Nicotinamide effects on the metabolism of human fibroblasts and keratinocytes assessed by quantitative, label-free fluorescence imaging

: Alterations in metabolism are central to the aging process. Therefore, understanding the subcellular functional and structural changes associated with metabolic aging is critical. Current established methods for exploring cell metabolism either require the use of exogenous agents or are destructive to the tissue or cells. Two-photon excited fluorescence (TPEF) imaging has emerged as a method for monitoring subtle metabolic changes non-invasively. In this study, we use TPEF imaging to acquire high-resolution fluorescence images from two coenzymes, NAD(P)H (reduced form of nicotinamide adenine dinucleotide) and FAD (flavin adenine dinucleotide), within human fibroblasts and keratinocytes in response to B3 (a nicotinamide precursor) supplementation and/or UV irradiation, without addition of exogenous labels. In addition, multi-parametric analysis methods are used to extract functional information of cellular metabolism, including cellular redox state, NAD(P)H fluorescence lifetime, and mitochondrial organization. Our results demonstrate that such optical metabolic assessments can serve as sensitive, label-free, non-destructive reporters of known effects of B3 to maintain and in some cases even enhance the respiratory function of mitochondria, while lowering oxidative damage. Thus, TPEF imaging, supported by highly-quantitative analysis, can serve as a tool to understand aging-dependent metabolic changes as well as the effect of actives on human epidermal and dermal cells.


Introduction
Cellular metabolism includes all the chemical reactions supporting key biological pathways required to maintain life. The complex sequences of these well controlled biochemical reactions underpin the cell's ability to transform energy and quality of life. The proper functioning of these processes is especially critical to maintaining a healthy aging trajectory. In human skin, the aging process is associated with changes in metabolism and cellular energy transduction [1]. As a person ages, dysfunction of energy-yielding metabolic processes lowers the capacity to regenerate, repair, and protect the skin itself, which results in undesired phenotypical changes in skin properties such as wrinkling, pigmentation, and loss of elasticity [2,3]. Since metabolic dysfunction correlates with skin appearance, repair of these processes to defy or even reverse the aging process has great interest in skin care research and development.
Nicotinamide adenine dinucleotide (NAD) exists in both reduced (NADH) and oxidized (NAD + ) forms in eukaryotic cells. It is a critical redox cofactor in the production and utilization of cellular energy. During aging, NAD is gradually depleted in multiple tissues and has been proposed as a master regulator of age-dependent pathology [4,5]. Its depletion induces mitochondrial mitochondrion is the main organelle for energy production in skin cells, energy metabolism is likely a key factor affecting the organization of mitochondria during aging [20].
Label-free optical measurements not only enable live assessments on living cells with time course of treatments without sacrificing the samples but also have many potential applications on human skin in vivo [21][22][23]. Based on TPEF images of cellular NAD(P)H and FAD, we have developed multi-parametric analysis methods to assess cellular metabolism [24][25][26][27]. The optical biomarkers that we developed and validated include the optical redox ratio, defined as the fluorescence intensity ratio of FAD/(FAD + NAD(P)H), the NAD(P)H bound fraction, which indicates the protein binding status of NAD(P)H generated from the NAD(P)H autofluorescence intensity decay profile (i.e. the fluorescence lifetime), and mitochondrial clustering, which reflects the fusion and fission status of mitochondria for optimizing energy production. These optical biomarkers provide complementary insights into cellular metabolism and underlying biological mechanisms experienced by the cells [28]. To gain a better understanding of cellular metabolism in skin, here we use TPEF microscopy as well as multi-parametric analysis to assess the bioenergetic and metabolic status of human keratinocytes and fibroblasts, the main cellular components of the skin epidermis and dermis, with these cells subject to supplementation of B3 and/or UV exposure. B3 is a nicotinamide precursor, which is known for its aging-related, metabolic role in protecting cells from oxidative stress and increasing mitochondrial efficiency [29][30][31][32]. It has also been reported to promote the repair of DNA damage and delay photoaging induced by UV [33,34].

