Differentiated Human Midbrain-Derived Neural Progenitor Cells Express Excitatory Strychnine-Sensitive Glycine Receptors Containing α2β Subunits

Background Human fetal midbrain-derived neural progenitor cells (NPCs) may deliver a tissue source for drug screening and regenerative cell therapy to treat Parkinson’s disease. While glutamate and GABAA receptors play an important role in neurogenesis, the involvement of glycine receptors during human neurogenesis and dopaminergic differentiation as well as their molecular and functional characteristics in NPCs are largely unknown. Methodology/Principal Findings Here we investigated NPCs in respect to their glycine receptor function and subunit expression using electrophysiology, calcium imaging, immunocytochemistry, and quantitative real-time PCR. Whole-cell recordings demonstrate the ability of NPCs to express functional strychnine-sensitive glycine receptors after differentiation for 3 weeks in vitro. Pharmacological and molecular analyses indicate a predominance of glycine receptor heteromers containing α2β subunits. Intracellular calcium measurements of differentiated NPCs suggest that glycine evokes depolarisations mediated by strychnine-sensitive glycine receptors and not by D-serine-sensitive excitatory glycine receptors. Culturing NPCs with additional glycine, the glycine-receptor antagonist strychnine, or the Na+-K+-Cl− co-transporter 1 (NKCC1)-inhibitor bumetanide did not significantly influence cell proliferation and differentiation in vitro. Conclusions/Significance These data indicate that NPCs derived from human fetal midbrain tissue acquire essential glycine receptor properties during neuronal maturation. However, glycine receptors seem to have a limited functional impact on neurogenesis and dopaminergic differentiation of NPCs in vitro.


Introduction
Glycine is an important inhibitory neurotransmitter in the adult central nervous system acting through ionotropic glycine receptors that are most prominently expressed in the brainstem and spinal cord [1][2][3]. These receptors belong to the superfamily of Cysloop receptors such as GABA A , nicotinic acetylcholine, and 5-HT 3 receptors [4]. As other members of this Cys-loop family, glycine receptors form homomeric or heteromeric pentamers with each of the five subunits arranged around a central ion-conducting pore [5][6][7]. Similar to GABA A receptors in the adult central nervous system, the strychnine-sensitive glycine receptors are involved in regulating inhibitory chloride influx to stabilise the resting membrane potential of neurons. In humans, only four functional glycine receptor subunits have been identified, a1-3 and b [8], which are likely to exist in heteromeric ab combinations [3]. The a4 subunit gene is a pseudo-gene in humans and there is little evidence for its functional expression in rats [7,9]. Besides inhibitory glycine receptors, a strychnine-insensitive excitatory glycine receptor contains NMDA receptor subunits from the NR3 family [10].
Defects in glycinergic neurotransmission can result in the neurological motor disorder hyperekplexia which is characterised by a disinhibited startle response [11,12]. This primary startle syndrome is mostly autosomal dominant and is caused by mutations in the glycine transporter, the a1 subunit, or less frequently by mutations in the b subunit [13][14][15][16]. However, the genetic basis of many cases of hyperekplexia and paroxysmal movement disorders remains unresolved [17]. Glycine receptors containing the a3 subunit in the spinal cord have been recognised as therapeutic target to treat inflammatory pain syndromes [18].
Glycine receptors play a major role in the excitability of spinal cord and brain stem neurons. During development, the receptor properties undergo molecular changes resulting in modifications of their physiological function. In rats, a developmental switch from a2 homomeric glycine receptors to a1b heteromeric subtypes occurs between birth and the third postnatal week [7]. The knockout of the a2 subunit, however, has no obvious effect on the behavioural phenotype and neuronal development although it eliminates a tonic glycine-gated chloride current in mouse embryonic cortical neurons [19]. The receptor subtype expression in mouse spinal neurons during in vitro development switches similarly from a monomeric a subunit or heteromeric a2b in immature neurons to an a1b isoform in mature neurons. Furthermore, the formation of postsynaptic glycine receptor clusters as well as the receptor affinity to glycine, strychnine, and Zn 2+ increases during development [20].
