Methylhydroxybenzoate, ethylhydroxybenzoate, m-cresol, propylhydroxybenzoate, butylhydroxybenzoate, sodium benzoate, chlorhexidine, benzoic acid, potassium sorbate, sorbic acid and cetylpyridinumchloride were obtained from Sigma-Aldrich (St. Louis, MI). Bronopol and phenylethyl alcohol were obtained from TCI Deutschland GmbH (Eschborn, Germany), benzalkonium chloride and benzethonium chloride from Fischer Scientific (Waltham, MA), chlorobutanol from MedChemExpress (NJ), benzyl alcohol from Biozol Diagnostica (Eching, Germany) and phenol from Boehringer Mannheim (Mannheim, Germany). Histamine was obtained from Fischer Scientific (Waltham, MA), Prostaglandin E2 from Adipogen (Füllinsdorf, Switzerland) and serotonin from Fluorochem (Hadfield, United Kingdom). Allyl isothiocyanate (AITC) was obtained from Fisher Scientific (Waltham, MA), capsaicin from Biotrend (Colonge, Germany). All substances were diluted in extracellular solution, composed of the following ingredients (in mM): 145 NaCl, 5 KCl, 10 glucose, 10 HEPES, 1.25 CaCl₂, and 1 MgCl₂ buffered to pH 7.4 with NaOH.

HEK293t cells were grown in Dulbecco’s Minimal Essential Medium (DMEM D5648, Sigma-Aldrich), supplemented with penicillin, streptomycin and L-glutamine (1% each, all from Lonza, Basel, Switzerland). HEK293t cells were transfected with human TRPA1 (hTRPA1) using jetPEI transfection reagent (Polyplus, Illkirch, France). The plasmid was obtained from Paul Heppenstall and cloned into a pcDNA3.1 vector (Heber et al., 2019; Meents et al., 2019). The cells were then spread on poly-D-lysine coated black 96 well plates (∼30.000 cells/well) and incubated overnight at 37°C and 5% CO₂.

The microfluorimetry of cytosolic calcium levels was performed as described before (Luo et al., 2011, p. 3). Briefly, cells were loaded with calcium 6 (Calcium 6 kit, Molecular Devices, San Jose, CA) in extracellular solution for 2 h according to the manufacturer’s protocol. Cells were not washed as fluorescence of the extracellular dye is absorbed by a dark dye in the kit. A pipetting fluorescence plate reader (FlexStation 3, Molecular Devices, San Jose, CA) was used to measure the change in intracellular Ca²⁺ concentration in hTRPA1 transfected and non-transfected HEK293t cells. Calcium 6 fluorescence, which was excited every 2 s at 485 nm, served as an index of intracellular calcium levels. Using a pre-set protocol, 50 µL of the assay solution was automatically pipetted onto the cells in 100 µL of medium. Fluorescence change is reported normalised to baseline fluorescence (dF/F0). Assays were performed at 37 °C (Heber et al., 2019). For experiments with histamine, PGE₂ and 5-HT, these mediators were mixed with the preservatives and co-applied onto the cells.

Breeding, euthanasia and all procedures of animal handling were performed according to regulations of animal care and welfare. Experiments were carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC). Wild type C57BL/6 or TRPA1^−/− mice were anaesthetised by an exposure to isoflurane or rising CO₂ levels and euthanized by cervical dislocation. DRGs from all spinal levels were excised and transferred to DMEM containing streptomycin, penicillin and L-glutamine, treated with 1 mg·mL⁻¹ collagenase (Sigma-Aldrich) and 3 mg·mL⁻¹ Dispase II (Roche) for 55 min at 37 °C. DRGs were then mechanically dissociated with a Pasteur pipette, centrifuged at 1,200 rpm for 5 min and plated onto 12 mm glass coverslips previously coated with poly-D-lysine (100 µg·mL⁻¹, Sigma-Aldrich). Dorsal root ganglia neurons were cultured in DMEM supplemented with streptomycin, penicillin and L-glutamine and mouse nerve growth factor 100 ng·mL⁻¹ (Alomone Labs, Tel Aviv, Israel) at 37 °C and 5% CO₂ for 15–30 h.