Materials
Normal human dermal fibroblasts and keratinocytes were used in the studies (Thermo Fisher Scientific, Waltham, MA). The cells were cultured at 37°C in a humidified, 5% CO 2 incubator. The fibroblasts were seeded onto glass-bottom 12-well plates (MatTeK, P12G-1.5-14-F) at a confluency of ∼20%, and cultured up to 21 days. The culturing medium for fibroblasts was complete DMEM (Thermo Fisher Scientific, Waltham, MA) with 10% FBS (Thermo Fisher Scientific, Waltham, MA). The keratinocytes were seeded onto glass-bottom 12-well plates (MatTeK, P12G-1.5-14-F) at a confluency of ∼15%, and cultured for 7 days to avoid detachment from the glass substrate after reaching a high confluency level. The culturing medium for keratinocytes was EpiLife Media (Thermo Fisher Scientific, MEPI500CA) with supplement (Thermo Fisher Scientific, S-001-5). To assess the effect of B3, cells were cultured with medium supplemented with or without B3 at a concentration of 10 mM (Sigma Aldrich, St. Louis, MO). Metabolism of keratinocytes was also assessed in the presence or absence of UV irradiation. The media was changed every two days during cell culture. For both fibroblasts and keratinocytes, TPEF imaging was performed typically every two or three days, so that the disruptions to cell cultures would be minimized while sufficient time-dependent information could be preserved.
To study the effects of B3 on UV exposed keratinocytes, three conditions were used, i.e., keratinocytes in culture using standard media without UV (control), keratinocytes in standard media exposed to UV (UV), and keratinocytes cultured in media supplemented with B3 (10 mM) and exposed to UV (B3+UV). The cells were first incubated for 3 days with culture media and then on Day 4, the UV irradiation was started and applied every day from Day 4 to Day 6. Each time before UV exposure, the media was removed and replaced with PBS. The UV exposure lasted for 14 min with dosage of 4 J/cm 2 of UVA + UVB. Each time we finished the UV exposure, we changed the medium from PBS back to the corresponding treatment group media. The TPEF images were acquired on Day 1, 3, 5, and 7, where Day 1 was considered a time approximately 24 hours following seeding; thus Day 5 corresponded to 24 hours post the onset of the first irradiation treatment. On Day 5, we first performed the imaging, and then applied the second UV exposure.

Image acquisition and data analysis
TPEF imaging was used to assess cellular metabolic state. HEPES (Sigma) was added to the culture media at a final concentration of 20 mM before imaging to maintain the pH of cell cultures throughout the entire imaging session. During the imaging session, the temperature was maintained at 37°C and the environment was kept humidified. Images were obtained using a Leica TCS SP8 confocal microscope equipped with a tunable (680-1300 nm) fs laser (InSight Deep See; Spectra Physics; Mountain View, CA). Images (1024 × 1024 pixels; 386 × 386 µm) were acquired using water-immersion 25× objective (NA 0.95; 2.4 mm working distance), with simultaneous collection by two non-descanned Hybrid Detectors (HyDs). To isolate NAD(P)H fluorescence, a 460(±20) nm emission filter, corresponding to the NAD(P)H emission peak, was placed before one of the non-descanned detectors. NAD(P)H fluorescence images were acquired from this 460 nm channel using 755 nm excitation. FAD fluorescence was isolated using a 525(±25) nm emission filter for the other non-descanned detector and 860 nm excitation. The fluorescence lifetime images (512 × 512 pixels; 386 × 386 µm) corresponding to NAD(P)H were acquired under the same excitation and emission settings, using a commercial time-correlated single photon counting (TCSPC) electronics module. The incident laser power was ∼15 mW at the sample. Both the gain of HyDs and laser power were recorded for each image and used to normalize the recorded fluorescence intensity. In this study, the gain was kept the same for the two HyDs and throughout all the experiments; therefore, the image intensity was calculated and normalized by the square of the incident laser power, as measured at the sample plane prior to each experiment. We prepared three independent experimental repeats for the fibroblast study, and two independent repeats for keratinocyte study, either without or with UV exposure. For each experimental repeat, four wells of cell culture were prepared for each group, and at least three fields were imaged per well. The imaging time was approximately 5 min for each well.