In contrast to the adult central nervous system, a high expression of the Na + -K + -Cl 2 co-transporter 1 (NKCC1, a Cl 2 importer) and a low expression of the K + -Cl 2 co-transporter 2 (KCC2, a Cl 2 exporter) in neural progenitors and immature neurons determine a high intracellular Cl 2 concentration leading to GABA-induced depolarisations [21][22][23][24]. Glycine-induced activation of strychnine-sensitive glycine receptors can also lead to hyper-or depolarising responses of the target cells depending on the intracellular Cl 2 concentration [7]. During neocortical development a depolarising glycine-gated Cl 2 efflux stimulates the calcium influx [25] necessary for the development of numerous neuronal specialisations including glycinergic synapses [26]. However, the involvement of glycine receptors in human neurogenesis and dopaminergic differentiation as well as their molecular and functional characteristics in human neural progenitor cells (NPCs) are largely unknown.
The proliferation and differentiation of NPCs enables to study human neurogenesis in vitro [24,[27][28][29][30][31][32] which shall help to translate neural stem cell therapy for neurodegenerative diseases [33][34][35]. Long-term expanded human mesencephalic NPCs maintain their proliferative capacity and continue to give rise to neurons that express tyrosine hydroxylase (TH) and also release dopamine [36]. The present study analyses human midbrainderived NPCs in respect to their glycine receptor expression and function. We also investigated whether glycine, the glycinereceptor antagonist strychnine, or the NKCC1-inhibitor bumetanide are able to influence neurogenesis and dopaminergic differentiation in vitro.

Cell Culture
Human neural progenitor cells were derived from CNS tissue of aborted human fetuses (gestational week [10][11][12][13][14][15][16] with the informed written consent of all mothers involved in this study. All experiments were approved by the Ethics Committees of the University of Leipzig and the Hannover Medical School, Germany and are in accordance with all state and federal guidelines. The human fetal tissue was washed with sterile Hank's buffered salt solution and dissected into mesencephalic and nonmesencephalic primary tissue samples. The tissue samples were mechanically separated into small pieces, incubated in 0.1 mg/ml papain solution (Roche, Mannheim, Germany), supplemented with 10 mg/ml DNase (Roche) for 30 min at 37uC, then washed three times with Hank's buffered salt solution followed by an incubation with 50 mg/ml antipain solution (Roche) for 30 min at 37uC. After three further washing steps the samples were homogenised by gentle trituration using fire-polished pasteur pipettes. The quality of the tissue was assessed as described previously [27,28].
All solvents and chemicals for pharmacological experiments were purchased from Sigma or Tocris (Germany). The stock solutions were prepared in DMSO or external recording solution as appropriate (1-300 mM). A fresh stock solution of tropisetron (1 mM) was prepared at the day of experiments. The drugs were dissolved in external solution containing DMSO at a maximal final concentration of 0.1%. All drugs were applied rapidly via gravity using a modified SF-77B perfusion fast-step system (Warner Instruments Inc., Hamden, CT, USA) as described previously [39]. For the glycine dose-response curve seven increasing concentrations (10 mM-10 mM) were applied for 2 sec on NPCs. For pharmacological characterisation of glycine receptors, positive and negative modulators were co-applied for 2 sec with an EC 70 of glycine (300 mM). The intervals between applications were 30 sec, after co-applying strychnine 1 min intervals were allowed for wash out.
Whole-cell currents were low-pass filtered at 1-5 kHz, digitized at 10 kHz, and analysed with PulseFit (HEKA) and GraphPad Prism (GraphPad Software, San Diego, CA, USA). Peak currents of each investigated cell were normalised to the maximal glycineevoked peak current (for glycine dose-response curves) or to the glycine EC 70 control that was applied prior to the co-application of each tested modulator. To obtain nonlinear regression concentration-response plots mean peak currents 6 SEM were fitted to a sigmoidal function using a four parameter logistic equation (sigmoidal concentration-response) with a variable slope. The equation used to fit the concentration-response relationship was I = I max /1+10 (LogEC502Logdrug)xHill slope where I was the peak current at a given concentration. Numerical data of all experiments were expressed as means 6 SEM. Statistical differences were calculated by using Student's t test (two tailed, unpaired) and considered significant at p,0.05 (Table 1).

Quantitative Real-time PCR
After 1 and 3 weeks of differentiation in vitro, human mesencephalic NPCs from 3 cell lines were harvested and total RNA isolated using the Trizol reagent (Invitrogen). Reverse transcription of 800 ng total RNA per reaction was carried out using oligo-dT primer and Superscript II reverse transcriptase (Invitrogen).
