Adult human dental pulp stem cells promote blood–brain barrier permeability through vascular endothelial growth factor-a expression

Abstract

Stem cell therapy is a promising new treatment option for stroke. Intravascular administration of stem cells is a valid approach as stem cells have been shown to transmigrate the blood–brain barrier. The mechanism that causes this effect has not yet been elucidated. We hypothesized that stem cells would mediate localized discontinuities in the blood–brain barrier, which would allow passage into the brain parenchyma. Here, we demonstrate that adult human dental pulp stem cells express a soluble factor that increases permeability across an in vitro model of the blood–brain barrier. This effect was shown to be the result of vascular endothelial growth factor-a. The effect could be amplified by exposing dental pulp stem cell to stromal-derived factor 1, which stimulates vascular endothelial growth factor-a expression. These findings support the use of dental pulp stem cell in therapy for stroke.

Introduction

The chronic cognitive and motor–sensory dysfunction resulting from stroke presents a major challenge to healthcare in much of the developed world. As of 2010, there were approximately 33 million people living with the effects of stroke, making stroke responsible for 4% of the total disability-adjusted life years of disease burden.¹ Given the scope of this issue, it is clear that much importance must be placed on developing efficacious treatments to return patients to a healthy and functional state.

Cell-based therapy is emerging as a possible treatment option for stroke.² This field of study was sparked by the ability of stem cells to populate areas and replace the function of many divergent tissue types. Transplantation of stem cells provides the potential to extend the post-stroke recovery of function beyond what is possible by endogenous recovery alone. This has been demonstrated over the last decade in animal models.^2–6 However, this effect is not well explained by the tissue replacement model alone. The paracrine secretion of various factors may underlie the functional improvement seen in animal stroke models of cell-based therapy whereby immunomodulation, neuroprotection, neurogenesis, neuroplasticity and angiogenesis are supported and enhanced.⁷

Research into cell-based therapy for stroke has advanced to a point where a number of early phase clinical trials have been undertaken.^8–11 However, several basic questions remain unanswered, with the focus from this study being on how stem cells may transmigrate the blood–brain barrier (BBB) if administered via the vasculature.

One potential source of stem cells for therapy is dental pulp stem cell (DPSC), which was first described by Gronthos et al.¹² These are a population of highly proliferative, undifferentiated cells residing in perivascular niches within the dental pulp of adult teeth. DPSCs have been demonstrated to have the capacity to differentiate into neurons, adipocytes, myocytes and chondrocytes in vitro.^13–15 Intracerebral (IC) transplantation following stroke in a rat model showed an improvement in functional outcomes.²

There are two paradigms for the administration of therapeutic stem cells in stroke research: intravascular (IV) and IC. IC transplantation delivers stem cells directly to the site of damage, which may result in a greater surviving population of viable cells. However, this is a highly invasive procedure and is associated with an increased mortality rate of approximately 10% when administered acutely in rodent stroke models.² As such, acute IC transplantation may not be a clinically viable model.

IV administration of stem cells in animal models of stroke has been validated as efficacious. It has even been demonstrated in some studies that small populations of transplanted stem cells can be detected within the brain parenchyma.⁵ A likely mechanism for this is that stem cells migrate along a stromal-derived factor-1 (SDF-1) gradient towards the ischaemic border zone via the receptor CXCR4. SDF-1 is upregulated for at least a month following stroke, so the effective window for treatment could hypothetically be extended significantly.^16,17 However, this depends on how stem cells interact with the BBB.

The BBB refers to the specialized structure of the neurovascular unit. The BBB is composed of a continuous layer of endothelial cells, which are surrounded by a basement membrane, pericytes and astrocyte processes. The main function of the BBB is to segregate the environment of the brain parenchyma from peripheral circulation. In general, the BBB is selectively impermeable to large polar molecules and cells.¹⁸

The mechanisms by which stem cells transmigrate the BBB have not yet been fully characterized. It is possible that IV stem cells administered post-stroke take advantage of existing damage to the BBB to gain access to the therapeutic target. However, even stem cells administered after the usual period of post-stoke BBB opening are able to gain access to the brain parenchyma.^5,19 This study will therefore investigate the possibility that DPSCs are capable of mediating passage through the BBB by causing a temporary opening of the barrier.