Optical metrics of redox ratio, NAD(P)H bound fraction, and mitochondrial clustering were generated as quantitative measures of cellular metabolism, as previously reported [24]. The redox ratio was defined as FAD/(FAD + NAD(P)H) using the endogenous fluorescence intensity from FAD and NAD(P)H molecules. The mean redox ratio was acquired by averaging the pixel-wise redox ratio values within the cell cytoplasm areas only, excluding background and nucleus. It was demonstrated previously that this optical redox ratio linearly correlated with the biochemical redox ratio NAD + /(NAD + +NADH) that was typically measured using LC-MS [17,25]. This redox ratio was an index of the reducing power and oxidative stress in cells [35]. NAD(P)H bound fraction was the estimated fractional amount of the NAD(P)H bound to proteins acquired from the analysis of the fluorescence lifetime images using the phasor approach [24,36]. Briefly, real and imaginary parts of the Fourier transform of the fluorescence intensity decay profile at each pixel were used to determine the two coordinates of a phasor. The fluorescence lifetime profile characterized by a mono-exponential decay was mapped onto a point that exactly fell on the universal semicircle of the phasor plot, while more complicated decay curves were represented by points within the semicircle. If a fluorescence decay curve was a bi-exponential function, its phasor fell on a line within the semicircle, with the two points where the line intersected the semicircle representing the short and long lifetime components, respectively (Fig. S1). The relative distance of the point on that line provided an estimate of the fractional contributions of the free (short lifetime) and bound (long lifetime) NAD(P)H components; therefore, NAD(P)H bound fraction was estimated based on the location of the centroid of ellipses that represented the distributions of the fluorescence decay data from cell cytoplasm areas (Fig. S1) [24]. Bound NAD(P)H had longer fluorescence lifetime compared to free NAD(P)H [37]. Since bound NAD(P)H molecules were primarily found in mitochondria, this index may be used to infer the state of energy (in the form of reducing power) shuttling between mitochondria and the cytosol [38].
Mitochondrial clustering was used as a metric to represent the dynamic changes in mitochondrial organization resulting from fission and fusion [27,39,40], which regulated mitochondrial networks in response to environmental stimuli, developmental status, and energy requirements of the cell [41,42]. Briefly, we normalized the NAD(P)H TPEF intensity image so that the resultant intensity varied between 0 and 1. Then we applied a custom bandpass filter to remove nuclear and interstitial space by combining 3 separate bandpass filters. Specifically, the first bandpass filter was formed by multiplying a Gaussian high-pass (0.01 µm -1 ) and a Gaussian low-pass (0.1 µm -1 ) filter. The second one was also a combination of Gaussian high-pass (0.021 µm -1 ) and low-pass (0.143 µm -1 ) filters, respectively. The final one was created by combining 3rd order Butterworth low-pass (0.2 µm -1 ) and high-pass (0.021 µm -1 ) filters. After each individual bandpass filter was computed, they were multiplied to create the final custom bandpass filter. Then the image intensity patterns within the cell cytoplasmic areas were cloned and randomly positioned in the image background to create a new image without distinct cell borders and only cell mitochondrial patterns spanning the entire image (Fig. S2). Upon Fourier transformation a power spectral density (PSD) curve was created for each image. The PSD followed a power law behavior at high spatial frequencies (> 0.1 µm −1 , corresponding to the size of mitochondria), suggesting a fractal organization of mitochondria ( Fig. S2(d)). We then fitted this linear portion with an equation of the form R(k) ∼ k −β between 0.1 µm −1 and the frequency at 98% of the entire PSD region (marked by the blue asterisk in Fig. S2(d)), and acquired the exponential power, β, as a quantitative measure of the mitochondrial clustering throughout this study [39,40]. To account for the impact that the random nature of the clone stamping procedure has on the value of β, we repeated the process twenty times for each field and recorded the average value. Increased mitochondrial clustering values typically represented more fissioned/fragmented organization, and lower clustering corresponded to more highly-networked mitochondria, as validated by studies using exogenous fluorescent mitochondrial markers [17,27]. This technique worked since images of NAD(P)H fluorescence predominantly reflected contributions from the bound form of NAD(P)H that resided primarily within mitochondria, owing to its increased fluorescent yield in that state [27,43].
To assess the cell confluency, we created a binary mask which removed the background area only, while keeping the nucleus area, based on the intensity gradient in NAD(P)H intensity image between nucleus and background area, since a certain level of NAD(P)H still existed in the nucleus. Briefly, six-level Otsu intensity thresholding was applied to each NAD(P)H image, and five thresholds divided the intensity to six levels. The lowest level was designated as background noise, and regions assigned to the upper five levels defined the cytoplasm and nucleus areas. All the pixels within this mask were summed, and the ratio between this summed number and the total pixel number of the image was calculated as the cell confluency. Cell confluency was recorded for both fibroblasts and keratinocytes for all treatment groups and measurement time points.