Quantitative real-time PCR was done using cDNA from 30 ng total RNA, 0.6 mM forward and reverse primers, Platinum-SYBR GreenH qPCR Supermix (Invitrogen), and 100 nM 6-carboxy-Xrhodamine (ROX) using the following protocol in an MX 3000P instrument (Stratagene, La Jolla, CA, USA): 2 min 50uC, 2 min 95uC and 50 cycles of 15 sec 95uC, 30 sec 60uC. To confirm a single amplicon a product melting curve was recorded. The correct amplicon size was asserted by agarose gel electrophoresis using low molecular weight DNA ladder (New England Biolabs, Ipswich, MA, USA). Oligonucleotide primers for human glycine receptor subunits a1-4 and b (Table 1) were designed to flank intron sequences, if feasible, using Primer 3 software.
Cycle threshold (Ct) values were placed within the exponential phase of the PCR as described previously [40]. Ct values of 3 independent experiments, each performed in duplicate, were normalised to b2-microglobulin (Ct -Ct b2-microglobulin = DCt). DCt values were used to calculate the relative subunit expression levels and are given as means 6 SEM (Table 2). Note, a low DCt value represents a high expression level. The expression of glycine receptor subunits was statistically evaluated by subjecting DCt values to a one-way ANOVA and Newman-Keuls post test for multiple comparisons taking statistical significance as p,0.05.

Calcium Imaging
Measurements of the intracellular Ca 2+ concentration [Ca 2+ ] i in human mesencephalic NPCs were performed as described previously [24] using a monochromator-based imaging system (TILL Photonics, Grä felfing, Germany). Cells were loaded with 5 mM fura-2-AM (Invitrogen) supplemented with 0.01% Pluronic F127 for 30 min at 20-22uC in a standard bath solution containing (mM): 140 NaCl, 5 KCl, 2 CaCl 2 , 10 glucose and 10 HEPES, adjusted to pH 7.4 with NaOH. For measurements of [Ca 2+ ] i , cells were placed in a recording chamber and continuously perfused with standard bath solution at a rate of 5 ml/min. Fluorescence was excited at 340 and 380 nm and emitted fluorescence intensities from single cells were acquired at intervals of 2 s.

Drug Treatment of NPCs
To investigate the influence of additional glycine, the glycinereceptor antagonist strychnine, and the NKCC1-inhibitor bumetanide on NPCs in vitro, separate experiments were performed in which some cells were treated with two distinct concentrations of glycine (1 and 10 mM), strychnine, or bumetanide (1 and 10 mM) during proliferation (2 weeks) and differentiation (1 and 3 weeks). The stock solution was prepared by dissolving glycine in distilled water at a concentration of 100 mM, strychnine in ethanol and bumetanide in DMSO at a concentration of 10 mM. The drugs were renewed with every media change. Besides using the standard differentiation medium, NPCs were also differentiated for 1 week by a novel mitogen-free medium to promote dopaminergic neurogenesis consisting of Neurobasal (Invitrogen) supplemented with 2% B27 (Invitrogen), 1% Glutamax, 100 mM dbcAMP (Sigma), 10 mM forskolin (Sigma), 100 mM fusaric acid (Sigma), and 0.1% gentamycin (Invitrogen).
Staining patterns were visualised by fluorescence microscopy (Axiovert 200, Zeiss, Jena, Germany). Digital images were acquired with an AxioCam MRc camera using the image-analysis software AxioVision 4 (Zeiss). The portion of cells immunoreactive for GFAP, b-tubulin III, MAP2, and TH was determined by counting the number of immunopositive cells in relation to the number of DAPI stained nuclei. Approximately 1000 cells were counted within four randomly selected fields per well.

Results
We studied the functional and molecular glycine receptor properties of human midbrain-derived NPCs following differentiation using standard conditions for 1 or 3 weeks in vitro. The effect of glycine on Ca 2+ signalling in differentiated mesencephalic NPCs was determined using fura-2-based Ca 2+ imaging. Besides its agonistic action on glycine receptors, glycine is also a co-agonist along with glutamate for most NMDA receptors and can activate NR3 subunit containing NMDA receptors in the absence of glutamate [10]. To test for a depolarisiation-induced entry of Ca 2+ due to activation of glycine receptors as well as for a possible involvement of glycine-activated NMDA receptors, we subsequently applied glycine alone or in combination with the antagonists strychnine or D-serine. The latter, a co-agonist of conventional NMDA receptors, was shown to inhibit currents of NMDA receptors containing NR3 [10].