There are a number of cytokines known to cause increases in BBB permeability. Notable examples are members of the vascular endothelial growth factor (VEGF) family, including VEGF-a and placental growth factor, both of which are potent permeability factors involved in repair of cerebral circulation following stroke.^20,21 A study by Díaz-Coránguez et al. implicates VEGF-a as one of the mediators of permeability in glioma-induced BBB permeability.²² VEGF-a acts by causing a downregulation of occludin and claudins, both components of tight junctions.²¹ Stem cells, in general, and DPSC, in particular, have been shown to express VEGF-a and stimulate angiogenesis.^23,24

Therefore, a reasonable explanation for trans-neuroendothelial migration by stem cells is that as circulating transplanted cells pass near ischaemic tissue, they secrete VEGF-a onto adjacent endothelial tissue. VEGF-a binds to VEGFR2 on the surface of endothelial cells, causing downregulation of tight junction proteins and ultimately a local increase in BBB permeability.

Co-cultures of endothelial cells and astrocytes have been used previously to model the BBB, allowing for a highly reproducible experimental construct.^25–28 It has been proposed that (an) unknown factor(s) expressed by astrocytes stimulate the adoption of BBB characteristics in an endothelial monolayer. This type of model has been shown to have a similar permeability profile to the BBB in vivo and also to respond to some permeability-inducing drugs.^26,28,29

This study aims to develop an in vitro model of the BBB, to investigate the ability of DPSC to mediate BBB permeability and to determine the molecular mechanism underlying this activity.

Materials and methods

Cell culture

Cells were passaged when near 80% confluency. Adherent cells were harvested from plastic surfaces by incubating in trypsin/EDTA (Invitrogen) in phosphate-buffered saline (PBS) and resuspended in the appropriate media for experimental use. Cells were cultured in media described in Table 1.

Table 1.

Culture media and supplements for DPSC, BMEC and astrocytes.

Human DPSC
a-modified Eagle’s media		Sigma-Aldrich, St. Louis
FCS	10% (v/v)	Invitrogen
l-Glutamine	2 mM	Invitrogen, 250030081
Penicillin	100 µg/ml	Invitrogen cat#15140122
Streptomycin	100 µg/ml	Invitrogen cat#15140122
l-ascorbic acid 2-phosphate	100 µM	Wako, Richmond, VA
Human BMEC
M199 medium		Sigma-Aldrich, St. Louis
FCS	16% (v/v)	Invitrogen
Penicillin	100 µg/ml	Invitrogen cat#15140122
Streptomycin	100 µg/ml	Invitrogen cat#15140122
Non-essential amino acids	0.8X	Invitrogen cat#11140050
Sodium pyruvate	0.8 mM	Sigma cat#S8636
Murine astrocytes
Dulbecco’s modified eagle’s medium		Sigma-Aldrich, St. Louis
FCS	10% (v/v)	Invitrogen
Penicillin	100 µg/ml	Invitrogen cat#15140122
Streptomycin	100 µg/ml	Invitrogen cat#15140122

BMEC: bone marrow endothelial cell; DPSC: dental pulp stem cell; FCS: Foetal calf serum.

Human DPSC

DPSCs (Figure 1(c)) were previously isolated from molar teeth extracted during routine dental procedures from otherwise healthy adults. Before use, DPSCs were stored in 10% DMSO (Merck, 10323.4L) in foetal calf serum at a liquid nitrogen storage facility.

Figure 1.

Phase contrast micrographs of astrocytes (a), endothelial cells (b) and DPSC (c). (d) A timeline showing the procedure followed during the preparation of an in vitro BBB. (e) A diagram showing the arrangement of the BBB model ready for testing at day five.

Use of human tissue was approved by the University of Adelaide Human Research Ethics Committee (ethics approval number: H-2012-164). This study adheres to guidelines set out in the National Statement of Ethical Conduct in Human Research (2007) and sections 95 and 95 A of the Privacy act (1988). Written, informed consent was given by all human donors.

Human bone marrow endothelial cells (BMEC)

The BMEC line (Figure 1(b)) was kindly provided by Prof. Andrew Zanettino (SA Pathology), who had previously acquired it from Dr Babette Walker (Weill Medical College of Cornell University).

Murine cerebellar astrocytes

Astrocytes (ATCC, CRL-2541) (Figure 1(a)) were originally obtained from the cerebellum of an eight-day-old mouse and the permanent line was selected following spontaneous transformation.

