Natural remedies are increasingly viewed as potentially valuable complements to conventional drugs in integrated treatment strategies for a number of disorders, and many consumers use natural health products alongside prescription medications . Anxiety and depression are among the ailments most frequently reported by patients seeking complementary and alternative medical remedies and/or naturopathic care [9, 52, 53]. Gelsemium s. is a traditional remedy used in complementary and alternative therapies for treating patients who exhibit neurological complaints such as headache and anxiety-like symptoms [9, 52, 53], but evidence-based clinical studies are few and with contrasting results [53, 54].
Homeopathy is a 200-year-old therapeutic system that uses extremely small doses of various substances to stimulate auto-regulation and self-healing processes . Although some conventional physicians find such notions implausible , use of highly diluted drugs from homeopathic pharmacopoeia has recently seen a worldwide revival [57, 58] and laboratory investigations are increasing in this field [26, 27], but scientific evidence of underlying molecular mechanisms is still lacking. Moreover, the experimental approaches adopted to study these remedies, particularly for highly diluted solutions, have suffered from problems with replicability between different laboratories. It is therefore important for any results in this field to be confirmed and consolidated through further investigations by independent laboratories, using rigorous protocols and statistical evaluations. The expression microarray analysis on whole genome, as other high-throughput technologies assisted by bioinformatics, could provide a strong clue as to the mechanism of action and the biological relevance of ultra-low doses and high dilutions interactions.
This is the first comparative transcriptomics approach to investigate changes in the human neurocytes induced by a natural plant remedy, traditionally used for anxiolytic-like effects. The chief innovation of our experimental design is that it employs a wide range of doses/dilutions. This enabled us to explore changes in gene expression from low dilutions (2c or 3c), where the active substances can still be expected to exert their normal pharmaceutical action, to high dilutions (9c or 30c), where the most controversial principles of high dilution pharmacology come into play. Thus, both conventional and 'alternative’ pharmacological theories were evaluated and compared in the same investigation. In previous recent trials, Gelsemium s. showed anxiolytic-like effects in mouse emotional response models and appeared to work even at the high dilutions 9c and 30c [26, 27]. Two other studies have also found that high dilutions of Gelsemium s. exert a preventive action against experimental stress (electric shock) in mice  and against convulsions provoked by lithium and pilocarpine in rats . Other researchers have reported an anti-anxiety activity of Gelsemium s. and of the alkaloids gelsemine, koumine, and gelsevirine [14, 16], but have not explored the effect of ultra-low doses and high dilutions/dynamizations.
To follow up on these in vivo studies, we decided to investigate the action of Gelsemium s. at the cellular and transcriptional level. We adopted a validated microarray protocol and applied it to a series of replicate experiments designed to test: a) the null hypothesis that the effect of any Gelsemium s. dilutions is similar to that of the control vehicle, b) whether any dose-dependence of the putative effects can be demonstrated. As our model system, we chose the SH-SY5Y and IMR-32 neurocyte cell lines because these have been previously employed for investigations of natural compounds , neurotrophic factors , mood stabilisers , and antipsychotics . In our conditions, this line proved to be more responsive to Gelsemium s. than IMR-32 and was used to compare the effects of high dilutions. The cells were functional, as demonstrated by intracellular calcium increase following treatment with the neurotransmitter carbachol, and none of the Gelsemium s. dilutions affected their growth rate or metabolic activity.
The most evident and statistically significant modification of cellular gene expression was induced by the lowest dilution of the medicine that we tested, namely Gelsemium s. 2c, as is to be expected with a dose-dependent effect. Spectroscopic analysis of the tested samples confirmed that the starting 1c solution supplied by manufacturer contained a considerable amount of original extract compounds, and proportionate quantities were also detected in the 2c dilutions prepared both by the manufacturer and in our laboratory. However, dilution equal or beyond 3c brought the concentration of detectable compounds below the minimum sensitivity threshold of optical spectroscopy.
Although Gelsemium s. contains several different compounds [2, 3], the major active alkaloid of this plant is gelsemine, which was present in a concentration of 6.5 × 10-9 M in the final incubation mixture of cells treated with Gelsemium s. 2c. This nanomolar dosage is much lower than the toxic doses that have been reported in poisoning cases  and in experimental evaluations of LD50 . In fact, Gelsemium s. 2c at the lowest dilution (highest dose) tested in this model system did not cause cell toxicity or viability impairment. This evidence is in agreement with recent hypotheses explaining the homeopathic effects (in the range of very low doses) in the framework of hormesis, where substance which are toxic at high doses turn into therapeutic when diluted to low and ultra-low doses . According to the hormetic theory, ultra-diluted drugs and nanoparticles will act as low-dose stress conditions that could possibly evoke an adaptive response process producing effects that might modulate gene expression [62–64].
