Stress-sensing in the human greying hair follicle: Ataxia Telangiectasia Mutated (ATM) depletion in hair bulb melanocytes in canities-prone scalp

Canities (or hair greying) is an age-linked loss of the natural pigment called melanin from hair. While the specific cause(s) underlying the loss of melanogenically-active melanocytes from the anagen hair bulbs of affected human scalp remains unclear, oxidative stress sensing appears to be a key factor involved. In this study, we examined the follicular melanin unit in variably pigmented follicles from the aging human scalp of healthy individuals (22–70 years). Over 20 markers were selected within the following categories: melanocyte-specific, apoptosis, cell cycle, DNA repair/damage, senescence and oxidative stress. As expected, a reduction in melanocyte-specific markers in proportion to the extent of canities was observed. A major finding of our study was the intense and highly specific nuclear expression of Ataxia Telangiectasia Mutated (ATM) protein within melanocytes in anagen hair follicle bulbs. ATM is a serine/threonine protein kinase that is recruited and activated by DNA double-strand breaks and functions as an important sensor of reactive oxygen species (ROS) in human cells. The incidence and expression level of ATM correlated with pigmentary status in canities-affected hair follicles. Moreover, increased staining of the redox-associated markers 8-OHdG, GADD45 and GP-1 were also detected within isolated bulbar melanocytes, although this change was not clearly associated with donor age or canities extent. Surprisingly, we were unable to detect any specific change in the expression of other markers of oxidative stress, senescence or DNA damage/repair in the canities-affected melanocytes compared to surrounding bulbar keratinocytes. By contrast, several markers showed distinct expression of markers for oxidative stress and apoptosis/differentiation in the inner root sheath (IRS) as well as other parts of the hair follicle. Using our in vitro model of primary human scalp hair follicle melanocytes, we showed that ATM expression increased after incubation with the pro-oxidant hydrogen peroxide (H2O2). In addition, this ATM increase was prevented by pre-incubation of cells with antioxidants. The relationship between ATM and redox stress sensing was further evidenced as we observed that the inhibition of ATM expression by chemical inhibition promoted the loss of melanocyte viability induced by oxidative stress. Taken together these new findings illustrate the key role of ATM in the protection of human hair follicle melanocytes from oxidative stress/damage within the human scalp hair bulb. In conclusion, these results highlight the remarkable complexity and role of redox sensing in the status of human hair follicle growth, differentiation and pigmentation.


Scientific Reports
| (2020) 10 Double immunohistochemistry of scalp tissue. This assay was conducted as described previously and adapted to the antibodies used 61 . Briefly, ten micron sections were air-dried onto slides for at least 1 h prior to fixing in ice-cold acetone for 10 min at − 20 °C. Non-specific antibody binding was reduced by incubating in 10% donkey serum (Sigma-Aldrich, UK) diluted in PBS for at least 30 min, followed by simultaneous incubation in primary antibodies (a melanocyte marker with a target antibody; see Supplementary table 1) diluted in PBS containing 1% donkey serum overnight at 4 °C. After washing sessions in PBS for 3 × 10 min, these were then incubated in donkey Alexa-488 and 594 conjugated secondary antibodies (1:100 dilutions, Thermofisher, UK) for 1 h at RT. After washing the sections were mounted for confocal microscopy under sealed coverslips in fluorescent mounting medium containing DAPI nuclear stain (VectorLabs, UK). Images were collected using the 365 nm (DAPI), 488 nm (Alexa-488) and 543 nm (Alexa-594) channels on a Zeiss LSM confocal microscope by sequential line scanning. Images were processed using the LSM confocal image browser software (Zeiss, UK) and ImageJ (freeware).
