The Effect of Short-Wavelength White LED Illumination throughout the Night on the Milk Fatty Acid Profile of High-Yielding Dairy Cows

Simple Summary Exposure to artificial white light at night (LAN) disrupts circadian rhythms, yet many dairy cows continue to be exposed to LAN for historical and practical reasons. We explored the effect of whole-night illumination, using white light-emitting diode (LED) bulbs, on the production and composition of milk collected during the day (day milk) and night (night milk). Compared to a natural light–dark cycle, white LED reduced the differences in milk composition between milking hours, yet did not affect milk production. Specifically, white LED increased the percentage of saturated fatty acids in day milk at the expense of unsaturated fatty acids. Therefore, we conclude that natural light is preferable to white LED for cows’ well-being and for preserving the natural milk fat profile. Maintaining the natural milk fat profile results in obtaining day milk fat with higher levels of unsaturated fatty acids, mainly oleic acid, which are known for their health-promoting characteristics. In conclusion, reducing the exposure of dairy cows to white LED during the night will reduce energy use in dairy farms, save electricity costs, and yield day milk with a healthier fat composition, and possibly a higher health-promoting market value. The welfare conditions of the cows will also improve. Abstract Fatty acid levels in milk vary between day and night milking. Many dairy cows are still kept under white light-emitting diode (W-LED) illumination throughout the night, although it is known to disrupt endogenous circadian rhythms. We investigated the effects of whole-night W-LED illumination (125 lux) on milk yield and circadian composition, compared to a natural light–dark (LD) cycle of 10 h light. Mid–late lactation cows (n = 34) that were exposed to natural LD cycle showed circadian variation in milk fat composition, characterized by higher health-promoting monounsaturated fatty acid (MUFA; 24.2 ± 0.4 vs. 23.2 ± 0.4 g/100 g fat, p < 0.001) and lower saturated fatty acid levels (71.2 ± 0.4 vs. 72.5 ± 0.4, p < 0.001) at 13:30 h (day milk) than at 03:30 h (night milk). Compared to natural LD (n = 16), W-LED (n = 18) did not affect milk production or milk fat yields, yet abolished the milking time variation in milk fat composition towards a less healthy fatty acid profile. This lowered MUFA levels of day milk (23.8 ± 0.4 vs. 26.7 ± 0.4, p < 0.01). Therefore, W-LED has no commercial advantage over the tested natural LD cycle, and conversely, even shows circadian disruption. Accordingly, a natural LD cycle of 10 h light is preferable over W-LED from the perspective of cost savings, the cows’ well-being, and preserving the natural milk fat profile, as the nutritional value of the day milk is slightly higher.

. Illumination conditions and selected parameters collected at the end of the experimental phases. The data are presented as mean ± SD.  Cow feed (Table A1) was provided ad lib and dispensed daily at 10:00 (or 1 h earlier during daylight savings time, which ended on 27 October 2019, the 30th day of the experiment). During the pre-experimental phase, the yard was illuminated throughout the night using fluorescent bulbs ( Figure 1 and Table 1). The cows were regularly milked at 05:00, 13:00, and 20:00 h standard time throughout the pre-experimental phase and all the following experimental phases. During the ca. 0.5 h milking, cows were exposed to white fluorescent illumination (dominant wavelength of 545 nm at 140 lux, measured at cows' eye height using AvaSpec-2048-FTSDU, Avantes, Eerbeek, Netherlands). As justified above, we examined the response to illumination manipulations in mid-late lactation cows [33][34][35]. Accordingly, the overall length of the experiment was limited to about 60 days, thus avoiding sampling of cows reaching the dry period ( Figure 1). We sought to examine the DM vs. NM variation in fatty acid composition under natural illumination, and the effects of W-LED LAN on this variation and on cows' milk yield. To this end, we first transferred all the cows to a natural light-dark (LD) cycle regime of ca. 11 h light and ca. 13 h dark ( Figure 1 and Table 1). The control group was held under this naturally changing LD cycle throughout the experiment, during which the light hours were shortened by ca. 1 h to ca. 10:14 LD cycle ( Figure 1 and Table 1). Following 30 days under the natural LD cycle, the illumination conditions of the treatment group were changed to a W-LED illumination regime (dominant wavelength of 462 nm, day light 6500 k, 34 w, Soul LTD, 124.5 Lux). W-LED lights were only turned on throughout the night, resulting in a 24 h illumination regime (24:0 LD cycle, Figure 1 and Table 1).