NPCs differentiated for 3 weeks showed glycine-induced Ca 2+ transients that were completely suppressed by the glycine receptor antagonist strychnine but were only slightly affected by D-serine (Fig. 1A). Increases in [Ca 2+ ] i were also evoked by application of 50 mM KCl, indicating depolarisation-dependent Ca 2+ entry. The percentage of cells responding to glycine and KCl determined  (Fig. 1B). These data suggest that glycine evokes Ca 2+ signals by activation of glycine receptors and receptor-dependent membrane depolarisation. Human midbrain-derived NPCs differentiated for 1 week were less responsive to application of glycin and KCl (Fig. 1C). The percentages of cells responding to glycine determined from these experiments were 5% (n = 7 of 139) and 70% (n = 60 of 85) for shorter (1 week) and longer (3 weeks) differentiation, respectively (Fig. 1D). The corresponding amounts of cells showing KClinduced Ca 2+ signals were 21% (n = 29 of 139) and 87% (n = 75 of 85) for shorter and longer differentiation, respectively (Fig. 1D). We also tested the ability of glycine to induce [Ca 2+ ] i changes in undifferentiated NPCs. However, all tested cells displayed no measurable responses to application of 100 mM glycine (n = 60 from 4 experiments, data not shown).
Patch-clamp electrophysiology was performed to measure glycine-evoked currents as well as voltage-gated sodium and potassium currents. Whole-cell recordings revealed a current response during rapid applications of glycine in 67% of differentiated NPCs (n = 53 from 79), which is similar to the percentage of glycine-responsive cells in calcium imaging experiments. The average peak sodium currents of glycine-responsive NPCs (-73 pA) were significantly smaller in comparison to cells without detectable glycine-induced current (-119 pA; Table 1). However, NPCs expressing functional glycine receptors showed a smaller cell membrane capacitance than cells without current evoked by glycine (9.1 pF and 12.0 pF, respectively; Table 1). Therefore, the relative sodium and potassium peak current densities in pA/pF were not significantly different between these groups of NPCs. Furthermore, there were no significant differences for peak potassium currents, resting membrane potentials, and input resistances between both cell types (Table 1).
Rapid applications of increasing glycine concentrations (10 mM-10 mM) on differentiated NPCs elicited desensitising currents in a dose-dependent manner (Fig. 2A). The glycine concentration-response plot (Fig. 2B) indicated an EC 50 of 99.2 mM (95% confidence interval of Log EC 50 value -4.181 to -3.826, hillslope 0.95, n = 10). Mean peak currents evoked by glycine were 241 pA641 pA (n = 53). For further pharmacological characterisation of glycine receptors, we used co-applications of 300 mM glycine corresponding to the EC 70 according to the concentration-response plot.
All glycine-induced currents of NPCs were markedly blocked by strychnine and partly inhibited by atropine, the neurosteroid pregnenolone sulphate as well as by the GABA A receptor antagonists picrotoxin and bicuculline (Fig. 3A). In contrast to the slight inhibition of glycine currents by 100 mM Zn 2+ , a small positive modulatory effect was induced by a lower Zn 2+ 2concentration (10 mM ; Fig. 3B). Furthermore, a current potentiation of the glycine EC 70 was induced by co-application of the 5-HT3receptor antagonist tropisetron. The pharmacological profile of glycine receptors in NPCs with moderate picrotoxin-sensitivity, a differential sensitivity to Zn 2+ , and a positive modulation by tropisetron, suggests the expression of heteromeric isoforms, most likely containing a2b subunits [41,42].
