Assessment of Metabolic Adaptations in Periparturient Dairy Cows Provided 3-Nitrooxypropanol and Varying Concentrate Proportions by Using the GreenFeed System for Indirect Calorimetry, Biochemical Blood Parameters and Ultrasonography of Adipose Tissues

Methanogenesis in ruminants contributes to both greenhouse gas emissions and feed energy losses whereby the latter becomes specifically important in energy-deficient periparturient cows. It was hypothesized that increased concentrate feed proportions (CFP) and feeding with the methane inhibitor 3-nitrooxypropanol (3-NOP), as well as their potential synergism, improve the energy status of peripartal cows. Periparturient dairy cows were fed low or high dietary CFP either tested without or combined with 3-NOP. The GreenFeed system was used to calculate the metabolic respiration quotient (RQmetabolic) and tissue energy retention (ERtissue) by methods of indirect calorimetry. The calorimetrically estimated ERtissue coincided with a conventionally calculated energy balance except for the antepartal period. Neither CFP nor 3-NOP affected the ultrasonographically assessed lipomobilization in adipose depots. In the group fed 3-NOP and a high concentrate feed proportion, the RQmetabolic significantly rose over the course of the experiment and the ERtissue was also increased. Serum non-esterified fatty acid concentrations were lower in the 3-NOP groups albeit ß-hydroxybutyrate (BHB) remained unaffected. Higher CFP reduced BHB and increased blood glucose levels. In conclusion, 3-NOP and high CFP improved the energy budget of the cows in an interactive manner, which was, however, not apparent in all of the examined parameters. The application of the GreenFeed system for indirect calorimetry is a promising approach, which needs further validation in the future.


Introduction
In ruminants, feed is mainly converted to volatile fatty acids (VFA) by the rumen microbiota, thereby yielding hydrogen (H 2 ) and carbon-dioxide (CO 2 ), which are redirected to methane (CH 4 ) formation in methanogenic archaea [1]. 3-nitrooxypropanol (3-NOP), a structural analogue of methyl-coenzyme M, is currently supposed to be one of the most potent CH 4 inhibitors in cattle [2,3]. The CH 4 -mitigating effect of 3-NOP potentially amounts to 39.0 ± 5.40% in dairy cows [4] but ranges widely from 7 [5] to 60% [6] depending on the provided ration type (neutral-detergent fibre (NDF) content), administration technique (mixing in with the total-mixed ration (TMR), infusion, pulse-dose) and dosage level [4]. Table 1. Chemical composition, peNDF and energy (means) of the total rations offered during the experimental period from d 28 ante-partum until d 120 post-partum (reproduced from and with permission from Schilde et al. [15] at Taylor & Francis Group https://www.tandfonline.com/) © 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group (https: //www.tandfonline.com/) under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (https://creativecommons.org/licenses/by-nc-nd/4.0/), which permits noncommercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.    11.0 11.5 11.0 11.5 NE L 6.6 7.0 6.6 7.1 † Control (CON) groups were provided a placebo and low (LC) or high (HC) concentrate feed proportion in the ration. §   aNDF om , α-amylase treated neutral detergent fibre without residual ash. # peNDF >8 mm , physically effective NDF in the partial mixed ration defined as the proportion of DM retained by a 8-mm screen multiplied by the dietary NDF content [28]. $ Calculations for concentrates based on table values according to DLG [29], silages according to VDLUFA [27] analyses and GE calculated according to GfE [30]. Gas samples and background gases were analysed by the already installed GF sensors. CH 4 and CO 2 concentrations were analysed by non-dispersive infrared absorption sensors and O 2 was analysed using a paramagnetic sensor. Sensor calibration was performed automatically on a daily basis using a zero (O 2 = 200,000 mg/kg, N 2 = 800,000 mg/kg) and a span gas (CH 4 = 1004 mg/kg, CO 2 = 10,000 mg/kg, O 2 = 210,000 mg/kg H 2 = 9.50 mg/kg, H 2 S = 9.80 mg/kg, while the remainder of the gas was nitrogen). The air velocity in the pipe was measured by an anemometer to determine total mass flow of all gases. CO 2 recovery tests were conducted once a month (recovery rate ± SD: 101% ± 5.7). The amount of bait feed delivered per feed drop was calibrated on a weekly basis.

Energy Turnover Estimated by Indirect Calorimetric and
Serum samples were photometrically analysed (Indiko Plus, Thermo Scientific GmbH, Dreieich, Germany) for concentrations of BHB, NEFA, triacylglycerides (TAG) and glucose. An automated blood gas and electrolyte analyser (GEM Premier 4000, Werfen GmbH, Kirchheim, Germany) was used to determine the temperature-corrected pH, hydrogen carbonate ions, haemoglobin and lactate concentrations immediately after sample collection.