Treatment of hair follicle melanocytes in culture. Prior  www.nature.com/scientificreports/ Membranes were incubated in either Abcam Cat# ab32420 Rabbit anti-ATM [Y170] antibody diluted 1:4000 in 5% milk/TBS-T overnight at 4 °C, then rinsed in TBS-T. Membranes were then incubated for 30 min at RT in HRP-conjugated secondary antibody (Santa Cruz Biotechnology, Inc, USA) diluted in 5% milk or BSA in TBS-T. Separated membranes containing WesternC marker were incubated in Streptavidin HRP-conjugated label (1:20,000) and secondary antibody only. Excess secondary antibody and Streptavidin-HRP was removed by washing in 4 changes of TBS-T. Cut membranes were pieced back together and incubated in Clarity ECL detection reagent (Bio-Rad, UK) for 5 min then imaged using the ChemidocMP imaging system (Bio-rad, UK). Protein bands were quantified by densitometry using Image Lab v4.1 software (Bio-Rad, UK). For Catalase and GP-1 Western blotting samples were prepared as described earlier, however proteins were electrophoresed through Bio-Rad Any kD mini-TGX PAGE gels alongside Kaleidoscope and Western C markers (Bio-Rad, UK), then transferred to PVDF using a Turbo Transfer blotter (2.5 mA, 25 V constant for 10 min). Membranes were blocked as earlier in 5% Milk/TBS-T then probed overnight with mouse anti-Catalase primary antibody (Sigma #C0979, diluted 1:1000 in 5% milk/TBS-T) at 4 °C then rinsed in TBS-T buffer. Membranes were then incubated for 30 min at RT in HRP-conjugated secondary antibody (Santa Cruz, Biotechnology, Inc, USA) diluted in 5% milk in TBS-T and imaged as described previously. Where necessary, membranes were stripped using Restore reagent (Thermofisher, UK) then reprobed for the housekeeping gene GAPDH (1:2000 Cat# MCA4740, Bio-Rad, UK) as a loading control.
Cell viability test. Primary HFMs were seeded into 96-well plates at a density of 1 × 10 4 cells/well (n = 12) in full 2:1 Melanocyte media. Cells were incubated at 37 °C with 5% CO 2 overnight to attach to plates, rinsed in PBS, then placed in starved media overnight before treatment. Treatments were prepared immediately prior to addition with a final dilution into starved media then added to cells. DMSO vehicle (basal) controls (0.00005% v/v) were also used in parallel. Following incubation, cells were rinsed in PBS then incubated with WST-1 reagent (diluted 1:10 into fresh starved media (Roche, UK)) and incubated at 37 °C for 4 h in a cell culture incubator according to manufacturer's instructions. Plates were scanned for absorbance at A 450nm on a Tecan M200 plate reader using a reference wavelength of A 690nm . Results were analysed using Magellan 6.0 software (Tecan, Germany). Mean absorbance relating to cell viability was calculated then plotted on a bar chart (Graphpad Prism v6.0).
Statistics. Data were expressed as mean ± SEM. Error bars also show SEM. Differences between data were evaluated by Student's t test or ANOVA where stated after testing the data normality using Graphpad Prism v6.0.

Results
Human scalp from multiple normal healthy donors was assessed for the expression of canities-relevant modulators of melanocyte phenotype using a panel of 23 different antibodies by immunofluorescence. Antibodies were directed against proteins implicated in key pathways such as apoptosis, cell cycle regulation, senescence, DNA repair/damage and oxidative stress. Double-staining was performed using a melanocyte-specific marker to investigate colocalization in bulbar HFM. Brightfield imaging was used in parallel to monitor the pigmentation status (melanin level) of each stained hair bulb.

Expression of melanocyte specific markers in greying human hair follicles. The correlation
between the pigmentation status and the melanocyte-specific markers (GP100, Tyrosinase related protein-1 (TRP-1) and Tyrosinase (TYR) was evaluated in human scalp hair bulbs. The first observation was that scalp tissues from donors aged from 22 to 70 years old, contained variably pigmented follicles. Multiple donors had fully pigmented follicles co-located with intermediary to weakly/non-pigmented hair follicles (data not shown).
Comparison of melanocytes staining with melanin content of hair bulbs showed the expected gradual decrease in melanocyte labelling from fully pigmented to weakly/non-pigmented hair bulbs ( Fig. 1 and supplementary  figure 1). The intensity and the distribution level of the staining of TRP-1 and GP100 was higher in the hair follicle melanocytes of fully pigmented follicles, sporadic in melanocytes with reduced pigmentation and lacking in weakly pigmented/white hair follicles (Fig. 1). The intensity of staining of TYR was also correlated to the pigmentation status of the hair bulb. However, the protein exhibited lower levels of expression and distribution in the hair bulb melanocytes compared to those of TRP-1 and GP100. For instance, TYR staining was markedly reduced in the melanocytes of hair follicles with mild or moderate canities, despite strong TRP-1 expression in the same hair follicle (see supplementary figure 1). The lack of full overlapping between the three-melanocyte lineage-specific markers suggests differential labelling of melanocyte populations. The melanocyte-specific marker TRP-2 was also investigated. However, the hair bulb of anagen hair follicles lacked TRP-2 expression as previously reported 64 .