Sampling of the Milk Fat and Analysis of Its Fatty Acid Profile
Two milk samples were collected 10 h apart, 1-3 days before the end of each of the two experimental phases (Figure 1). One milk sample was taken during the regular milking time of 13:30 (i.e., DM) and another during a novel milking time of 03:30 (i.e., NM). The latter time was set to enable sampling of the milk formed in the mammary gland during the night hours, before sunrise. At each milking event, one 50 mL milk tube from each cow was sent to an external ISO-certified laboratory (Israeli Cattle Breeders' Association, Caesarea, Israel). There, total fat was analyzed using FTIR (Foss MilkoScan™, Hilleroed, Denmark). Additionally, two 15 mL tubes were sub-sampled and frozen at −20 °C for the fatty acid profile analysis.
Samples were prepared as previously described [36]: First, 0.5 mL of each milk sam-retreatment luorescent aseline Natural light hite L light il sampling il sampling Natural light ay ay ay

Sampling of the Milk Fat and Analysis of Its Fatty Acid Profile
Two milk samples were collected 10 h apart, 1-3 days before the end of each of the two experimental phases (Figure 1). One milk sample was taken during the regular milking time of 13:30 (i.e., DM) and another during a novel milking time of 03:30 (i.e., NM). The latter time was set to enable sampling of the milk formed in the mammary gland during the night hours, before sunrise. At each milking event, one 50 mL milk tube from each cow was sent to an external ISO-certified laboratory (Israeli Cattle Breeders' Association, Caesarea, Israel). There, total fat was analyzed using FTIR (Foss MilkoScan™, Hilleroed, Denmark). Additionally, two 15 mL tubes were sub-sampled and frozen at −20 • C for the fatty acid profile analysis.
Samples were prepared as previously described [36]: First, 0.5 mL of each milk sample was pipetted into a 15 mL tube, followed by 10 mL of hexane (analytical grade, J.T. Baker) spiked with 50 µg/mL of benzophenone (Cat #: 427551, Sigma, Darmstadt, Germany) as an internal standard, and 50 µg/mL tritridecanoate (T3882, Sigma, Darmstadt, Germany) as the transesterification standard. Then, 1 mL of a saturated solution of sodium methoxide (TCI Co., Ltd.) in methanol (HPLC grade, J.T. Baker) was added. The sealed tubes were placed in a rack and vigorously shaken for 20 min while lying horizontally on an orbital shaker. An aliquot of the upper organic layer was filtered through a syringe filter (PTFE filter matrix, 13 mm , 0.45 µm pore size) directly into a glass amber vile for gas chromatography (GC) analysis. External calibration standards were prepared. First, a standard mix containing 37 fatty acid methyl esters (cat #: 18919, Sigma, Table A2) was dissolved in hexane to a 20 mg/mL concentration and stored in a capped vile as a stock solution at −20 • C. An aliquot from the stock was used to prepare standard dilutions of 4, 3.2, 2.4, 1.6, 0.8, 0.4, and 0.1 mg/mL in hexane. Additionally, a standard of conjugated linoleic acid (CLA, C18:2) and methyl esters (cat #: O5632, Sigma, Darmstadt, Germany) was dissolved in pentane to a 50 mg/mL concentration and stored in a capped vile as a stock solution at −20 • C. An aliquot from this stock was used to prepare standard dilutions of 1, 5, 10, 15, 20, and 25 µg/mL in hexane. Each calibration solution, the standard mix, and the CLA standard contained 50 µg/mL of benzophenone as the internal standard.