For quantitative expression analysis of glycine receptor subunits, human midbrain-derived NPC lines (n = 3) were investigated by real-time PCR after differentiation for 1 and 3 weeks in vitro. Statistical comparisons of DCt values (n = 3 independent experiments, each performed in duplicate) are summarised in Fig. 4. While the expression of various glycine receptor subunits was not yet markedly different after 1 week of cell maturation (Fig. 4B), we found a predominant expression of a2 and b subunits with significant difference to the other glycine receptor isoforms after differentiation for 3 weeks (Fig. 4A; p,0.001, ANOVA and Newman-Keuls post test). The expression of a2, a4 and b subunits was significantly higher in NPCs differentiated for 3 weeks than for 1 week (Table 2; p,0.05, t-test). Interestingly, b subunit expression was most pronounced and significantly different from all a subunits after 3 weeks of maturation (Fig. 4A). The expression of a1, which is the most abundant glycine receptor a subunit in the adult nervous system, was barely detectable, whereas the a4 subunit sparsely occurred and a3 was expressed to a moderate extent. In line with the pharmacological results, the quantitative PCR data indicate a predominant expression of glycine receptors containing a2 and b subunits.
Drug treatment of human mesencephalic NPCs with additional glycine (1 and 10 mM), the glycine-receptor antagonist strychnine, or the NKCC1-inhibitor bumetanide (1 and 10 mM) during expansion (2 weeks) did not induce significant differences compared to untreated controls (n = 3-6) in the number of living cells, protein content, or the cell proliferation as determined by Western blot analysis of the proliferation marker PCNA and the survival marker Bcl2 as well as by immunocytochemistry of the proliferation marker Ki67 and the neural stem cell marker nestin (data not shown). After differentiation with the standard protocol (1 and 3 weeks) or with a novel medium to promote the dopaminergic neurogenesis (1 week), Western blot and immunocytochemical results of drug treated NPCs (n = 3-6 cell lines) using the markers GFAP, b-tubulin III, MAP2, and TH were not significantly different from controls (data not shown).
The amount of untreated NPCs (n = 3 cell lines) immunopositive for TH (0.9%), b-tubulin III (34.4%), and GFAP (31.3%) after 1 week increased moderately during a total differentiation of 3 weeks with the standard protocol to 1.6%, 39.6%, and 32.6%, respectively. The number of MAP2-immunoreactive NPCs was significantly higher after 3 weeks (30.5%) than 1 week (13.7%) of maturation (n = 3, p,0.05, t-test). Using a novel medium to enhance dopaminergic neurogenesis of NPCs (n = 6 cell lines) resulted in a marked increase of TH-immunoreactive cells to 2.7% after differentiation for 1 week. Immunocytochemical stainings after 3 weeks of standard NPC-differentiation show that neuronal MAP2-expressing cells are immunopositive for glycine receptor subunits (Fig. 5).

Discussion
In this study, calcium imaging and electrophysiological recordings demonstrated the expression of excitatory strychnine-sensitive glycine receptors in differentiated human midbrain-derived NPCs. The similar voltage-gated peak currents/pF in cells with functional glycine receptors (67%) compared to glycine-nonresponsive NPCs (33%, Table 1) suggest that the maturation of functional glycine receptor properties and voltage-gated channels is not a simultaneous process contrasting GABA A receptor expression in these cells [24]. Previously, we could show that GABA induces [Ca 2+ ] i increases in NPCs due to membrane depolarisation mediated by GABA A receptors [24]. Activation of glycine receptors corresponded with depolarising effects leading to a similar rise of [Ca 2+ ] i in NPCs which could be completely inhibited by coapplication of strychnine. Calcium responses in NPCs revealed only a weak sensitivity of excitatory glycine receptors to D-serine that blocks glycine-evoked currents at NMDA receptors containing NR3 subunits [10]. Although human mesencephalic NPCs express low amounts of NR3 [32], our calcium imaging data suggest that glycine induces depolarisations by activation of strychnine-sensitive glycine receptors.
Previous approaches to analyse human fetal neural progenitors revealed intracellular Ca 2+ responses to glycine with a lower magnitude and frequency (6%) after 2 to 4 weeks of differentiation [43,44] suggesting an immature state of glycine receptor expression and function. Furthermore, a subpopulation of human corneal stem cells responded to glycine in electrophysiological experiments [45]. Strychnine-sensitive glycine receptors with a depolarising function were detected indirectly by glycine-induced suppression of inwardly-rectifying potassium channels in 80% of neural stem-like cells derived from human umbilical cord blood [46]. However, the functional relevance of glycine receptor-mediated depolarisations for neurogenesis remains unclear. High levels of glycinergic transmission may modulate neuronal excitability causing membrane depolarisation and changes in intracellular calcium that possibly regulate gene activity and affect neurite outgrowth in immature mouse spinal neurons [20].