The coverslips were incubated with 3 µM Fura-2 AM (Biotium, CA, United States) for 30 min at 37°C and 5% CO₂, before placement in 35 mm glass-bottom dishes in extracellular solution. After 10 min, allowing removal or cleaving the ester of the dye, dishes were mounted onto an inverted microscope (IX73, Olympus, Tokyo, Japan) and imaged using a ×10 objective. Cells were continously superfused with extracellular solution using a software-controlled 8-channel, gravity-driven, common-outlet system (ALA Scientific Instruments Inc., NY, United States). This superfusion was switched to different solutions after baseline recording. After exposure to the preservative, TRPA1 agonist AITC and TRPV1 agonist capsaicin were applied. This allowed to calculate for single neurons a correlation of the response to the preservative and a subsequent stimulus, using a product-momentum correlation. Finally, a positive control detecting viable cells was added at the end of each recording, using depolarization by an extracellular solution with 60 mM KCl (isotonic replacement of NaCl).

Fura-2 was alternatingly excited for 30 ms by a 340-nm LED (50 mW, used at 100%) and by a 385-nm LED (1,435 mW, used at 5%) by an Omicron LEDHub (Laserage-Laserprodukte GmbH, Rodgau-Dudenhofen, Germany). Fluorescence emission was long-pass filtered at 495 nm, and pairs of images were acquired at a rate of 1 Hz with a 5.76-megapixel 16-bit CCD camera (6.5-μm pixel edge length, 22-mm sensor diagonal, Kinetix 22; Teledyne Photometrics AZ, United States). The hardware was controlled by the μManager 1.4 plugin in ImageJ (Schneider et al., 2012). The background intensity was subtracted before calculating the ratio between the fluorescence emitted when the dye was excited at 340 nm and at 385 nm (F340/F385 nm). The time course of this ratio was analysed for regions of interest adapted to individual cells.

An area under the curve (AUC) was calculated for 30 s after application in HEK293t cells, as these respond rapidly and show a high level of TRPA1 expression. In contrast, to detect also slower but steady rises in calcium, a longer area under the curve of 60 s was used of for sensory neurons. As baseline for AUC calculations, the arithmetic mean of the 30 s before application were used. Whether preservatives activate TRPA1 was assessed using dose-response curves with the concentration of a preservative as predictor and the AUC value described above as the dependent variable. For each preservative, a 3-parameter dose response curve with the parameters bottom, EC₅₀ and top was fitted. In case residuals were not approximately normally distributed, or fit parameters were unstable, a 4-parameter dose response curve containing the Hill slope as a fourth parameter was fitted. In case this also did not result in adequately distributed residuals or unstable parameters, a 5-parameter dose response curve was fitted, which includes an asymmetry factor as a fifth parameter. When the most adequate type of dose response curve was not apparent, the best model was chosen based on Akaike’s information criterion. Whether responses to preservatives in TRPA1-transfected were above naïve cells was tested by unpaired student’s t-tests. For aggregation of results in Figure 1S, the response at the lowest commonly used preservative concentration was taken from the experiment, or interpolated from the curve fit.

Sensitization of TRPA1 by inflammatory mediators was assessed using a mixed linear model. The preservative concentration and the inflammatory mediator were used as a factor. A random factor accounted for the dependency of values derived from the same 96 well plate. To stabilise the residual distribution, data were transformed prior to the analysis using log [x - minimum(x) + 1]. No adjustment for multiplicity was performed. p-values ≤0.05 were considered statistically significant. Curve fittings and graphs were generated by GraphPad 9, the mixed linear model was estimated by IBM SPSS statistics 29.

TRPA1 activation has already been demonstrated for the widely used parabens such as methylparaben, ethylparaben, propylparaben and butylparaben as well as for o-cresol (Fujita et al., 2007; Startek et al., 2021). This includes evidence for pain-related behaviour elicited in mice by methylparaben (Fujita et al., 2007). In the present study, activation of human TRPA1-transfected HEK293t by methylparaben was not far from the 4.4 mM reported before, and the minor difference might be well explained by non-identical experimental settings (Fujita et al., 2007).