Calculation of Energy Metabolism Parameters by Indirect Calorimetry and Ultrasonography
Processing and validation of gas exchange data were conducted by C-Lock Inc. Gas measurements were converted from g/d to L/d according to the gas density of 0.717 kg/m 3 for CH 4 , 1.977 kg/m 3 for CO 2 and 1.729 g/m 3 for O 2 under standard conditions (1013.25 hPa). The cow s visiting time and head position in the GF were used to check for data plausibility [31]. Daily means of GF data were averaged to weekly means using the previously described arithmetic averaging method [32]. Due to technical reasons, O 2 consumption in CON groups was estimated from weekly means of CO 2 production and DMI using the following regression equation: O 2 (g/d) = 2056 − 72.5 × dry matter intake (kg/d) + 0.62 × CO 2 (g/d) (1) with R 2 = 0.90 and a residual standard error (RSE) of 371 g/d on 337 degrees of freedom. In ruminants, total CO 2 production (VCO 2 ) is the sum of fermentative (VCO 2fermentative ) and metabolic (VCO 2metabolic ) CO 2 derived from microbial fermentation in the rumen and the intermediary metabolism, respectively [33]. A differentiation between the two is essential in order to refer to the intermediary substrate oxidation [34]. As proposed by Chwalibog et al. [33], VCO 2fermentative was calculated by applying the stoichiometrically derived factor of 1.7, which was confirmed to be applicable for a variety of ration compositions [35].
Then, VCO 2fermentative was subtracted from VCO 2 to obtain VCO 2metabolic , which was used to calculate the RQ metabolic mirroring the intermediary oxidation of the macronutrients of carbohydrates, fat and protein [36]: The total RQ (RQ total = total VCO 2 production (L/d) ÷ VO 2 consumption (L/d)) reflected the cow s nutritional plane. Gross energy (GE) content of the feedstuffs was calculated according to GfE [24]. The metabolizable energy (ME) content of the concentrates was derived from tabular values according to DLG [29] and that of silages was derived according to VDLUFA [27] analyses.
HP was quantified according to the Brouwer [22] formula: whereby urinary nitrogen excretion (N U ) was set to 50 g/d [37] even though the real N U in dairy cows varies between 75 and 150 g/d [38]. However, the N U contribution to HP is negligible and an error of about 0.3% in the absolute HP values was accepted [39].
Methane energy (CH 4 E; MJ/d) was derived from the multiplication of the energy equivalent value of 39.54 kJ/L of CH 4 [22] and the daily CH 4 production (L/d).
The partitioning of EE for energy retention (ER) was computed as follows: ER in body tissues and milk (ER total ) (MJ/d) = ME intake − HP ER in body tissues (ER tissue ) (MJ/d) = ME intake − HP − ME E − NE P (6) Analyses and calculations of milk energy excretion (ME E ; MJ/d) according to GfE [24] were used from Schilde et al. [15]. Net energy demand for pregnancy (NE P ; MJ/d) was averaged for period 1 according to constants proposed by GfE [24] with 13 MJ NE L /d for week 4 a.p. and 18 MJ NE L /d for week 3 a.p. until calving resulting in an average of 17 MJ NE L /d for period 1. The EB GFE data were extracted from Schilde et al. [15] in which EB GFE was calculated according to GfE [24].
The residual ER in body protein and intramuscular fat was assessed as: ER fat depot was calculated from UM as described in the following: The daily mobilization of fat depot masses from each AT depot was described by the difference in AT masses between d 3 p.p. and d 28 p.p. divided by the number of days. The energy release from mobilized fat depots being used for milk production was calculated based on the assumption that 1 g of body fat corresponds to 39.8 kJ of GE [22], whereby 16% is lost as heat energy when body tissue energy is used for milk synthesis [40].
Depot masses of each AT, namely the retroperitoneal (RAT), mesenteric (MAT) and omental (OAT), collectively referred as the visceral AT (VAT), and the subcutaneous AT (SAT) were estimated in kg from ultrasonographically measured distances of the different sites as described in Schäfers et al. [26] The efficiency of utilization of ME for lactation (k l ) was calculated according to AFRC [41]: where ME E is adjusted to zero energy balance with a coefficient of a = 0.84 for negative ER tissue or a = 1/0.95 for positive ER tissue . ER tissue is the energy balance obtained by indirect calorimetry using the GreenFeed system. The maintenance requirement (ME m ) was estimated using the equation from GfE [24]: where BW 0.75 is the metabolic body weight.