The fatty acid profile was analyzed on a GC instrument (7890A, Agilent Tech. Santa Clara, CA, USA) coupled with a mass spectrometer (5975C, Agilent Tech. Santa Clara, CA, USA). Separation was performed on a polar column (30 m length, 0.25 mm internal diameter, Zebron ZB-FAME, Phenominex). The conditions were as follows: First, 1 µL of the sample or calibration standard was injected into the GC inlet, heated to 250 • C, and set to a split ratio of 10:1. The oven program was set to 100 • C for 2 min, then a 10 • C/min increase to 140 • C, then a 3 • C/min increase to 180 • C, ending with a 30 • C/min increase to 260 • C, with another 2 min holding time (the total runtime was 24 min). Fatty acids were quantified as methyl esters using individual calibration curves, as previously described [37]. A calibration curve was plotted with values calculated for each fatty acid: x i = C i /C is , y es = A i /A is . Here, X i is the calibrant's value for a given FAME (plotted on the x-axis); C i is the calibrant concentration for a given FAME as mg/mL; and C is is the standard internal concentration in the calibrant as mg/mL.

Statistical Analysis
Kolmogorov-Smirnov normality tests were applied to assess the normality of the data. As the data were not normally distributed, we used the Mann-Whitney non-parametric test to evaluate differences between the control and treatment groups at each experimental period, and the Wilcoxon non-parametric test to evaluate differences between milking times within each experimental group using SPSS, version 27 (IBM). The significance level was set at p < 0.05.
Body weight gradually increased throughout the experiment, and milk yield gradually decreased ( Figure 3 and Table 1). As a baseline, after another three weeks under natural light conditions, the milking hour still affected most (11 of the 15) quantified fatty acid levels ( Figure 2 and Table A3). As was found for baseline milk, SFA levels were lower at DM than NM (69.8 ± 0.8 vs. 71.8 ± 0.8 g/100 g milk fat, p = 0.003), while MUFA levels were higher (26.7 ± 0.9 vs. 24.3 ± 0.7 g/100 g milk fat, p = 0.003). These differences were due to the same fatty acid that accounted for the corresponding differences measured at baseline ( Figure 2 and Table A3). In all, these milking-hour-related differences in milk fat composition, found under natural light conditions of ca. 10-11 illumination hours, are robust. The differences were not eliminated by the combination of a 1 h reduction in the hours of natural light, increased days-at-milking, or decreased milk yield ( Figure 2 and Table A3). were higher (26.7 ± 0.9 vs. 24.3 ± 0.7 g/100 g milk fat, p = 0.003). These differences were due to the same fatty acid that accounted for the corresponding differences measured at baseline ( Figure 2 and Table A3). In all, these milking-hour-related differences in milk fat composition, found under natural light conditions of ca. 10-11 illumination hours, are robust. The differences were not eliminated by the combination of a 1 h reduction in the hours of natural light, increased days-at-milking, or decreased milk yield ( Figure 2 and Table A3). , capric (f), lauric oleic (g) and conjugated linoleic (i) acid. The data are presented as mean ± standard error. Control, n = 15; treatment, n = 18. SFA, saturated fatty acids; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; CLA, conjugated linoleic acid.*, p < 0.05; #, p < 0.01; $, p < 0.001, between milking hours, by Wilcoxon non-parametric test. a, p < 0.05; b, p < 0.01; c, p < 0.001, between treatments within milking hours, by Mann-Whitney non-parametric test. , capric (f), lauric (g) oleic (h) and conjugated linoleic (i) acid. The data are presented as mean ± standard error. Control, n = 15; treatment, n = 18. SFA, saturated fatty acids; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; CLA, conjugated linoleic acid. *, p < 0.05; #, p < 0.01; $, p < 0.001, between milking hours, by Wilcoxon non-parametric test. a, p < 0.05; b, p < 0.01; c, p < 0.001, between treatments within milking hours, by Mann-Whitney non-parametric test.