Statistical analysis
An ANOVA with post-hoc Tukey HSD test was performed to assess significant differences among different treatment groups using JMP 15 (SAS Institute). For statistical analysis, n was 12 wells for the fibroblast study, and 8 wells for keratinocyte study, either without or with UV exposure. There was nesting of wells and fields to avoid discarding any data information. Differences were considered significant at p < 0.05.

B3 effect on human dermal fibroblasts
Since B3 was known to have beneficial anti-aging effects on skin hyperpigmentation and wrinkles, we sought to assess its optical metabolic effects first on fibroblasts and then on keratinocytes. Time-dependent imaging of the same fibroblast cultures was performed over the course of 21 days. Representative images highlighting key metabolic changes recorded over 21 days of culture with or without B3 supplementation are shown in Fig. 1. Changes in the hues of the images in panels (a) and (b) corresponded to differences in redox ratio and NAD(P)H bound fraction values, respectively; changes in hues in panel (c) were more representative of differences in mitochondrial organization. Fibroblasts reached confluency under the conditions examined around Day 8 (Fig.  S3). The redox ratio hues of the control group became more yellowish to reddish at later time points, corresponding to a higher level of redox ratio, while supplementation with B3 led to maintenance of redox at more greenish to bluish hues (i.e., lower redox ratio) after Day 14 compared with control ( Fig. 1(a)). Similarly, the switch from blue to green and reddish hues in the NAD(P)H bound fraction maps corresponded to higher levels of bound fraction, which was observed in the B3 group compared to corresponding controls at later time points (Fig. 1(b)). As indicated in Methods, the levels of mitochondrial clustering were extracted from Fourier analysis of NAD(P)H intensity images that were processed to represent only the cytoplasmic NAD(P)H intensity fluctuations. These processed (intensity normalized, cytoplasmic area selected, and clone stamped) images are shown in Fig. 1(c). The patches of higher intensity (yellow hues) that were easily observed especially at the later culture time points within the control group corresponded to higher levels of mitochondrial clustering. B3 supplementation led to more uniform intensity distributions compared to control cultures at later time points (Fig. 1(c)). We note that when the mitochondrial distribution within the cytoplasmic area is not uniform, the clone stamped images maintain some patterns that are related to nuclear and cytoplasmic dimensions that may impact the value of the extracted clustering metric. Repetition of the clone stamping procedure twenty times for each field, at least partially mitigates such potential artifacts. Figure 2 shows quantitatively the changes in optical metabolic metrics detected from analysis of TPEF images acquired from fibroblasts cultured in the presence or absence of B3. The redox ratio increased with time as cells proliferated and differentiated under control media culture, with the change becoming statistically significant on Day 8 relative to Day 1 (Table S1). B3 supplementation led to lower redox ratio levels than corresponding controls after Day 8 ( Fig. 2(a)). The NAD(P)H bound fraction, shown in Fig. 2(b), increased in the first week and then levelled off (Table S1). It declined towards the third week for the control cells to levels similar to those at Day 1. Cells treated with 10 mM B3 showed a significant increase in NAD(P)H bound fraction starting on Day 8 (Table S1) which was sustained for the duration of the experiment and was significantly higher than the controls on Day 18 and Day 21 ( Fig. 2(b)). Consistent with the redox ratio trends, there was a statistically significant elevated level of mitochondrial clustering from Day 8 relative to Day 1 in the control group (Table S1), which was suppressed with B3 supplementation, resulting in significantly lower levels of mitochondrial clustering than what was observed in the control cultures as early as Day 8 (Fig. 2(c)).

B3 effect on human dermal keratinocytes
Similar investigations were conducted on human dermal keratinocytes. Representative keratinocyte images color-coded with hues that correspond to distinct levels of redox ratio, NAD(P)H bound fraction, and mitochondrial clustering are shown in Fig. 3. B3 led to similar metabolic alterations in keratinocytes as in fibroblasts. Specifically, when compared to control keratinocytes, cells treated with B3 exhibited more greenish to bluish hues in optical redox ratio maps (Fig. 3(a)), redder hues of NAD(P)H bound fraction levels ( Fig. 3(b)), and more uniform NAD(P)H intensity distributions in clone-stamped maps (Fig. 3(c)). The quantitative analysis results shown in Fig. 4 further confirmed the qualitative observations. The redox ratio gradually became lower and mitochondrial clustering became higher relative to Day 1 in the control group, while no significant differences were detected in the NAD(P)H bound fraction as a function of time (Table S2). As early as Day 3, keratinocytes showed significantly lower redox ratio ( Fig. 4(a)), increased NAD(P)H bound fraction (Fig. 4(b)), and decreased mitochondrial clustering (Fig. 4(c)) in response to B3 supplementation. However, the timing and consistency of the differences were not uniform across all metrics, suggesting that the activity of several metabolic pathways might be changing dynamically over the course of the assessments. Moreover, the overall levels of these optical metrics were different between keratinocytes and fibroblasts, potentially indicating that despite similar trends in changes of these optical assessments in terms of B3 supplementation versus control, different levels of metabolic regulation and/or distinct pathways might take place for different cell types.