Preparation of the in vitro BBB model

The luminal surface of an 8 µm pore cell culture insert (BD Falcon TM, 353182) was coated with bovine collagen (Sigma, C4243) at a density of 20 µg/cm² and left to dry. On day two, astrocytes were harvested and 1 × 10⁵ cells were seeded onto the underside of the membrane. The astrocytes were allowed to adhere for 4 h. BMECs were harvested and 1 × 10⁵ cells were seeded onto the upper surface of the membrane. Astrocyte medium supplemented with EGF (BD Biosciences, BD356006) and heparin (Sigma, H-0777) at a dilution of 1:333 was added to both compartments; this was done to increase proliferation of the BMEC. On day four, the medium was replaced with unsupplemented astrocyte medium to slow down BMEC proliferation and allow BBB characteristics to form. On day five, the co-culture was confluent and resembled the BBB in terms of permeability (Figure 1(d) and (e)).

Immunocytochemistry

To confirm the presence and regulation of the molecular components of the BBB, a monolayer of endothelial cells was cultured on 12 mm glass coverslips. A selection of these was co-cultured with astrocytes. These were then treated with 100 ng/ml VEGF-a (ebioscience®, 14-8359-80), DPSC-conditioned media or vehicle for 24 h. Additionally, rodent brain sections were obtained for use as a positive control. The slides were fixed in 4% (w/v) PFA (Sigma, P6148) in PBS for 30 min and stored in PBS. Samples were permeabilized by washing with 5% (v/v) TX-100 (Sigma, T8787) in PBS for 5 min, then blocked with 2% (w/v) BSA (Sigma, A7906) and 0.2% (v/v) TX-100 in PBS for 2 h, then washed 3X in PBS. Each slide was incubated overnight with 1:200 rabbit anti-occludin (Abcam, ab31721) in 1% (w/v) BSA, 0.05% (v/v) TX-100 and 0.5% (v/v) normal donkey serum (Sigma, D9663) in PBS (blocking buffer B). The slides were then washed 3X in PBS and once in blocking buffer B before being incubated for 2 h with 1:200 Cy2 anti-rabbit (Jackson ImmunoResearch, 109225008). The coverslips were then washed twice with PBS, inverted and fixed to microscope slides with ProLong® Gold Antifade containing DAPI (Life Technologies, P-36931). Capture of micrographs was not subject to blinding; however, images were analysed using a computer algorithm (ImageJ, 1.49C) to prevent human error. Micrographs were taken from areas on each slide containing sufficient cells for statistical analysis.

VEGF-a-induced permeability across different sized markers

In vitro BBB models were prepared as described above. On day five, nine assemblies were incubated with 100 ng/ml VEGF-a for 5.5 h and nine were used as controls. Two microlitres of 5% (w/v) Evan’s Blue (EB) (Sigma, E2129) in 12.5% (w/v) BSA (Sigma, A-7906), 2 µl of 5% (w/v) methylene blue (Probing & Structure, C124) or 2 µl of 5% (w/v) bromophenol blue (Sigma, B-8026) in PBS was added to each luminal compartment. After 30 min, the micropore inserts were removed and 2 × 100 µl samples were removed from the abluminal chamber and transferred to a 96-well plate. Absorbance for EB, bromophenol blue and methylene blue was read at 620, 590 and 600 nm, respectively, using a plate reader (BioTek®, Synergy MX).

VEGF-a-induced permeability

BBB models were prepared as described above. Assemblies were incubated with VEGF-a at a concentration between 0 and 100 ng/ml for 5.5 h. Two microlitres EB-BSA was then added to the luminal chamber and the assembly was incubated for a further 30 min. The transwell inserts were then removed and 2 × 100 µl samples were taken from the abluminal chamber and transferred to a 96-well plate. Absorbance was read at a wavelength of 620 nm.

DPSC-conditioned media-induced permeability

DPSCs were cultured in two 75 cm² flasks (samples 1 and 2) until they had reached ∼80% confluency. Media were changed then harvested after three days and sub-sequently stored at -80℃. In vitro BBB models were prepared as described above. On day five, the assemblies were incubated for 30 min with 1 ml of DPSC-conditioned medium (sample 1 or sample 2) or unconditioned DPSC medium. Each assembly was incubated for 30 min with 2 µl of EB-BSA added to the luminal compartment. The transwell inserts were removed and 2 × 100 µl samples were taken from the abluminal compartments and transferred to a 96-well plate. Absorbance was read at a wavelength of 620 nm.