Whole-Night White LED Illumination Results in a Higher Level of SFA at DM and Abolishes the between-Milking-Time Variation in Milk Fat Composition Found under the Natural Light Regime
White LED illumination did not affect the cows' body weight, milk yield, or milk fat levels at each milking hour (Table 1 and Figure 2a,h). W-LED illumination affected the milk fat composition, mainly in DM. This abolished the differences between DM and NM level for eight of the nine fatty acids, and the differences in SFA (72 ± 0.5 and 72 ± 0.8 g/100 g milk fat, respectively, p = 0.206) and MUFA (23.8 ± 0.4 and 23.0 ± 0.4 g/100 g milk fat, respectively, p = 0.122) (Figure 2 and Table A3). Differences were still observed between DM and NM in control cows: SFA (69.8 ± 0.8 and 71.8 ± 0.8 g/100 g milk fat, respectively, p = 0.003) and MUFA (26.7 ± 0.9 and 24.3 ± 0.7 g/100 g milk fat, respectively, p = 0.003; Figure 2 and Table A3). This lack of a time-related difference in SFA levels under W-LED was due to the abolishment of time-related differences in butyric (C4:0), caproic (C6:0), caprylic (C8:0), capric (C10:0), lauric (C12:0), margaric (C17:0), and stearic (C18:0) acid levels, which were still present in the control cows, due to their higher levels at DM under W-LED compared to natural light ( Figure 2 and Table A3). The complementary effect was found in MUFA levels, which were now lower in the DM of the W-LED treated than in the control cows ( Figure 2c and Table A3). This W-LEDinduced lower MUFA level at DM was mainly due to a lower oleic acid (C18:1) level than in the control cows (20.9 ± 0.4 vs. 24.0 ± 1.0 g/100 g milk fat, p < 0.01); (Figure 2h and Table A3).

Discussion
In cows and other organisms, many behavioral and physiological outputs show a circadian rhythm that is disrupted by exposure to artificial illumination at otherwise dark hours (i.e., LAN) [11,17]. Our present study shows that exposing high-yield dairy cows to a natural LD cycle of ca. 10-11 h of natural light is accompanied by a circadian variation in milk fat composition. This circadian variation is reflected in higher levels of the healthpromoting UFA in DM (milked at 13:30) than in NM (milked at 03:30). This natural betweenmilking-time variation was abolished under whole-night short-wavelength illumination by W-LED, but 24 h milk yield was not affected. These results suggest that a natural LD cycle of ca. 10-11 h of the natural light regime is preferable over W-LED for cows' well-being, DM quality, and reducing unnecessary illumination costs.
In dairy cows, as in other organisms, most physiological parameters show a circadian output affected by photoperiod [15,16]. This has prompted investigations of withinday variations in milk composition, to differentiate between DM and NM from healthpromoting, nutritional, and commercial perspectives [17,[29][30][31][32]38]. As expected, most of these studies found circadian oscillation in milk yield and in some of the milk's constituents, yet the presence and extent of the circadian variation varied between studies and depended on milking and feeding frequency [17,[29][30][31][32]38]. In this study, we mainly focused on the composition of fatty acids, due to their high nutritional value [39]. In cows exposed to natural light conditions, we found 1-2 percentile point higher UFA levels and lower complimentary SFA levels in DM than in NM. This lower level of SFA in DM was due to a difference in short-to mid-chain SFA (C8 to C14). In contrast, the higher level of UFA in DM was due to different levels of all the detected UFA, as found by others [32], mainly oleic acid levels (e.g., a 2.9 percentile point difference at the second time point). These circadian rhythms in milk fat level and composition are suggested to be controlled by the feeding patterns and mammary circadian clock, which synchronizes to the central circadian clock; the latter synchronizes to the photoperiod [16,40]. In all, under natural light conditions of ca. 10-11 illumination hours, our results show a modestly higher nutritional value of DM milk fat, which was richer in UFA compared to NM milk fat. This difference was not affected by the combination of a 1 h reduction in the hours of natural light, increased daysat-milking, or decreased milk yield. Yet, due to the scant nutritional difference between DM and NM, the economic and commercial feasibility is doubtful, regarding the separation of DM from NM milk during milking, and the production of 'premium milk' characterized by health-promoting nutritional values.