B3 effect on keratinocytes with UV irradiation
Since keratinocytes are the primary cellular targets of solar UV radiation that reaches the earth's surface [44], TPEF imaging was also conducted on UV-irradiated keratinocytes to assess if B3 supplementation could delay or protect cells from photoaging caused by UV. The representative maps of these three optical metrics (data not shown) were morphologically identical to what was shown in Fig. 3, with some differences in hues that represented different levels of readouts. Fig. 5. Generally, UV resulted in increased redox ratio, increased NAD(P)H bound fraction, and increased mitochondrial clustering in keratinocytes cultured using standard media (Table S3). When the media were supplemented with B3, we detected a decrease in the redox ratio ( Fig. 5(a)), and in mitochondrial clustering (Fig. 5(c)). Interestingly, no significant differences were identified in the NAD(P)H bound fraction values (Fig. 5(b)). As the B3-induced metabolic changes trended towards the behavior of the cells without UV irradiation (with complete statistical analysis results among different treatments shown in Table S4), our results suggested that B3 treatment might protect or repair cells from UV-induced damage. It was also noted that the time-dependence of the changes in the optical redox ratio and mitochondrial clustering were distinct. The redox ratio differences were observed in the B3

B3 effects on human dermal fibroblasts
In this study, we use three optical metrics to quantify the metabolic changes of human fibroblasts and keratinocytes, i.e., redox ratio, NAD(P)H bound fraction and mitochondrial clustering. We have established in our previous work the relationship between the combination of these three optical metrics and seven typical metabolic pathways, including glycolysis and glutaminolysis, extrinsic and intrinsic mitochondrial uncoupling, and fatty acid oxidation (saturated and unsaturated substrates) and synthesis [24], which enables us to understand TPEF findings from this study in the context of potential metabolic perturbations. The complementarity of these optical metrics can be used to assess cellular function from the cell to the tissue level with high classification accuracy. We found that the three-metric combination quantitatively yielded the highest original (91.6%) and cross-validated accuracy (91.6%) in classifying these seven metabolic perturbations, in contrast to varying accuracies ranging from 35.5 to 70.9% with only one metric or 66.5 to 87.8% with two metrics [24]. The classification improvement provides concrete evidence of the complementary nature of the metrics, which is also indicated by the differences in the direction of changes in each of the metrics depending on the metabolic pathway perturbation. For example, while both enhanced glycolysis and fatty acid synthesis lead to a decrease in the optical redox ratio and an increase in mitochondrial clustering, the NAD(P)H bound fraction decreases in the former and increases in the latter case. Glycolysis occurs in the cytosol and leads to enhanced levels of NADH in that cell compartment, where NADH is free, while fatty acid synthesis relies on the export of citrate from mitochondria, whose enhanced production leads to an accumulation of NADH in the mitochondria, where it is primarily found in bound form. However, the relationship among potential changes in these metrics is rather nuanced. For example, the redox ratio depends on the intensity of NAD(P)H TPEF signal. A small change in the fraction of NAD(P)H that is bound in mitochondria (i.e. NAD(P)H bound fraction) will lead to a much more significant change in the redox ratio, because the efficiency of TPEF intensity of bound NAD(P)H is approximately ten fold that of unbound NAD(P)H [43]. On the other hand, there are factors, such as pH changes or changes in the free/bound NADH vs. NADPH pools (likely to occur as a result of changes in the balance of glycolysis and oxidative phosphorylation vs. oxidative stress and anti-oxidant responses, respectively), that are likely to be reflected in much more significant differences in lifetime than in intensity [16,45]. Finally, we have found that enhanced oxidative phosphorylation upon re-instatement of normoxia (after hypoxia induction) led to a restoration in the levels of NAD(P)H TPEF intensity and mitochondrial networks to pre-hypoxia levels, but the latter changes occur later than the former [27].