Statistical Analyses
Prior to statistical evaluation, means were calculated per cow and week for variables used in indirect calorimetry. A.p. blood samples were retrospectively assigned to the actual day relative to parturition by tolerating a deviation of 24 h for the d −3 sample and a deviation of 2 days for the d −7 and d −14 samples. Due to gas leakage through the fistula, cannulated cows were excluded from statistics except for blood and ultrasonic variables. The statistical analyses were conducted using the SAS software package (version 9.4; SAS Institute Inc., Cary, NC, USA) and a repeated measures mixed model (PROC MIXED) fitted by a restricted maximum likelihood [42]. The sequence of day, week of sampling or period (PER) was a repeated measure. 3-NOP, CFP, time relative to parturition and the interaction between them were set as fixed effects and each cow within treatment was set as a random effect. Data on indirect calorimetry and gas measurements were evaluated according to periods fixed at d 28 a.p. until parturition (period 1), d 1 until d 28 p.p. (period 2) and d 29 until d 120 p.p. (period 3). For clinical chemistry parameters, the autoregressive variance-covariance structure was selected based on the best fit according to the lowest Akaike Information Criterion and the result of the first measurement at d 28 before 3-NOP supplementation was regarded as a covariate. Parameters of indirect calorimetry and ultrasonic measurements were tested using a compound symmetry structure. Effects were regarded as statistically significant at p-values ≤ 0.05 and a trend was implied at p-values between 0.05 and 0.10. Multiple t-tests (PROC PDIFF) with Tukey-adjusted p-values were computed to evaluate significant means.
The R software package (version 3.6.1, R Foundation for Statistical Computing, Vienna, Austria) was used to calculate Pearson correlation coefficients and to perform linear regression of EB GFE and ER tissue data. Further, R was applied to estimate the O 2 consumption in CON groups in a linear regression model, whereby significant independent variables and related regression coefficients were estimated in a forward stepwise manner.
The degree of agreement between both methods to estimate the energy balance (ER tissue and EB GFE ) was assessed according to Bland and Altman [43]. The differences between EB GFE and ER tissue were plotted against the arithmetic mean of these pairs at each cow and period. The bias as the mean difference including a 95% confidence interval and the standard deviation (SD) of the differences were calculated. The lower and upper limits of agreement (LoA) were calculated (bias ± 1.96 SD) and used to define the range within which 95% of the differences lay. A regression line was plotted through the differences to detect changes in the bias depending on the magnitudes of the measurements themselves.
The normality of differences was tested using the Shapiro-Wilk test in R.

Methane Emission and Respiratory Gas Exchange Measured with the GreenFeed SYSTEM
Data on the emitted fermentation gases and metabolic respiratory gas exchange are presented in Table 2 while detailed dynamics of VCH 4 and VCO 2 emissions on a weekly basis are presented in Schilde et al. [15]. The VCO 2metabolic expressed as a percentage of total VCO 2 decreased after parturition until week 4 p.p. in the LC groups, as depicted in Figure 1A. VCO 2metabolic in NOPLC continued to decline over the course of the experiment ( Figure 1A; 3-NOP × CFP × PER; p < 0.001). The mean VCH 4 over the three periods was reduced by 24.2% in NOPLC and 29% in NOPHC when compared to the respective CON group (Table 2). VCH 4 increased from an average of 315 L/d in period 2 by 27.7% to 438 L/d in NOPLC and by 8% to 341 L/d in NOPHC in period 3 and, therefore, to a greater extent in the NOPLC than in the NOPHC group. VCH 4 (L/d) positively correlated to VCO 2 (L/d) (r = 0.67; p < 0.001; N = 917). Both VCH 4 and VCO 2 production (total and metabolic VCO 2 ) decreased from the a.p. period to parturition but increased thereafter over the p.p. period, which was also the case for VO 2 consumption (Table 2). In period 3, VCO 2 and VCH 4 emissions were affected by 3-NOP and the high CFP (Table 2; 3-NOP × CFP × PER; p < 0.001). Hence, VCO 2 was significantly higher in CONHC, whereas VCH 4 decreased, which was most apparent for NOPHC. VCO 2 production was positively correlated to VO 2 consumption (r = 0.92; p < 0.001; N = 915). VO 2 consumption decreased in the 3-NOP groups over the course of the experiment ( Table 2; 3-NOP × PER; p < 0.001), whereas CFP did not exert an influence. Both the RQ total and RQ metabolic were affected by the 3-NOP × CFP × TIME interaction (Table 3; p < 0.001). The RQ metabolic markedly dropped from approximately 0.92 ± 0.03 during the a.p. period to its lowest point of 0.90 ± 0.007 in week 1 p.p. ( Figure 1B; Table 2). Afterwards, the RQ metabolic increased to 0.99 in NOPLC and 0.94 in CONLC until week 4 p.p. and remained more or less constant. In contrast, the RQ metabolic continued to slightly increase to 1.01 in NOPHC and 0.98 in CONHC until week 9 p.p., respectively (Table 2; Figure 1B; 3-NOP × CFP × TIME; p = 0.024). Table 2. Fermentation and respiration gases (L/d; measured using the GreenFeed system), total (RQ total ) and metabolic respiration quotient (RQ metabolic ) of the experimental groups during period (Per) 1 (d 28 ante-partum until day of calving), 2 (d 1 until d 28 post-partum (p.p.)) and 3 (d 29 until d 120 p.p.).