ATM staining correlates with pigmentary status in hair follicles. A major finding of the immunofluorescence study on scalp tissues was that in addition to the melanocyte specific markers, another protein appears to be strongly related to the hair bulb melanocytes and their pigmentary status. Expression of the DNArepair coordinator and oxidative stress sensor ATM was exclusively restricted to hair follicle melanocytes (i.e. was not expressed in follicular keratinocytes and fibroblasts) with a pattern that correlated with pigmentary status of the hair follicle within the hair bulb (Fig. 2). Heavily pigmented hair follicles showed strong nuclear ATM colocalisation within bulbar melanocyte nuclei. In hair follicles with reduced overall pigmentation, hair bulb melanocytes displayed variably reduced expression or absent ATM expression. ATM expression was completely absent in white hair follicles without pigmented cells (i.e. full canities). It was also noted that frequent strong ATM nuclear staining was colocalised within the nuclei of GP100 positive melanocytes and occasional www.nature.com/scientificreports/ basal keratinocytes in the UVR exposed epidermis of multiple scalp samples. Nuclear ATM staining was not detected within dermal fibroblasts (supplementary figure 2), and although some cytoplasmic expression was visible. Among the biomarkers implicated in cell cycle, DNA damage/repair, apoptosis, senescence and oxidative stress, only ATM was consistently expressed in hair follicle melanocytes in situ and correlated to the pigmentary status of the hair bulb. By contrast, only occasional colocalisation with isolated bulbar melanocytes was detected for GP-1, 8-OHdG and GADD45; however, their expression pattern or level did not correlate strongly with overall pigmentary status or with age of donor (Fig. 3). All other markers tested did not appear to be expressed to any significant extent with melanocytes.

Novel expression of multiple markers for ROS and apoptosis in differentiating inner root sheath (IRS) of anagen scalp hair follicles.
A striking feature of this study was the intense but highly  www.nature.com/scientificreports/ unclear if it was located within peripheral hair follicle melanocytes or in DP cells. We did not detect any CASP-3, phospho-p53, p27/phospho-p27 and phospho-p38 expression in these samples.

ATM protein levels increase in hair follicle melanocytes in vitro in response to elevated oxidative stress.
In vitro analysis of primary hair follicle melanocytes (HFMs) isolated from scalp tissue showed strong nuclear expression of ATM (Fig. 4a). To investigate if ATM protein levels change in these melanocytes during elevated oxidative stress, these cells were treated with 40 µM H 2 O 2 for 1 h to induce ROS. Protein levels of total ATM and the ROS-activated S1981-phosphorylated ATM were investigated using Western blot analysis. Results showed increased total ATM levels in HFMs after H 2 O 2 treatment, and this increase was blocked by preincubation of HFMs with antioxidants VitE/Q for 1 h (*p < 0.05 VitE/Q pretreatment versus untreated control; Fig. 4b,c). However, expression of ROS-activated phosphor-ATM S1981 could not be detected despite elevated levels of catalase in these cells in response to H 2 O 2 treatment (Fig. 4d,e and supplementary figure 4).

Inhibition of ATM kinase activity in HFM in vitro reduces cell viability. To investigate if ATM
activity influenced melanocyte survival under sustained elevated ROS conditions in culture we pre-treated cells with the ATM-specific kinase inhibitor KU60019 (which abrogates the kinase function of the protein and prevents downstream phosphorylation of ATM targets 65 ) and incubated these cells with an increased concentration   66 ) to raise ROS levels, for short (1 h) and long (24 h) durations then measured cell viability. As seen in Fig. 5, results showed that in the absence of ATM kinase activity, sustained H 2 O 2 treatment for 24 h significantly reduced cell viability compared to controls (***p < 0.001 ANOVA). Treatment with the potent ROS inducer menadione 67,68 also significantly reduced cell viability at both 1 h and 24 h time points compared to untreated controls (***p < 0.001 ANOVA). Treatment of HFM with only 5 µM KU60019 was found to reduce total mean ATM protein levels compared to untreated cells (supplementary figure 5).