ELISA

DPSCs were cultured until they reached ∼80% confluency. Media were then changed and cells were cultured for a further 72 h. The conditioned media were subsequently aspirated and stored at -80℃. Prior to testing, 4 ml samples were thawed at room temperature and concentrated using Amicon filters (Millipore, UFC801024) and 100 µl samples were set aside for total protein quantitation. ELISAs were conducted using a duoset kit (R&D, DY293B-05) as per the manufacturer’s instruction. Samples were tested at serial dilutions between concentrations of 16X and 0.25X and the peak reading within the sensitivity of the assay was taken. Protein quantitation was carried out using a colourimetric protein quantitation kit (Bio-Rad, 500-0006EDU) as per the manufacturer’s instruction. The amount of protein sequestered by cells in culture was determined by subtracting total protein in each conditioned medium sample from total protein in unconditioned medium. The VEGF-a concentration determined by ELISA was then standardized to sequestered protein for each sample. This was done to control for cell number.

Pharmacological control of permeability

BBB models were prepared as described above. Assemblies were incubated with either VEGF-a at a concentration of 100 ng/ml (as a positive control) or DPSC-conditioned medium for 5.5 h. Cediranib (life research, S1017), a VEGFR2 antagonist was added immediately prior to treatment at a concentration of 10 µM in DMSO. Three replicates from each treatment group were exposed to DMSO only as a negative control. EB-BSA (2 µl) was added to the luminal compartment and assemblies were incubated a further 30 min. The inserts were removed and 2 × 100 µl samples of media from the abluminal compartment were transferred to a 96-well plate. Absorbance was read at a wavelength of 620 nm.

Statistics

Graphpad Prism (V5.04) was used to calculate all statistics. For the in vitro BBB model, the concentration of dye present in the abluminal chamber was determined by interpolating the absorbance within serial dilutions, results were expressed as percentage of maximum equilibration. One-way ANOVA with a Tukey post hoc test was used to compare grouped data. Dose–response assays were assessed by testing the strength of their linear correlation. Normality was confirmed by calculating quantile–quantile plots with the pooled data from each experiment. Power calculations indicate a minimum sample size of three BBB model systems per group is required to detect a 44% difference between groups to account for a standard deviation of 4.15%, with 80% power and 95% confidence. Parameters for this power calculation were set using data obtained from the initial permeability experiment described above.

Results

Characterization of an in vitro BBB model system

Tight junction modulation is the major physiological mechanism underlying changes in BBB permeability. In order to assess the validity of this in vitro BBB model, immunocytochemistry was used to characterize the mechanisms that underlie modulation of permeability (Figure 2(a) to (d)). The tight junction protein occludin appeared to be upregulated when BMECs were co-cultured with astrocytes. However, this failed to reach statistical significance (Figure 2(e)).

Figure 2.

Immunocytochemistry of BMEC monolayers stained for the tight junction protein occludin (Cy2/green) and nuclei (DAPI/blue). (a) BMEC and astrocyte co-culture negative control for the occludin primary antibody. (b) Positive control for the occludin antibody, a section of rodent brain showing expression of occludin in vascular tissue of the choroid plexus. (c) Co-culture of BMECs and astrocytes. (d) Monolayer of BMECs grown without astrocytes. Images a, c and d were taken at 200X magnification, image b was taken at 100X magnification. (e) Quantitation of relative occludin expression for above immunocytochemistry. Error bars depict standard deviation.

To investigate whether this in vitro model could approximate the permeability characteristics of the BBB, three dyes with different molecular sizes were added to the luminal side of the assembly along with VEGF-a, a known inducer of permeability. As shown (Figure 3(a)), EB-BSA was least able to equilibrate under normal conditions. EB-BSA was also the only marker to increase significantly in equilibration when the in vitro model was treated with VEGF-a. Hence, EB-BSA was used as the marker in all subsequent in vitro BBB experiments.

Figure 3.

(a) A representation of the percentage of equilibration detected for three markers of different molecular weight with or without 100 ng/ml VEGF-a treatment. Methylene blue = 319 Da, bromophenol blue = 670 Da and EB-BSA = 66,423 Da. For EB-BSA VEGF-a: Control, P < 0.0001, n ≥ 5 experimental replicates per group. Error bars depict standard deviation. (b) The linear association between the concentration of VEGF-a added to the model and the percentage of equilibration. Discontinuous lines show 95% confidence intervals. P < 0.0001, R²= 0.64, n = 23 experimental replicates total.