It is known that B3 has anti-oxidative benefits [46] and is an earlier precursor of NAD + in the salvage pathways of NAD + biosynthesis [47,48]. In agreement, in our experiments, treatment with 10 mM B3 results in significant metabolic effects when compared to control/untreated keratinocytes and fibroblasts. Specifically, the induced redox ratio changes suggest that the oxidative stress is lowered and reducing power is increased for the cells treated with B3, in agreement with the known anti-oxidative benefits of B3 [30,46,[49][50][51]. The NAD(P)H bound fraction assessments suggest that the cells treated with B3 maintain the utilization of NAD(P)H in enzymatic reactions throughout the 3-week treatment. Given the fact that most bound NAD(P)H is located at the mitochondrial membrane and matrix, the results indicate that B3 may keep up the energy production through oxidative phosphorylation. In addition, the lower mitochondrial clustering levels indicate that cells treated with B3 have more highly-networked mitochondria, suggesting more efficient energy production and delivery [52,53]. All these observed properties indicate the maintenance or even enhancement of respiratory mitochondrial function and containment of oxidative damage.
In addition to the significant antioxidant effects indicated by the decreased redox ratio, B3 seems to enhance oxidative enzymatic metabolism as well. In the control fibroblast group (Fig. 2(b)), the bound NAD(P)H fraction increases during the early proliferating periods (week 1 of measurements), and then it starts to decrease gradually, suggesting a decrease in the shuffling of reducing power by oxidative phosphorylation. Consistently, the mitochondrial clustering increases and then plateaus. This suggests that mitochondrial function may deteriorate with time, likely due to the oxidative stress and damage to mitochondria [54]. During the process of cell aging, it is known that oxidative phosphorylation and mitochondrial function decline [55,56] and the energy production mostly shifts to anaerobic glycolysis [3]. Under treatment with 10 mM B3, the cells maintain higher NAD(P)H bound fraction levels and lower mitochondrial clustering (i.e. more networked mitochondria), suggesting improved maintenance of mitochondrial function and efficiency in energy production. This is consistent with studies reporting that the lack of mitochondrial DNA exchange through mitochondrial fusion in fusion-deficient cells may greatly reduce respiratory capacity and the functional mitochondrial mass [57]. Measuring the activity of mitochondrial complex I and α-ketoglutarate dehydrogenase, Jia et al. showed that B3 could improve mitochondrial function in a neuroblastoma cell model [49]. Moreover, Regmi et al. used C. elegans to study mitochondrial morphology and suggested that mitochondrial fragmentation (i.e., clustering) was a consequence of the aging process [54]. Mellem et al. also showed that keratinocytes in young skin had less clusters than those in old skin [19], suggesting that mitochondrial clustering may be a phenotypical change that was age-dependent.

B3 effects on keratinocytes
Compared to the fibroblast studies, keratinocytes seem to be more responsive to the given B3 treatment than fibroblasts, since significant metabolic effects are detected as early as 3 days of treatment in keratinocytes, while typically after 8 days in fibroblasts. This might be due to the fact that the keratinocytes are perturbed by metabolic stimuli mostly in the proliferative state during the experiments. Cells are seeded at low confluency ∼15-20% in both fibroblasts and keratinocytes experiments. However, fibroblasts typically quickly reach high confluency (around 90%) in less than 5 days (Fig. S3), while keratinocytes reach less than 90% confluency after the 7-day course of an experiment (Figs. S4 and S5), indicating that the keratinocytes are maintaining a proliferative state within these 7 days. Such differences, partly explain a decreasing trend in redox ratio observed within keratinocytes as a function of time, since cells at a proliferative state prefer glycolysis for ATP production [58]. Unfortunately, because keratinocytes tend to lift off the culture dish once they reach confluency, it is not really possible to design an experiment that yields results that are not impacted by the overall different proliferation rates of these two cell types. Cells at proliferative state are sensitive to metabolic perturbations possibly because extensive metabolic rewiring occurs at this stage in an organized way in order for cells to acquire sufficient nutrients to support cell growth, in contrast to cells at non-proliferative state [59].