The effects on gene expression observed here are specifically targeted to the regulation of certain functions, possibly linked with the plant’s pharmacological activity. Of a total of 45033 transcripts, 49 were down-regulated and 7 up-regulated by the 2c dilution. This effect was quantitatively small (absolute value of fold change between 0.5 and 1.0) but statistically significant (adjusted p < 0.05). In general, the prevalence of down-regulation seems to indicate a tendency to reduce cell excitability, especially because several of the genes in question belong to surface receptors involved in GPCR signaling and calcium homeostasis. Moreover, this first microarray screening of the effects of Gelsemium s. on neurocytes revealed a significant down-regulation of genes for inflammatory response, olfactory transduction and neuron differentiation. Clearly, this plant species contains a variety of active chemical principles which are presumably involved in different pathways of cell regulation besides the pure neural function, as suggested by reports of possible anticancer and immunomodulating activities [17–19].
A hypothetical neurological target of Gelsemium s. has been suggested by studies showing that gelsemine stimulates the biosynthesis of allopregnanolone in the rat brain [30, 65], but the genes of neurosteroid enzymatic pathways were not modulated in our cell system. This apparent discrepancy may depend either on the fact that we used a cell line, whereas Venard et. al. [30, 65] used slices of spinal cord and limbic system, or on the fact that they studied a post-translational level of regulation, linked to enzyme function and not to gene expression. In any case, since in our model the effects of Gelsemium s. were quantitatively small, as confirmed by RT-PCR results, no definite conclusions regarding the role of single genes in the action mechanism of this plant can be drawn at this stage.
These microarray findings can be regarded as a preliminary screening of the sensitivity of SH-SY5Y cellular system to Gelsemium s., while more robust conclusions about the possible role of the implicated genes will require to determine whether proteins encoded by the affected genes are similarly changed, through proteomic and phosphoproteomic approaches, and/or further studies using plant purified active compounds.
Ultra-low doses and high dilutions
The second major goal of this investigation was to study the dose-effect relationship, which is of central importance in any kind of pharmacological approach. As noted above, the Gelsemium s. 2c dilution yielded statistically significant results for 56 genes. This raised the question of whether those same genes, which appeared to be most sensitive to Gelsemium s., would also be modified by higher dilutions. Since the quantitative changes for the 3c and higher dilutions were quite low (Table 3 and Figure 7B), the 4 replicates were insufficient to yield statistical confidence for analysis of single transcripts. We therefore employed cluster analysis to separately describe the trends of 6 gene subsets with similar expression profiles. All 4 down-regulated clusters included genes with negative mean fold changes, though of varying magnitude. Most notably, we found two clusters (2 and 3 in Figure 7) that included a total of 24 genes clearly responsive to Gelsemium s. 30c, and characterized by a bell-shaped dilution-effect curve. Exploring results accurately (Table 3), some genes showed an interesting pattern of expression in function of Gelsemium dilution. For instance, the EN2 gene that was under-expressed in treated cells exhibited a bell-shaped curve. This tendency can be seen in other genes in the cluster analysis. Moreover, it seems that after 9c, another wave of expressions or no-expressions is recovered. Maybe future testing even higher dilutions, such as 100c or 200c, the bell-shaped curve could be more evident and, thus, the hypothesis of ultra-sensible genes could be checked.
For the high dilutions, due to the small changes of gene expression, the only hypothesis statistically evaluable is the global effect of Gelsemium s. dilutions on the 49 down-regulated and 7 up-regulated genes, considered as gene-sets. Using the Fisher exact test (Gelsemium s. dilutions vs. their respective control solutions), the null hypothesis was rejected for every dilution in the down-regulated gene-set. This outcome of our microarray analysis is astounding if we consider that the 9c and 30c dilutions were obtained from MT extract by dilution factors of 1018 and 1060 respectively. Starting from a crude MT containing the active principle gelsemine at a concentration of 6.5 × 10-4 M, the 9c dilution would theoretically contain 6.5 × 10-22 M gelsemine, corresponding to less than 1 molecule per ml in the final working solution; even in the case of the 5c dilution, where the theoretical gelsemine concentration is 6.5 × 10-15 M, it can be calculated that this would correspond to 3.9 × 107 molecules per culture plate, i.e. about 13 molecules per seeded cell. These results suggest that neurocytes have a number of genes with extreme sensitivity to Gelsemium s. effects, even if those effects of high dilutions are quantitatively very small (decrease in expression by approximately 10% to 20% compared to the control). The physiological or pharmacological implications of this observation remain to be clarified, but the rejection of the null hypothesis furnishes a new input for the open debate on this kind of therapeutic approach.