Our demonstration of a between-milking-time variation in milk fat composition set the ground for exploring our main aim. Specifically, we investigated the effect of whole-night illumination, using W-LED, on milk yield and milk fat composition, compared to a natural illumination regime of about 10 h of natural light. The result of this comparison is of scientific importance for several reasons. For example, although illumination throughout the night does not increase milk production efficiency beyond that achieved under a long-day regime [8,9], some dairy farms still practice it, and, more recently, even use short-wavelength LED illumination of various spectrums and intensities. This usage of LED lighting highlights another important aspect of this study: examining the effect of W-LED LAN per se, which is dominated by short wavelengths, on productivity and the existence of daily endogenic rhythms. Illumination with W-LED showed that whole-night illumination using W-LED bulbs did not affect cows' body weight, milk yield, or milk fat content; all resembled those of control cows, which continued to be exposed to the natural LD cycle. This corroborates another study that compared outcomes of W-LED LAN to those of a natural photoperiod of 13 h illumination [9]. Therefore, as W-LED LAN, which requires a high energy investment in illumination overnight, did not yield a commercial advantage, we cannot cite any advantage in its usage. It should be noted that we tested the effect of W-LED LAN vs. natural light during the fall. Hence, the study should be repeated under high summer temperatures at which W-LED LAN-exposed cows may shift their feeding hours to the cooler, now-illuminated, "night" hours, hence inducing a more positive energy balance resulting in a higher milk yield. This proposed experiment is also needed as the seminal work showing that a 24 h light regime does not affect milk yield and composition in Holstein cows, compared to an 18L:6D, was conducted during the cold winter time (December to April) rather than in summer time (1), or under comfortable ambient temperature, which characterizes South Korea during the summer and autumn months (2).
Our lack of finding a robust W-LED-induced disruption in body weight and milk yield is somewhat surprising. This contrasts with LAN disruption observed over a range of physiological parameters and circadian rhythms in human and animal models [11], and cows' circadian rhythm of milk melatonin and heart rate [17]. Hence, we suggest that macro parameters, such as body weight, milk yield, and milk fat level, appear more resistant to a disturbance resulting from W-LED LAN exposure, at least one that lasts for three weeks. This resistance may result from the short, only three-week-long manipulation, or the endless selection of high-yielding cows. The latter favors cows that produce high volumes of milk and milk fat despite their exposure to LAN and its detrimental effects. Nevertheless, the observed 'resistance' to the W-LED illumination (spectrum and intensity) used in our study does not infer sweeping resistance to W-LED. This is because W-LED bulbs differ in the composition of their wavelengths. In fact, the dominant wavelength in our W-LED illumination was 462 nm ('blue'), yet it included other (longer) wavelengths. The usage, however, of W-LED illumination with a higher representation of short wavelength may have resulted in a disruption of such macro parameters as body weight, milk yield, and milk fat, as described in previous work [9].
Despite the above, the lack of a W-LED-induced disruption did not apply to the between-milking-time variation in milk fat composition. Milk and its products are a substantial source of dietary fat in many human populations [41]. However, milk fat comprises a high proportion of SFA [23,24], suggested as a risk factor for cardiovascular diseases (CVD, e.g., atherosclerosis) [42,43]. Nevertheless, reducing SFA consumption by reducing milk consumption may not be the best approach, as milk is rich in essential minerals and amino acids, and milk fat per se is a carrier of fat-soluble vitamins and a source of several essential fatty acids [44]. Both feed-related factors, i.e., dietary intake and seasonal and regional effects, and animal-related factors, i.e., genetics (breed and selection), stage of lactation, mastitis, and ruminal fermentation, are capable of modifying the fatty acid composition, as well as the overall quantity of lipids present in milk [45]. Therefore, manipulating fatty acid composition in milk fat by genetic selection or dietary modification seems beneficial for increasing UFA percentage on account of SFA percentage [28,39].