Energy Turnover Estimated by Indirect Calorimetric and Ultrasonic Methods
The parameters of BW 0.75 , GEI, metabolizable energy intake (MEI), EE and ER are presented in Table 3. GEI and MEI significantly increased in the HC groups by an average of 28% from period 1 to period 2 and by 31% up to period 3, whereas in the LC groups, GEI and MEI increased, on average, by 35% from period 1 to period 2 and by 23% up to period 3 (CFP × PER; p < 0.001). The experimentally intended gradual increase in energy intake resulted, during period 3, in significantly higher daily GE and ME uptakes in the HC groups, by about 0.4 MJ and 0.32 MJ/kg BW 0.75 and d, respectively, when compared to the LC groups.

Energy Turnover Estimated by Indirect Calorimetric and Ultrasonic Methods
The parameters of BW 0.75 , GEI, metabolizable energy intake (MEI), EE an presented in Table 3. GEI and MEI significantly increased in the HC groups by a of 28% from period 1 to period 2 and by 31% up to period 3, whereas in the L GEI and MEI increased, on average, by 35% from period 1 to period 2 and by 2 period 3 (CFP × PER; p < 0.001). The experimentally intended gradual increase intake resulted, during period 3, in significantly higher daily GE and ME upta HC groups, by about 0.4 MJ and 0.32 MJ/kg BW 0.75 and d, respectively, when com the LC groups.   [22], gas volumes used from Table 4. ‡ CH 4 E (MJ/kg) = CH 4 (L/d) × 0.03954 (MJ/L) [22]. ♦ ER total (MJ/d) = MEI − HP. ER tissue = MEI − HP − NE P and ME E , resp. Dairy 2021, 2, x FOR PEER REVIEW Table 3. Energy intake, expenditure and retention of the experimental groups (d 28 ante-partum until day of calving), Per 2 (d 1 until d 28 post-partum (p.p.) d 120 p.p.) estimated according to gas measurements presented in Table 4.  Schilde et al. [15] and calculated according to GfE [24] (footnote Figure 2).   [22], whereby 16% is lost as heat when body tissue energy is converted into milk [40], 1 kg of body fat corresponds to 39.8 MJ of GE [22]. ∏ ER residual (MJ/d) = ER tissue − ER fat depot .

Biochemical Blood Parameters
Lactate peaked on the day of calving (Figure 3A; TIME; p < 0.001). Hydrogen carbonate and the temperature-corrected blood pH marginally fluctuated around their mean of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and 7.39, respectively. However, a slight drop in blood pH values was observed at d 1 p.p. (TIME; p < 0.001; Figure 3C). Antepartal haemoglobin levels slightly increased, on average, from 10.6 to 12.0 g/dL on the day of parturition but continuously decreased by approximately 24% afterwards. From d 49 until termination of the experiment, haemoglobin diverged to constant levels of 9.5 g/dL in the HC groups but still decreased to approximately 8.5 g/dL in the LC groups ( Figure 3D; CFP × TIME; p < 0.001).

Biochemical Blood Parameters
Lactate peaked on the day of calving (Fig bonate and the temperature-corrected blood pH of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and in blood pH values was observed at d 1 p.p. (T moglobin levels slightly increased, on average, rition but continuously decreased by approxim mination of the experiment, haemoglobin dive HC groups but still decreased to approximately × TIME; p < 0.001).

Biochemical Blood Parameters
Lactate peaked on the day of calvin bonate and the temperature-corrected blo of 28.9 mmol/L (TIME; p = 0.208; Figure 3B in blood pH values was observed at d 1 p moglobin levels slightly increased, on ave rition but continuously decreased by app mination of the experiment, haemoglobi HC groups but still decreased to approxim × TIME; p < 0.001).

Biochemical Blood Parameters
Lactate peaked on the day of calving (Figure 3A; TIME; p < 0.001). Hydrogen carbonate and the temperature-corrected blood pH marginally fluctuated around their mean of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and 7.39, respectively. However, a slight drop in blood pH values was observed at d 1 p.p. (TIME; p < 0.001; Figure 3C). Antepartal haemoglobin levels slightly increased, on average, from 10.6 to 12.0 g/dL on the day of parturition but continuously decreased by approximately 24% afterwards. From d 49 until termination of the experiment, haemoglobin diverged to constant levels of 9.5 g/dL in the HC groups but still decreased to approximately 8.5 g/dL in the LC groups ( Figure 3D; CFP × TIME; p < 0.001).

Biochemical Blood Parameters
Lactate peaked on the day of calving (Figure 3A; TIME; p < 0.001). Hydrogen carbonate and the temperature-corrected blood pH marginally fluctuated around their mean of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and 7.39, respectively. However, a slight drop in blood pH values was observed at d 1 p.p. (TIME; p < 0.001; Figure 3C). Antepartal haemoglobin levels slightly increased, on average, from 10.6 to 12.0 g/dL on the day of parturition but continuously decreased by approximately 24% afterwards. From d 49 until termination of the experiment, haemoglobin diverged to constant levels of 9.5 g/dL in the HC groups but still decreased to approximately 8.5 g/dL in the LC groups ( Figure 3D; CFP × TIME; p < 0.001).