VEGF-a was shown to mediate an increase in permeability of the in vitro BBB model. To determine whether this is dose dependent, a range of concentrations of VEGF-a were added to the luminal compartment of in vitro BBB models. The relationship between VEGF-a concentration and equilibration was subjected to linear regression analysis. As shown (Figure 3(b)), there was a significant positive correlation between VEGF-a dose and permeability.

DPSC mediates an increase in BBB permeability via VEGF-a expression

Next, this in vitro BBB model was used as a permeability assay to investigate whether soluble factors expressed by DPSC were capable of mediating an increase in permeability. It was found that DPSC-conditioned media caused a statistically significant increase in permeability above unconditioned media (Figure 4).

Figure 4.

Comparison between the equilibration allowed by barriers treated with DPSC conditioned and identical but unconditioned media. P = 0.0002, n ≥ 6 experimental replicates per group. Error bars depict standard deviation.

DPSC-conditioned media caused a significant increase in BBB permeability. To investigate the mechanism by which this downregulation occurred, occludin expression was assessed following treatment with DPSC-conditioned media and VEGF-a. It was shown (Figure 5(a) to (e)) that treatment with both DPSC-conditioned media and VEGF-a caused a similar downregulation in occludin expression in the in vitro BBB and this was found to be statistically significant.

Figure 5.

Immunocytochemistry of BMEC monolayers stained for the tight junction protein occludin (Cy2/green) and nuclei (DAPI/blue). (a) Untreated co-culture of BMECs and astrocytes. (b) BMEC and astrocyte co-culture negative control for the occludin primary antibody. (c) Co-culture of BMECs and astrocytes treated with DPSC-conditioned medium. (d) Co-culture of BMECs and astrocytes treated with VEGF-a. The images presented are representative samples from 24 co-cultures examined. Images were taken at 200X magnification. (e) Quantitation of relative occludin expression for above immunocytochemistry. Error bars depict standard deviation.

It was necessary to determine whether DPSC expressed VEGF-a in the conditioned media. This was assessed by sandwich ELISA of DPSC-conditioned media. Three samples were taken from each of two separate populations of DPSC. Figure 6(a) demonstrates that VEGF-a was present in the conditioned media at a concentration of 0.33 ng/ml, with a standard error of 0.27 ng/ml. These concentrations have been shown to be biologically relevant.³⁰

Figure 6.

(a) The concentration of VEGF-a detected in DPSC-conditioned media, standardized to total post-culture protein deficit. N = 3 biological replicates per sample. (b) Comparison between in vitro BBB treated with VEGF-a and DPSC-conditioned media, with or without cediranib. N = 9 experimental replicates per group. Error bars depict standard deviation.

Next it was investigated whether VEGF-a expression in DPSC-conditioned media was the dominant biological active factor using our in vitro BBB assay. Cediranib, a VEGF-R2 antagonist,³¹ was used to block VEGF-a activity. The in vitro BBB was treated with DPSC-conditioned media with or without the addition of cediranib. Recombinant human VEGF-a was used as a positive control. It was found that cediranib significantly attenuated DPSC-mediated BBB permeability by approximately fourfold (Figure 6(b)).

Discussion

In this study the data demonstrate the appropriateness of a previously characterized in vitro model of the BBB for assessing the permeability-mediating potential of soluble stem cell factors in culture medium. Using this model, we have demonstrated that DPSC expresses VEGF-a in biologically significant quantities to mediate an increase in BBB permeability.

The BBB is selectively impermeable to large, polar molecules, though penetration is dependent on a multitude of other factors.^32,33 This is consistent with the findings presented in Figure 3(a), which suggest that the in vitro BBB may be used to assess the ability of substances to diffuse across the BBB. The low level of background penetration may be attributed either to perfusion by unbound EB, which has a molecular weight of 960 Da, or to minor discontinuities in the endothelial monolayer. Obviously, molecular size is only one predictor of BBB penetration; other factors include polarity and lipid solubility.³⁴ This may explain the tendency for methylene blue to equilibrate more readily than bromophenol blue, despite being smaller.

