale, CA, USA). Varying concentrations of γ-aminobutyric acid (GABA, Tocris Bioscience, Bristol, United Kingdom), myosmine (Sigma, St. Louis, MO, USA) and anabaseine [31,32] in blue dye solution were added to the cells by the FlexStation. Programmed readings and data analysis were performed as previously described [58,59]. The cellular responses to GABA were normalized to the maximum dye response generated by 1 mM GABA. The fifty percent effective concentration for GABA was determined using a sigmoidal dose-response equation (log(agonist) vs. normalized response-Variable slope) with Prism version 6.03 (GraphPad Software, Company, San Diego, CA, USA, www.graphpad.com).