In differentiated human mesencephalic NPCs, the pronounced expression of the importing Cl 2 co-transporter NKCC1 contrasts KCC2 expression [24] which can result in a high intracellular Cl 2 concentration and depolarising glycineeffects. Neuronal maturation is associated with a downregulation of NKCC1 and a stronger appearance of KCC2 [21][22][23] that changes the direction of the ligand-gated Cl 2 current at strychnine-sensitive glycine receptors from a depolarising to a hyperpolarising one. We intended to test the role of glycine receptors and NKCC1 for proliferation and differentiation of NPCs in vitro by culturing the cells with the NKCC1-inhibitor bumetanide (1 and 10 mM, [23]), the glycine-receptor antagonist strychnine (1 and 10 mM), or additional glycine (1 and 10 mM) while the media contained moderate glycine concentrations (250-400 mM). However, such drug treatment of human mesencephalic NPCs did not reveal significant changes compared to untreated controls regarding markers for neurogenesis and dopaminergic differentiation suggesting that glycine receptors seem to have a limited functional impact on proliferation and maturation of NPCs in vitro. The marked increase of MAP2-immunopositive cells between 1 and 3 weeks of differentiation reflects neuronal maturation that is also apparent in glycine receptor function (Fig. 1) and subunit expression (Fig. 4) as well as in voltage-gated channel, GABA A -and glutamate receptor function of NPCs [24].
Mature glycine receptors in the adult CNS display molecular structures and physiological properties different from those in the immature CNS. Immature glycine receptors are usually equipped with a2 or a2b subunits while mature receptors are characterized by their content of a1b and an increased sensitivity to several modulators [7,20]. Most studies using recombinant glycine receptors determined higher EC 50 values for isoforms containing b subunits than for the corresponding subtypes devoid of b [41,[47][48][49]. The glycine EC 50 of receptors with a2 or a2b subunits mainly ranged from 62 to 96 mM and from 66 mM to 157 mM, respectively [41,49]. For the native glycine receptors in NPCs, we calculated an EC 50 of 99.2 mM. The glycine-induced currents (EC 70 :300 mM) were all blocked by strychnine and showed a rather low sensitivity to many modulators suggesting immature receptor subtypes. In addition, the results of quantitative PCR indicated a lack of a1 expression and a2 and b as predominant subunits leading to the assumption that NPCs express glycine receptors with a2 and/or a2b subunits.
In pharmacological experiments we tried to distinguish these subtypes, although there are few compounds with sufficient discriminatory capacity to identify the presence of either homomeric a2 or heteromeric a2b glycine receptors [7]. While the glycine responses of recombinant a2 receptors are inhibited by the 5-HT3-antagonist tropisetron [41,47], the responses of a2b subtypes can be potentiated by low tropisetron concentrations [41]. In human mesencephalic NPCs, co-application of tropisetron (1 mM) enhanced glycine-induced currents indicating the presence of b subunits which are likely to contribute to this modulatory glycine receptor site. A differential sensitivity to Zn 2+ that can potentiate glycine-evoked currents at lower micromolar concentrations and inhibit recombinant heteromeric and neuronal glycine receptors at higher concentrations [42,50] could also be demonstrated for the receptor isoforms of NPCs suggesting a2b subunit expression. Picrotoxin and the neurosteroid pregnenolone sulphate are more potent inhibitors at homomeric than at heteromeric glycine receptors [48,[51][52][53]. Similar to recombinant a2b subtypes [41,48], 10 mM picrotoxin and pregnenolone sulphate caused only a moderate reduction of glycine-evoked currents at native NPC receptors. Corresponding to the quantitative PCR data for human mesencephalic NPCs, the elucidated pharmacological profile indicates the expression of strychninesensitive glycine receptors with a2b subunits.
In conclusion, the analyses of NPCs derived from human fetal midbrain tissue demonstrate the expression of strychnine-sensitive glycine receptors with an excitatory function following differentiation but not during proliferation. Molecular and pharmacological evidence suggest a predominant role for heteromeric a2b subtypes which may indicate that cells are not fully maturated. This study suggests that human mesencephalic NPCs acquire essential glycine receptor properties during neuronal differentiation in vitro.