Biochemical Blood Parameters
Lactate peaked on the day of calving (Figure 3A; TIME; p < 0.001). Hydrogen carbonate and the temperature-corrected blood pH marginally fluctuated around their mean of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and 7.39, respectively. However, a slight drop in blood pH values was observed at d 1 p.p. (TIME; p < 0.001; Figure 3C). Antepartal haemoglobin levels slightly increased, on average, from 10.6 to 12.0 g/dL on the day of parturition but continuously decreased by approximately 24% afterwards. From d 49 until termination of the experiment, haemoglobin diverged to constant levels of 9.5 g/dL in the HC groups but still decreased to approximately 8.5 g/dL in the LC groups ( Figure 3D; CFP × TIME; p < 0.001).

Energy Turnover Estimated by Indirect Calorimetric and Ultrasonic Methods
The parameters of BW 0.75 , GEI, metabolizable energy intake (MEI), EE and ER are presented in Table 3. GEI and MEI significantly increased in the HC groups by an average of 28% from period 1 to period 2 and by 31% up to period 3, whereas in the LC groups, GEI and MEI increased, on average, by 35% from period 1 to period 2 and by 23% up to period 3 (CFP × PER; p < 0.001). The experimentally intended gradual increase in energy intake resulted, during period 3, in significantly higher daily GE and ME uptakes in the HC groups, by about 0.4 MJ and 0.32 MJ/kg BW 0.75 and d, respectively, when compared to the LC groups.
The 3-NOP × PER interaction (p < 0.001) of HP was driven by a decreasing HP from 3-NOP when compared to the CON groups in period 3 (Table 3). HP was positively correlated with MEI, which was not different between treatment groups (r = 0.37; p < 0.001; N = 895. During the course of the experiment, ME E decreased in the LC groups, whereas that of the HC groups increased (CFP × PER; p < 0.001). With regard to period 3, CH 4 E was lowest in NOPHC, in contrast with the NOPLC and the CON groups (3-NOP × CFP × PER; p < 0.001). During the course of the experiment ER total and ER tissue increased with elevated CFP in the diet (Table 3; CFP × PER; p < 0.05). ER tissue was more positive in the NOPHC group over the experimental periods (Table 3; Figure 1C: 3-NOP × CFP; p = 0.006). ER tissue is shown on a weekly basis in Figure 1C and a sharp drop can be seen in ER tissue starting from the initiation of the trial until week 1 p.p., when the tissue energy balance was the most negative, independent of the experimental group ( Figure 1C; TIME; p < 0.001). In all of the treatment groups, a continuous rise in ER tissue was observed from week 1 p.p. onwards, with this being the most distinctive in the HC groups (CFP; p < 0.001). In the NOPHC group, ER tissue reached a positive range in week 4 p.p. which was earlier when compared to CONHC (positive ER tissue from week 8 p.p.). In contrast, ER tissue in the LC groups remained in a negative range until termination of the trial.
The described group differences concerning the extent of energy retained in body tissues were, however, not recovered in the ER fat depot (Table 3; 3-NOP × CFP; p = 0.637) which was estimated ultrasonographically during period 2. The effect of time (Table A1; TIME; p < 0.001) was reflected by a decrease in each AT depot ( Table 4). Irrespective of treatment group, the average lipomobilization from the visceral and subcutaneous AT of 0.69 kg of fat depot masses per day contributed to a daily energy release of about 177.5 kJ/kg BW 0.75 and d being potentially utilizable for milk synthesis (Table 3). Correspondingly, back fat and rib fat thickness decreased, on average, by 0.14 and 0.15 cm/d, respectively (Table 4). In addition, the visceral fat deposit was mobilized to a larger extent when compared to the subcutaneous one (0.53 kg/d vs. 0.17 kg/d; Table 4).
Due to the described differences in ER tissue between groups but the missing effects of 3-NOP and CFP on depot fat mobilization from ultrasonic measurements, ER residual was higher in the 3-NOP and HC groups (Table 3; 3-NOP; p = 0.058; CFP; p = 0.001).

Validation of the ER tissue Outcome of the GreenFeed Indirect Calorimetry Method
The EB GFE varied between experimental periods, which was similar to ER tissue (Table 3; PER; p < 0.001). In contrast to ER tissue , the EB GFE of the NOPHC group was more positive  (Table 3; 3-NOP × CFP; p = 0.082). The Bland-Altman analysis (Figure 2A; mean bias of 70 kJ NE L /kg BW 0.75 and d; p < 0.001 over all of the experimental periods) and the slope of the regression line of the linear relationship indicated that the EB GFE was estimated to be approximately 33% ( Figure 2B) higher when compared to ER tissue . The slope of the regression line through the data points of differences was not significant (p = 0.756), indicating a constant bias over the experimental periods (Figure 2A). Nevertheless, greater differences between both methods with increasing magnitude of a positive energy balance can be visually identified regarding the a.p. period 1 (Figure 2A,B). Furthermore, the agreement between the EB GFE and the calorimetrically obtained ER tissue was most accurate concerning period 2 and 3 (Figure 2A,B). Hence, a non-significant mean bias of 21 kJ NE L /kg BW 0.75 and d (p = 0.051) was calculated for the agreement between both methods for period 2 and 3. In contrast, greater differences between EB GFE and ER tissue were found in period 1, with a mean bias of 167 kJ NE L /kg BW 0.75 and d (p < 0.001). The average k l over the experimental groups and periods totalled 0.61 (data not shown).