Another option for obtaining a higher level of UFA in milk fat could be manipulating the photoperiod. Notably, numerous studies searched for an effect of the natural photoperiod, long-day and even whole-night illumination on body weight, milk yield, milk composition, and circadian variation (reviewed above). Moreover, many studies investigated the effect of season-induced change in photoperiod on milk fat composition (e.g., [46,47]). Yet, only a few recent studies investigated circadian variation in milk fat composition and its attenuation by photoperiod [29,32]. Thus, the novelty of this study is our aim of bridging this gap in existing knowledge. Our results show that compared to a natural LD regime, W-LED LAN resulted in higher SFA levels in DM, hence abolishing the between-milking-time variation in SFA and MUFA levels maintained under a natural LD regime in control cows. The elevation in SFA and the circadian disruption were due to elevations in DM levels of butyric (C4:0), caproic (C6:0), caprylic (C8:0), and capric (C10:0) acids compared to control cows. The complementary effect for MUFA levels in DM was mainly due to an impact on the oleic acid (C18:1) level, which was lower in the DM of W-LED-exposed cows than in the control cows. Overall, these results show that W-LED LAN does not affect milk or milk fat yields, but results in a slightly less healthy fatty acid profile, by increasing SFA levels of DM, hence abolishing the between-milking-time variation in milk fat composition. However, as we did not measure the SFA vs. UFA ratio in the milk tank, we cannot evaluate the effect of W-LAN on this parameter. This limitation of the study's results should be addressed in future studies.
The biochemical mechanisms by which photoperiod, specifically W-LED LAN, affects the milk fat composition of DM are yet to be understood. Notably, the main effect on DM SFA levels was on the less abundant short-to mid-chain (≤C12) fatty acid levels, and not on the most abundant palmitic acid (C-16). This suggests several implications. In contrast to palmitic acid, short-to mid-chain fatty acids are mainly synthesized de novo within the mammary gland rather than partly transferred from the blood to the mammary gland [24]. Therefore, the observed LAN-induced change in DM fatty acid composition may be due to a LAN-induced effect on the within-mammary-gland synthesis of short-to mid-chain fatty acids. Mammary gland cells obtain an endogenous circadian clock and show a circadian rhythm in about 7% of their transcriptome, including core-clock-related and metabolic genes [16]. As these genes are affected by changes in photoperiod [16], we suggest that the LAN-induced changes in DM SFA levels are mainly due to a LAN-induced mammary gland shift in the clock-controlled de novo anabolism of short-to mid-chain fatty acids. Oleic acid originates mainly from feed [24] rather than the mammary gland. Accordingly, the source of the W-LED LAN-induced difference in DM oleic acid level seems external to the mammary gland. Regardless of the source of the changes in milk fatty acid levels, they are probably mediated by the already shown photoperiod-induced changes in feeding patterns [5] that were shown to affect milk levels of both oleic acid and short-to mid-chain fatty acids [40]. The described changes in milk fatty acids are also probably mediated by photoperiod-induced changes in blood melatonin levels, to be measured in future studies, as we already showed that milk melatonin is attenuated by short-wavelength white fluorescent LAN [17].

Conclusions
We showed no commercial advantage in W-LED LAN compared to the natural LD cycle during the autumn months, at 32 • north latitude, consisting of ca. 10 to 11 h light exposure period. However, we showed a disadvantage, namely, a circadian disruption that yielded slightly higher levels in DM, of SFA; increased dietary consumption of the latter has been suggested as a risk factor for metabolic diseases. Hence, our findings support using the natural LD cycle of ca. 10 to 11 h of light over W-LED. This is from the perspective of cost savings, cows' well-being, and preserving the natural circadian variation in milk fat composition, due to DM's slightly higher nutritional value.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.
Acknowledgments: We express our appreciation to the Volcani Center (Beit Dagan, Israel) dairy farm team and its manager Shamai Yaacob; Raviv Harris, the head of the analytic chemistry unit at MIGAL; and Dalia Niv and Larisa Yavel from the Laboratory of Integrative Physiology at MIGAL.

Fatty Acid Standard Fatty Acid Group Status
Butyric  Table A3. The proportions of total fat in whole milk (g/100 mL of milk) and of fatty acids in milk fat (g/100 g of milk fat), according to milking time and experimental phase (i.e., illumination condition).  The data are presented as mean ± SE. The Wilcoxon non-parametric test was used to assess the effect of milking time in each group. The Mann-Whitney non-parametric test was used to assess differences between the control and treatment groups, for the milking time at each experimental period. *, p < 0.05; #, p < 0.01, $, p < 0.001 between treatments within milking hours, by the Mann-Whitney non-parametric test.