Phenol and alkylated phenols are non-electrophilic TRPA1 agonists (Startek et al., 2021), their antimicrobial mechanism was suggested to rely on enzyme inhibition (Denver et al., 2011). Phenol has not been reported to be painful, although it is used at high concentrations for neurolysis in cancer patients (Koyyalagunta et al., 2016). An additional local anaesthetic effect was hypothesized to explain the lack of pain upon phenol injection (Zamponi and French, 1993). Bronopol and phenylethyl alcohol have not been previously investigated for human TRPA1 activation. Intraperitoneal bronopol injection was described to cause an acute ‘transitory aggressive behaviour in rats’, which was described as ‘painful state’ (Nieminen and Möttönen, 1975). Importantly, even the lowest concentration used for preservation caused TRPA1 activation. For bronopol, a reaction with bacterial thiols and disulfide formation has been described as a mechanism of antimicrobial action (Shepherd et al., 1988). This fits well with the covalent action on N-terminal TRPA1 cysteines. In contrast, for phenylethyl alcohol bacterial membrane alterations have been reported as a mechanism of action (Corre et al., 1990). However, an irritant action of phenylethyl alcohol 6 g/L added to eye drops has been described in a blinded experiment in six subjects (Boer, 1981). Meta-cresol is a known irritant, with a TRPA1-dependent activation at low concentrations, at preserving concentrations there is also a TRPA1-independent component. Different formulations of growth hormone showed that pain perception was more for 0.25% m-Cresol compared to 0.9%–1.5% benzyl alcohol (Kappelgaard et al., 2004).

Benzyl alcohol, cetylpyridinium chloride, benzalkonium chloride and benzethonium chloride caused TRPA1-independent calcium influx at target concentrations. Chlorobutanol induced calcium influx at the highest tested concentration, however, maximum solubility in extracellular solution was below the preserving range. The cellular mechanism of the TRPA1-independent calcium influx in HEK293t cells is unknown. However, extracellular calcium was not required to observe the calcium elevation caused by TRPA1-independent preservatives. This indicates a ubiquitous intracellular calcium source, like the endoplasmic reticulum or mitochondria.

More pain was reported after shoulder arthroscopy by patients with benzyl alcohol in the saline solution compared to only saline, indicating the choice of preservative matters (Storey et al., 2014). Paravasation of intravenous drugs is underdocumented, and although cytotoxic, proinflammatory and irritant reactions are known for many pharmaceuticals, a potential contribution of preservatives to that is unknown (Le and Patel, 2014). Benzalkonium instillation into the bladder causes pain (Xu et al., 2011). No literature suggesting nociceptor activation, pain-related behaviour in animals or reports of pain was found for cetylpyridinium, benzethonium and chlorobutanol.

Importantly, there are substances available which do not activate at the required concentration for preservation. Potassium sorbate, sodium benzoate, and chlorhexidine elicited no calcium influx in untransfected HEK293t cells at any of the test concentrations.

A database search for these three substances did not yield literature suggesting nociceptor activation or induction or facilitation of pain-related behaviour in animals. Therefore, these substances are favourable for avoiding cell activation.

Inflammatory conditions alter neuronal sensitivity, and this includes TRPA1 expression and responsiveness (Bautista et al., 2013). Parenteral drugs are rarely deliberately applied to inflamed tissues, but, e.g., for application of a proinflammatory drug, the induced inflammation and the preservative might cause an additive or supra-additive effect. In inflammation, several proinflammatory mediators are upregulated (Choi and Di Nardo, 2018). This has led to the common uses of histamine, serotonin prostaglandin E2 and bradykinin as an ‘inflammatory soup’ to mimic the biological changes in inflamed tissue. In the in vitro model, bradykinin was omitted, as it elevated intracellular calcium in HEK293t, as previously demonstrated (Kramarenko et al., 2009). Therefore, sensitisation cannot be easily quantified or might be obfuscated by cross-desensitisation. HEK293t cells express the respective receptors for proinflammatory mediators (Atwood et al., 2011). Histamine, serotonin or prostaglandin E2 alone did not cause a calcium influx, but sensitised the preservative-induced TRPA1 activation. For phenylethyl alcohol 10 μM, a supra-additive response was observed, as neither the preservative nor the inflammatory mediators alone cause a response. Sensitisation by inflammatory mediators was confirmed for bronopol, ethylparaben, propylparaben, and butylparaben.

The chosen substances reflect the currently used preservatives. Allowed substances may differ based on regulation of countries or regions and are subject to change. Further, some substances have been used in parenteral formulations but are currently only found in ophthalmic, nasal or otic formulations, or in creams, which applies, e.g., to phenylethyl alcohol, ethylparaben or butylparaben. Other substances, e.g., organomercury compounds like thiomersal have not been investigated, as their use in new parenteral formulations is rather limited. The presented experiments were mostly done on HEK293t cells and partially confirmed in sensory neurons; other cell types might have different sensitivity. Further, validation in vivo seems a necessary next step to judge the clinical relevance of the presented results.

Several preservatives commonly used in therapeutics activate TRPA1. This is particularly important in acute inflammation, where sensitization may add to hyperalgesia. The findings can serve as an orientation for the selection of preservatives to reduce TRPA1-dependent and independent activation.