Biochemical Blood Parameters
Lactate peaked on the day of calving (Figure 3A; TIME; p < 0.001). Hydrogen carbonate and the temperature-corrected blood pH marginally fluctuated around their mean of 28.9 mmol/L (TIME; p = 0.208; Figure 3B) and 7.39, respectively. However, a slight drop in blood pH values was observed at d 1 p.p. (TIME; p < 0.001; Figure 3C). Antepartal haemoglobin levels slightly increased, on average, from 10.6 to 12.0 g/dL on the day of parturition but continuously decreased by approximately 24% afterwards. From d 49 until termination of the experiment, haemoglobin diverged to constant levels of 9.5 g/dL in the HC groups but still decreased to approximately 8.5 g/dL in the LC groups ( Figure 3D; CFP × TIME; p < 0.001).
Blood serum concentrations of BHB, NEFA, TAG and glucose are presented in Figure 4. During the transitional period, the characteristic changes of BHB, NEFA, TAG and glucose were observed in all treatment groups (TIME; p < 0.001; Figure 4). In all experimental groups, TAG and glucose decreased by 68% from d 3 a.p. until d 3 p.p. and by 11% from d 1 p.p. to d 7 p.p., respectively. Starting from an initial value of 0.205 mmol/L, the NEFA concentration peaked to 0.856 mmol/L at d 1 p.p. followed by a decline to the a.p. baseline level until d 98 p.p. BHB increased from 0.63 mmol/L at d 3 a.p. to 1.11 mmol/L at d 7 p.p. in the CON groups, whereas a numerically lower peak of 0.88 mmol/L was observed in the 3-NOP groups. 3-NOP treatment did not impact the BHB, TAG and glucose concentrations but lowered that of NEFA by approximately 19.5% in the 3-NOP compared to the CON groups (3-NOP; p < 0.001). CFP affected neither NEFA nor TAG but did affect BHB (CFP × TIME; p = 0.009). Thus, a more pronounced decrease in BHB serum concentrations was observed in the HC compared to the LC groups from d 7 p.p. until termination of the experiment. Elevated blood glucose levels in the HC groups were considered significant from d 21 p.p. until d 73 p.p., in contrast with the LC groups (CFP × TIME; p = 0.073; CFP; p = 0.006). NEFA concentration was correlated with TAG after parturition (r = 0.47; p < 0.001; N = 350). Blood glucose was positively related to TAG (r = 0.49; p < 0.001; N = 532) and HP (r = 0.24; p < 0.001; N = 512) but negatively associated with both serum NEFA (r = −0.28; p < 0.001; N = 532) and BHB (r = −0.47; p < 0.001; N = 532). NEFA and BHB were significantly interrelated (r = 0.42; p < 0.001; N = 532) and decreased with elevated ER tissue (r = −0.52; p < 0.001 for NEFA and r = −0.29; p < 0.001 for BHB; N = 511) and MEI (r = −0.29; p < 0.001; N = 525 for NEFA and r = −0.18; p < 0.001; N = 514 for BHB). Accordingly, increased MEI went along with increased ER tissue (r = 0.39; p < 0.001; N = 914) and CO 2 yield (g CO 2 /kg DMI) (r = 0.41; p < 0.001; N = 914). CO 2 yield had a strongly positive correlation with TAG (r = 0.56; p < 0.001; N = 511) and postpartal NEFA levels (r = 0.44; p = 0.001; N = 350) but had a negative relationship with ER tissue (r = −0.35; p < 0.001; N = 914). Dairy 2021, 2, x FOR PEER REVIEW 14 of 25 Blood serum concentrations of BHB, NEFA, TAG and glucose are presented in Figure  4. During the transitional period, the characteristic changes of BHB, NEFA, TAG and glucose were observed in all treatment groups (TIME; p < 0.001; Figure 4). In all experimental groups, TAG and glucose decreased by 68% from d 3 a.p. until d 3 p.p. and by 11% from d 1 p.p. to d 7 p.p., respectively. Starting from an initial value of 0.205 mmol/L, the NEFA concentration peaked to 0.856 mmol/L at d 1 p.p. followed by a decline to the a.p. baseline level until d 98 p.p. BHB increased from 0.63 mmol/L at d 3 a.p. to 1.11 mmol/L at d 7 p.p. in the CON groups, whereas a numerically lower peak of 0.88 mmol/L was observed in