To characterize the molecular mechanisms by which the in vitro BBB may be modulated, occludin expression was studied. Occludin is one of the main components of the tight junction assembly and its expression is therefore characteristic of the BBB.³⁵ The in vitro BBB was treated with either VEGF-a, DPSC-conditioned media or vehicle and stained for occludin. BMEC grown in co-culture with astrocytes showed greater staining than BMEC grown in isolation. This is consistent with the observation that astrocytes induce and maintain BBB characteristics in BMEC. Co-cultures treated with VEGF-a showed a decrease in expression of occludin. This is consistent with previous studies showing that (i) VEGF-a causes BBB permeability by promoting tight junction dysregulation²¹ and (ii) VEGF-a expression is responsible for BBB leakage in the post-ischaemic brain.²⁰

It may be noted that in many studies, occludin expression is punctate in appearance and localized at the intersection between endothelial cells. This was not seen in the present study. Localization of occludin to the cell membrane is dependent on phosphorylation by protein kinase C, which appears only to occur in the presence of the brain extracellular matrix. The cytoplasmic expression of occludin is still an indicator of tight junction regulation.^36,37

VEGF-a was demonstrated to cause a dose-dependent increase in permeability of the in vitro BBB. This was expected as VEGF-a-mediated BBB permeability is a receptor-mediated response, so incremental increases in VEGF-a concentration should continue to cause increases in permeability until saturation is reached. VEGF-a treatment only took place within a physiologically relevant range and so the dose–response relationship was best described as linear. The correlation between VEGF-a dose and in vitro BBB permeability was moderate. Random error in equilibration was likely due to variance in unmeasured variables such as the final quantity of BMEC adhered to each membrane.

The allocation of individual in vitro BBB assemblies in this exploratory study was not subject to randomization. Operators were also not blinded to the conditions of the experiments. As the factors that contribute to the precision and power of this construct cannot be tested directly, the inclusion of randomization and blinding will improve the strength of conclusions drawn from subsequent investigations.

Occludin expression in response to treatment of the in vitro BBB to DPSC-conditioned media was investigated. The results from this suggest that DPSC expresses soluble factors capable of causing tight junction dysregulation. It was then confirmed that soluble DPSC factors are capable of causing an increase in permeability of the in vitro BBB.

As DPSC caused permeability of the in vitro BBB in a similar manner to VEGF-a, this cytokine was investigated as a major soluble component of the conditioned media from DPSC. VEGF-a was shown by ELISA to be expressed in biologically significant quantities by DPSC, though the amount varied between the two human donors. To determine whether VEGF-a was responsible for inducing permeability, the in vitro BBB was treated with cediranib.³⁸ Treatment with cediranib significantly attenuated the permeability-inducing effect of DPSC by a factor of four. This suggests that VEGF-a is the major soluble component of DPSC-conditioned media responsible for inducing permeability in the in vitro BBB. These findings are consistent with previous literature implicating VEGF-a in BBB permeability and oedema.²²

Another possible mechanism to explain neuroendothelial transmigration of stem cells is the interaction between selectins and their endothelial ligands. This is similar to the mechanism involved in leukocyte migration into the parenchyma. Membrane-bound integrins bind to endothelial VCAM-1 and ICAM-1, triggering diapedesis and transmigration. This mechanism was investigated by Rosenblum et al. as an explanation for mesenchymal stem cell migration into the CNS following intra-arterial transplantation.⁵ The effects noted by this study may occur downstream of VEGF-a upregulation. VEGF-a has been observed to increase expression of VCAM-1 by vascular endothelial cells,³⁹ as well as increasing transendothelial migration.⁴⁰ It may be the case that trans-neuroendothelial migration of stem cells via selectin–integrin interaction is VEGF-a dependent. In this case, the role of VEGF-a can be considered to be twofold; being both responsible for increasing BBB permeability and stimulating adhesion and transmigration of transplanted stem cells.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Australian National Health and Medical Research Council project grant. The Peter Couche Foundation

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

JNW was responsible for carrying out all experimental work and data collection in this manuscript, as well as compiling draft versions. KLK was responsible for guidance during design of the experiments described in this manuscript, for aiding in data analysis and for the major revision of content prior to submission. SAK allocated funding support to this project, guided the study design, was responsible for interpretation of results and critically revised drafts of this manuscript. All authors approved this manuscript prior to submission.