Consistent with the optical metabolic effects of B3 in fibroblasts, B3 supplementation in keratinocyte media also leads to a decrease in the redox ratio and mitochondrial clustering, along with an increase in NAD(P)H bound fraction. In agreement with the lower redox ratio, Rovito et al. suggested that B3 may protect cellular metabolism from oxidative stress by preferentially affecting glycolysis [30], and others suggested that B3 could be an antioxidant to reduce oxidative mitochondrial damages [46,[49][50][51]. The main source of reactive oxygen species (ROS) causing oxidative stress is produced from the enzymatic complexes at mitochondrial membrane that are involved in oxidative phosphorylation. If the cells switch from oxidative phosphorylation to a more glycolytic state, the oxidative stress may decrease due to the decrease of mitochondrial respiration. However, the combination of the lower redox ratio, increased NAD(P)H bound fraction and lower mitochondrial clustering is unique and possibly suggests that B3 might improve energy production and efficacy by enhancing both glycolysis and the efficiency of mitochondrial respiration.

B3 effect on keratinocytes with UV irradiation
When comparing the metabolic readouts from keratinocytes with and without UV irradiation, we note that UV leads to increased redox ratio, consistent with an increased oxidative stress reported in the literature [31,[60][61][62]. For example, elevated ROS levels were directly evident from keratinocyte cultures subject to UV irradiation, as represented by measured levels of two highly reactive free radical species, superoxide and nitric oxide [60]. ROS levels were also measured in skin organ cultures exposed to UV irradiation, and increased ROS production and decreased antioxidant cell capacity were found as a result of UV-induced photodamage [61]. In addition, Park et al. showed that UV resulted in glycolytic blockage [31], while Brace et al. indicated that increased oxidative phosphorylation was a cellular mechanism in response to the DNA damage caused by exogenous stresses, such as UV irradiation [62]. The observed increase of NAD(P)H bound fraction resulting from UV irradiation may indicate the increased energy demands in cells for fighting against photo-damage and cell repairs. In addition, UV exposure leads to an elevated level of mitochondrial clustering, which is probably associated with stress responses and apoptosis of the cell and may suggest the deterioration in mitochondrial function [63]. As more damage occurs under UV irradiation, the lifespan of keratinocytes may be shortened. Paz et al. evaluated mitochondrial function of keratinocytes after UV irradiation and concluded that mitochondrial membrane depolarized after UV irradiation, which led to apoptotic cell death [64].
Consistent with the previous experiments, 10 mM B3 leads to a lower redox ratio, thus slightly lowering oxidative stress when keratinocytes are irradiated by UV. These results are in agreement with those of Zhen et al., who investigated the protective effects of B3 on particulate matter-induced oxidative stress in human HaCaT keratinocytes, and detected a decreased ROS level via B3 supplementation [65]. NAD(P)H bound fraction is not significantly changed in this case, which might be a result of the balance between enhanced glycolysis and enhanced mitochondrial respiration in response to enhanced oxidative stress as mentioned above, with NAD(P)H bound fraction decreasing in the former while increasing in the latter case. Compared with UV-free keratinocytes where an increased level of NAD(P)H bound fraction is observed in response to B3 supplementation, these UV-irradiated keratinocytes are reported to upregulate glycolysis against oxidative stress [30,31], thus potentially leading to an unchanged bound fraction level, although future experiments are needed to validate this possibility. B3 also significantly lowers the mitochondrial clustering of UV-irradiated cells, indicating that it enables improved mitochondrial function and more efficient energy production.

Conclusion
In this study, we use label-free, non-invasive TPEF imaging to assess cellular metabolism of human fibroblasts and keratinocytes in response to B3 supplementation, relying completely on endogenous contrast. Specifically, we use a multi-parametric analysis approach relying on three optical metrics, including the optical redox ratio, NAD(P)H bound fraction and mitochondrial clustering, which provide complementary insights into the underlying possible metabolic pathway changes. Overall, we find that B3 is able to aid cell energy production, to improve mitochondrial function [32,49], and to prevent stress-induced energy loss in cells [30,31]. All data support the important role of B3 as an ingredient that can boost cell energy, provide antioxidant benefits, and potentially extend the lifespan of cells [66]. In addition, B3 effects on antiaging or anti-photoaging, as implied from previous studies [67,68], are confirmed and supported by the experimental results. Further applications of such optical techniques may include but are not limited to characterizing the biological effects of actives, revealing modes of action, and utilization in a clinical environment for treatment evaluations. Continual advancement of such measurements will accelerate the understanding of cell energy dynamics while investigating other beneficial ingredients for skin care benefits.