Limitations of the GreenFeed Technology for its Use in Indirect Calorimetry
The accurate indirect calorimetric calculation of the HP and RQ depends on precise gas respiration measurements [44]. Due to technical reasons, the VO2 consumption of CON groups needed to be regressively predicted from VCO2 and DMI. Even though slightly increased VO2 consumption rates were temporarily observed in the CON groups, the highly predictive performance of the applied model (R 2 = 0.90; RSE = 371 g/d) confirmed its validity. In contrast to RC, a reliable within-day gas exchange pattern could not be obtained from GF measurements, which precluded investigations on intraday HP and RQ kinetics [23] and potentially explained some of the variations, as shown for RQmetabolic and HP. Over the present trial period, the coefficients of variation for the within-day GF spot measurements of VCH4, VO2 and VCO2 were, on a weekly average, 22.1, 10.0 and 11.1%, respectively. Correspondingly, this could have resulted in deviations of HP of approximately ± 1.3 MJ (1.02% of total HP), ± 9.7 MJ (7.65% of total HP) and ± 3.5 MJ (2.76% of total HP), respectively, indicating that HP estimation is most sensitive towards variations in O2 consumption. In conclusion, the overall variability of the present GF gas mass

Limitations of the GreenFeed Technology for Its Use in Indirect Calorimetry
The accurate indirect calorimetric calculation of the HP and RQ depends on precise gas respiration measurements [44]. Due to technical reasons, the VO 2 consumption of CON groups needed to be regressively predicted from VCO 2 and DMI. Even though slightly increased VO 2 consumption rates were temporarily observed in the CON groups, the highly predictive performance of the applied model (R 2 = 0.90; RSE = 371 g/d) confirmed its validity. In contrast to RC, a reliable within-day gas exchange pattern could not be obtained from GF measurements, which precluded investigations on intraday HP and RQ kinetics [23] and potentially explained some of the variations, as shown for RQ metabolic and HP. Over the present trial period, the coefficients of variation for the within-day GF spot measurements of VCH 4 , VO 2 and VCO 2 were, on a weekly average, 22.1, 10.0 and 11.1%, respectively. Correspondingly, this could have resulted in deviations of HP of approximately ± 1.3 MJ (1.02% of total HP), ± 9.7 MJ (7.65% of total HP) and ± 3.5 MJ (2.76% of total HP), respectively, indicating that HP estimation is most sensitive towards variations in O 2 consumption. In conclusion, the overall variability of the present GF gas mass flux measurements was stated to be low. This was related to an accurate data acquisition, which was realized by a high-sampling frequency being evenly distributed throughout the day. In addition, GF data were averaged over seven days and validated for visiting time and head positioning of the cow in the GF hood. Hence, the measurement procedure applied herein (detailed in Schilde et al. [15]) was previously noted to produce comparable results to those obtained from RC [23,32]. In particular, both RC and GF used the same equations and sensor types for O 2 (para-magnetic), CH 4 and CO 2 (non-dispersive infrared) respiration measurements. However, in particular, further validation of the GF algorithm principles is needed as O 2 sensor validation data from RC measurements are lacking. In the present study, the VCO 2metabolic was differentiated from the fermentative VCO 2 to calculate RQ metabolic at the intermediary level. Indeed, this fractionation can be visually conducted for each cow visit from the VCO 2 gas-measurement trajectory depicted in the GF graphical online interface. In this way, a "baseline" CO 2 level reflects the amount of expired lung-derived CO 2 (VCO 2metabolic ) that needs to be corrected for background CO 2 gas concentration. The "baseline" CO 2 level is temporarily interrupted by CO 2 eructation peaks (VCO 2fermentative ) [45]. However, this visual evaluation is impractical for large datasets and, therefore, algorithms for an automatized graphical assessment should be developed in the future. As a consequence, the commonly applied factor of 1.7 [34,36] was used, resulting in VCO 2fermentative proportions of 12 ± 0.5% in CONHC, 9 ± 0.5% in NOPHC and, more incrementally, 13 ± 0.4% in CONLC and 10 ± 1.1% in NOPLC (mean ± SD) ( Figure 1A; Table 2). Comparatively, Caetano et al. [45] visually estimated the VCO 2fermentative from the GF online interface to be between 6 and 20% of the total VCO 2 production in beef cattle offered diets of varying energy density for ad libitum and restricted intake.