Discussion
Despite several studies into the biology and genetics of hair greying or canities, the mechanism by which human scalp hairs lose pigmentation remains poorly understood, due in part to the complexity of interaction between different cell types, complex protein interactions, differential gene expression during the hair growth cycle, and variable conditions that occur within the hair follicle microenvironment 1,7-10,13,41,60,69-74 . Here we have used a specific panel of antibodies chosen on the basis of their involvement in DNA repair, oxidative stress, cell cycle dynamics, senescence and apoptosis to investigate potential markers of canities within hair bulb melanocytes in the scalp of multiple normal healthy donors aged from 22 to 70 years old. One of the major findings of this study was that of all the biomarkers tested, only ATM was found to be exclusively expressed within hair bulb melanocytes and so correlated with hair follicle pigmentation status. ATM nuclear expression markedly decreased in the melanocytes of hair follicles with reduced pigmentation. ATM was also absent in some hair bulbs that contained few remaining GP100 positive melanocytes. www.nature.com/scientificreports/ ATM is a PI3K-kinase and one of the major components of the DNA damage response (DDR) pathway 46 but can also be activated during oxidative stress 50,51 . It exists as an active dimer 45 and targets phosphorylation of downstream targets such as CHK2, γH2AX and p53 44,[47][48][49] . Investigations on primary hair follicle melanocytes (HFM) in vitro showed that total levels of ATM were upregulated after exposure to hydrogen peroxide that increased oxidative stress. Nishimura and colleagues showed that loss of ATM in melanoblast stem cells within the hair follicle bulge region in mice could sensitize the stem cells to genotoxic stress, leading to early differentiation, depletion from the niche and progressive hair greying over subsequent cell cycles due to eventual depletion of mature melanocytes from the hair bulb 3,10 . Oxidative stress is associated with the auto-phosphorylation of ATM at S1981 51 , but this was not detected by Western blot analysis in the current study. Hydrogen peroxide is known to be a short-lived molecule in cell cultures that is degraded within 30 min. However during that time it can be a potent activator of cellular anti-oxidant response systems, and if high enough can even induce DNA damage and cause cell death 75 . It has been speculated that when peroxide activates ATM in vitro little autophosphorylation takes place and it may be partially dependent on complexing with the MRN complex recruited to DNA repair 51,76 . This may explain why we could not detect ROS activated S1981 phosphorylation of ATM, particularly if the H 2 O 2 concentrations were not high enough to induce DNA damage in HFMs. To determine if levels of ROS were increased during treatment's we measured catalase and GP-1 proteins levels by Western blot and both were markedly increased in HFM's after addition of H 2 O 2 indicating that the ROS stress response within cells was active. If auto phosphorylation of ATM S1981 is absent and the antioxidant response is functioning in HFM then does ATM play a role in melanocyte survival during ROS induced stress? To investigate this we used the ATM specific kinase inhibitor KU60019 65 to inhibit the downstream phosphorylation activity of ATM with both H 2 O 2 and menadione. The naphthoquinone menadione is metabolized by several enzymes through one electron reduction resulting in highly reactive unstable semiquinones 77 . These enter redox cycles to reform quinones but produce high amounts of free radical oxygen that are subsequently deactivated by dismutase's, producing elevated H 2 O 2 67,68,77 . We detected increased sensitivity to ROS and decreased cell viability in both cases, especially after a sustained increase in oxidative stress over 24 h. This indicates that although ATM may not be activated through traditional ROS associated S1981 phosphorylation in HFM, functional ATM is still important to bulbar melanocyte survival during elevated oxidative stress. The exact mechanism by which ATM is activated by and which downstream proteins are targeted needs further investigation. Casual observation during treatments showed that HFM appeared to exhibit mild cytotoxicity effects to external H 2 O 2 around the 250 uM concentration and levels above this would produce mass cell death/apoptosis in cultures within 24 h of treatment. We postulate that functional melanocytes may have multiple intrinsic robust mechanisms to process H 2 O 2 induced ROS due to the natural generation of H 2 O 2 during melanogenesis itself, although further work is needed to investigate this. KU60019 has been well documented as a very potent ATM kinase inhibitor, used in multiple ATM publications at similar concentrations as our study. Thus, we would expect ATM activity to be completely suppressed [78][79][80][81][82][83][84][85][86] . Treatment of HFM with the KU60019 kinase inhibitor alone reduced total mean ATM protein levels compared to untreated cells. This would be expected as other studies have documented that ATM phosphorylation can increase total ATM expression through a feedback mechanism [87][88][89] . However, we cannot www.nature.com/scientificreports/ completely rule out secondary effects on closely related cell signalling pathways such as ATR/Chk1 mediated DNA repair which can in part be influenced by reductions in ATM kinase activity 90,91 . In our study we only detected co-localisation and increased staining of three other proteins (GP-1, GADD45 and 8-OhDG) within bulbar melanocytes. Other protective mechanisms may cooperate with ATM expression/activity to protect bulbar melanocytes from oxidative stress-induced terminal differentiation similar to melanocyte stem cells within the murine hair follicle niche 10 . Kauser et al., showed that reduced expression of catalase in HFM cultured from elderly donors compared to HFM isolated from young donors, indicating that catalase expression/activity is an important effector in the response of melanocytes to aging-associated stimuli over time 92 . Other studies have shown the importance of the NRF2/HO-1 proteins in directing the ARE (antioxidant response elements) pathway in melanocytes in response to H 2 O 2 induced stress 41,42 . Coincidently, ATM is thought to regulate selective mitochondrial functions through NRF1 (a similar protein of the NRF2 family) in response to ROS 93 . It has been shown that melanocytes can produce peroxide as a consequence of melanin synthesis [25][26][27][28] . Hair bulb melanocytes already have robust mechanisms in place using the cells own intrinsic antioxidant enzymes to reduce elevated ROS before critical levels that damage DNA occur. We would speculate that ATM is part of a larger cohort of proteins that are tightly regulated to cope with elevated ROS levels and that together they contribute to protecting melanocytes from oxidative stress and ensure their survival within the hair bulb. Changes in expression levels of these marker proteins could imbalance ROS levels, leading to a gradual reduction in melanocyte protection and viability over time leading to hair greying. As loss of ATM correlates with decreased pigmentation in hair follicle melanocytes it is difficult to determine if loss of melanin contributes to a reduction in ATM expression or drives the process of ATM loss more rapidly. Melanin production is known to decrease and become less efficient with aging in melanocytes, with fewer melanosomes produced and poor quality/quantity of melanin 8,71 . We also specifically analysed scalp tissue from skin types II-III in this study and so could not detect differences between eumelanotic and phaeomelanotic phenotypes. Further work is therefore needed to investigate this.
Elsewhere in the hair follicle, several other cell cycle, DDR, apoptosis and ROS metabolising anti-oxidant markers appear to have very specific expression in specialised cell layers of the hair follicle, particularly those cells undergoing differentiation in the anagen-specific IRS during hair fibre formation. Different compartments or cell layers of the hair follicle may have their own individual 'microenvironment' , with varied requirements to protect cells from ROS caused by normal cell activity such as differentiation. Lemasters et al., showed that hair follicles have specific compartments of mitochondrial gradients and oxidative metabolism, which leads to increased ROS during the development of the hair shaft that they termed a 'ring of fire' 94 . Another study showed that mice with a keratinocyte-specific deficiency in mitochondrial transcription factor A (TFAM), a gene required for the transcription of mitochondrial genes, have abnormalities in hair follicle growth, and that mitochondriagenerated ROS was in fact necessary for the mediation of cell differentiation 95 . Indeed, a high proportion of the markers tested in our study localised to where cells differentiate to establish the IRS, and it appears their presence is necessary to prevent damage or to stabilise ROS in cells during the differentiation process. Further studies are needed to determine if regulated expression of p21, CASP-2, CASP-9, SOD-1 and HO-1, components of the anti-oxidant glutathione metabolic pathway GP-1 and GR-1, and the DNA damage marker GADD45 are necessary to complete differentiation in specialised keratinocyte layers.
In summary we have shown that ATM protein expression correlates with pigmentation status in greying hair follicles and contributes to sustained melanocyte viability in vitro during elevated ROS levels. We have also showed that other markers of DNA damage/cell cycle (p21, p53, p57 and phospho-p57, GADD45, 8-OHdG), apoptotic (BCL-2, CASP-2, CASP-9) and anti-oxidant pathways (GP-1, SOD-1, HO-1, p38 stress kinase), that are not specifically associated with hair follicle pigmentation, have distinct expression patterns and colocalise to parts of the hair follicle such as the IRS, ORS, DP and cortex. These results highlight the multicellular complexity of the hair follicle to cope with oxidative stress and indicate novel previously unreported roles for several proteins implicated in cell cycle/senescence and oxidative stress in IRS differentiation pathways and hair follicle growth. Further work is needed to establish how ATM and other proteins co-operate to reduce damage to the hair follicle caused by the regular cellular processes during hair growth and/or pigmentation and the external factor associated with the chronological aging/greying process.