Technical issues and confounding factors
The puzzling evidence of gene expression changes under the influence of homeopathic dilutions prompt an analysis of the possible confounding factors that might explain the effects observed. We adopted different measures to address the issue of possible experimental artifacts. To avoid dye-bias artifacts a single-channel microarray was employed. We adopted a microarray design with probes of the same probe-set located in not contiguous positions on the array, so that artifacts due to uneven hybridization would only affect a subset of probes for a probe-set. Anyhow, the absence of spatial biases in fluorescence signal was assessed by checking the coefficient of variation of the mean signal intensities of different portions of each array. The experimental set up could have introduced biases and “position effects” if handling of control and Gelsemium s. matched dilutions was not equivalent. Actually, we conducted four independent experiments in which Gelsemium s. dilutions and the corresponding vehicle controls were processed in tandem (from drug addition to RNA extraction and cDNA synthesis). In every subarray of the chip, each transcript was targeted with three separate probes, merging the fluorescence values and attributing a statistical score.
Regarding the statistical analysis, the large number of genes of the complete set causes some problems concerning the choice of “interesting” genes. The approach followed here was quite stringent and limited the number of genes considered, reducing the probability of “false positive” results, but forcing to discard some possibly interesting genes from the analysis. Moreover, the small entity of the expression changes observed with high dilutions unavoidably reduced statistical inference in the single genes, especially since multiplicity corrections were applied. The choice of analyzing the sign of the fold changes in a pool of genes, rather than the variance of a single gene, may lead to a loss of statistical information, to the advantage of greater precision in discarding the null hypothesis. Further research specifically oriented on the most responsive genes, with suitable sample sizes, could possibly overcome this limitation of the microarray approach.
Physico-chemical and biological hypotheses
Our results are in keeping with a number of experimental observations from a variety of research fields, confirming that highly diluted compounds exert statistically significant effects on biological systems [66–69]. Thus far there is no satisfactory or unifying theoretical explanation for these claims, though some have hypothesized that the dynamics of the solvent water (or water-ethanol) on a mesoscopic scale may play a part . Three major models for how this happens are currently being investigated: the water clusters or clathrates, the coherent domains postulated by quantum electrodynamics, and the formation of nanoparticles from the original solute plus solvent components. It has been suggested that a major role in the formation of water clusters is played by silica released from the glass containers which are usually employed in the preparation of homeopathic drugs . Silica nanostructures formed during succussion in glass and/or biosynthesized by specific plant extract tinctures may also acquire and convey epitaxial information from the remedy source materials into the higher potencies [21, 72, 73]. In our experimental model, since the verum were succussed samples, we used the succussed ethanol/water solutions as negative controls and evaluated preliminarily the variability of the negative system before assessing the biological effect of the succussed/diluted drug. Notably, in our experiments serial dilutions/succussions were performed in glass bottles, with the exception of the last step, which was developed in polystyrene tubes. Thus the hypothetical role of silicates in nanoparticle formation is pertinent, but also the contribution of polystyrene should not be excluded .
Recent evidence supports the plausibility that homeopathic Gelsemium s. in the potencies tested could contain crudely formed nanoparticles. Bel-Haaj et al.  demonstrated that just extended ultrasonication of plant starch can create starch nanoparticles in water. Moreover, electron microscopic evidence of nanoparticles has been obtained in several different plants prepared homeopathically . Gelsemium mother tincture itself, like many other plant extracts, can biosynthesize nanoparticles of silver metal from precursor substrate . Nanoparticles have unique biological and physicochemical properties, including increased catalytic reactivity, protein and DNA adsorption, bioavailability, dose-sparing, electromagnetic, and quantum effects that are different from those of bulk-form materials . As an example, Prakash and colleagues  compared in model animals the anti-anxiety effects of hypericum prepared as gold nanoparticles versus a bulk form and observed more significant effects with the nano-hypericum, even at a 10-fold lower dose. Higher cellular uptake of nano-encapsulated (poly lactide-co-glycolide) Gelsemium s. than of its bulk form has been observed by Bhattacharyya et al. .
The hypotheses regarding the possible biological mechanisms of highly diluted/dynamized solutions (beyond Avogadro-Loschmidt limit) at the level of DNA expression variously invoke sensitivity to bioelectromagnetic information, participation of water chains in signaling, stochastic resonance, and regulation of bifurcation points of nonlinear systemic networks [64, 80–83]. Based on microarray data, it has been suggested that gene regulatory networks may be regarded as dynamically 'critical’ systems poised near the phase transition between order and chaos [80, 84, 85], where extreme sensitivity to initial conditions and small perturbations is well known to occur. Chaotic regimes have been found in a number of physiological systems, including neural systems [86–88], and this would result in enhanced susceptibility to extremely low energy inputs and to small changes of regulatory factors. According to this argument, the highly diluted drug might be regarded as a solution endowed with water clusters and/or nanoparticulate structures capable of communicating some pharmacological information, through a resonance process, to biological fluids and to cell critical systems such as macromolecules, alpha-helixes, filamentous structures, receptors and DNA networks. This effect could be mediated by the participation of a dynamic intracellular water network which may be presumed to exist in living cells .