Validation of the Energy Partitioning Estimated by Indirect Calorimetry and Ultrasonography
The present GF method of indirect calorimetry resulted in ER tissue values that strongly corresponded to the EB GFE values measured for period 2 and 3. However, both methods significantly differed with regard to the a.p. period (period 1; Figure 2A; compare ER tissue and EB GFE in Table 3). Erdmann et al. [39] compared the EB GFE with that calculated from indirect calorimetric RC measurements over the same antepartal period as the present period 1 and also reported higher EB GFE values (by about 33 MJ/d) when compared to the RC energy balance. The higher EB GFE could have been a result of an underestimation of EE during the ante-partum period 1 when compared to the calorimetrically derived ER tissue . Thus, the dynamically increasing antepartal energy requirements for the onset of lactogenesis and foetal growth could have been captured more accurately by continuous calorimetric measurements in contrast with the constants applied in EB GFE calculations. Furthermore, the impact of maintenance requirements on the EB outcome was proportionally higher during the dry period when compared to the lactation period. The factors applied in the German NE system for calculating ME m were derived from 40-year-old data. Meanwhile, the breeding of higher genetic merit cows resulted in generally increased body sizes of cows and a greater proportion of liveweight as body protein mass while back fat thickness decreased [46]. As a consequence, the increased feed intake resulted in greater digestive loads and blood flow-rates in the total splanchnic tissues being paralleled by increased metabolic rates, internal organ masses and O 2 consumption [47]. These metabolic changes are related to higher energy demands for maintenance metabolism, which implies an underestimation of maintenance energy requirements in the German NE feeding system and a further explanation of the higher EB GFE when compared to the ER tissue values.
The mean k l value of 0.61 is within the range of k l values (0.60 to 0.67) summarized in a literature review by Agnew and Yan [47] and close to the k l value of 0.60 reported by Van Es [48], which confirms the suitability of the GreenFeed system as an indirect calorimeter.
The estimated energy released from the ultrasonographically assessed lipolysis in AT depots (ER fat depot ; Equation (8)) was subtracted from the negative ER tissue in period 2 (Table 3) yielding the remaining fraction of glycogen, triglycerides and proteins deposited in skeletal muscles and organs (ER residual ). It was supposed that protein and lipid breakdown in skeletal muscles around parturition partially compensated for the observed negative ER tissue ( Figure 1C; Table 3) and contributed to the decreased RQ metabolic ( Figure 1B) [11]. Thus, gluconeogenesis from the oxidation of alanine, one of the most important glucogenic amino acids (AA) [11], and intramuscular lipids result in very low RQ metabolic values of 0.13 and 0.7, respectively. Tamminga et al. [49] estimated the fractional rate of skeletal muscle protein breakdown in dairy cows to be 0.38, 0.22, 0.04 and 0.02 kg per day in week 1, 2, 3 and 4 p.p. From a rough calculation, this would correspond to a total of 4.6 kg mobilized body protein (92 MJ NE L ) during the complete period 2 and an energy equivalent of 3.3 MJ NE L /d (92 MJ NE L from proteolysis divided by 28 days in period 2 = 3.3 MJ NE L /d; 1 g of body protein = 23.8 kJ [22]; energy efficiency of 84% [50]). Comparatively, von Soosten et al. [51] reported a lower energy yield from body protein mobilization, which amounted to an average of 2.1 MJ/d over the period from d 1 until d 42 p.p. in primiparous cows measured by the comparative slaughter technique. However, those results are not directly comparable to the present periparturient pluriparous dairy cows and observation period (d 1 until d 28). Von Soosten et al. [51] assumed protein accretion in the growing primiparous cows (BW of approximately 490 kg) and protein mobilization is generally supposed to change to repletion from d 35 p.p. onwards [52]. The non-explained remainder of the difference between ER residual and the estimated energy supply from skeletal muscle proteolysis can be partially assigned to energy mobilized from inter-and intramuscular and organ tissues. Furthermore, the corresponding models estimating the fat depot masses are, to some extent, prone to error. Although ER tissue was more positive in the HC groups ( Figure 1C), the ultrasonographically assessed lipolysis from AT and serum NEFA levels (indicative for negative EB) were not different between the HC and LC groups (Table 3; Figure 4B). Raschka et al. [25] validated the ultrasonographic-based multiple regression model for the predicted weights of the SAT and VAT depots as highly accurate with R 2 values of 0.88 and 0.94 and root mean square errors of 3.4 and 6.1 kg, respectively. In the present experiment, an assumed ± 10% variation between the predicted and actual daily changes in SAT and VAT would result in ER fat depot variations of approximately ± 2.3 MJ NE L /d.

Effects of 3-NOP, CFP and Parturition on Energy Metabolism Parameters
Both RQ and HP notably depend on the magnitude of MEI (r = 0.69 and r = 0.22 resp.; p < 0.001; N = 914), its utilization for maintenance and productive purposes [53], whether substrates are either deposited or mobilized in tissues ( Figure 5A) and, finally, on the type of the metabolized substrate itself [22,34]. In contrast to LC groups, the higher RQ metabolic ( Figure 1B) and ER tissue ( Figure 1C) in HC groups reflected their increased GEI:EE ratio (Table 3) and dietary content of non-fibre carbohydrates (Table 1) being microbially degraded into gluconeogenic substrates. Correspondingly, increased blood glucose ( Figure 4C) and reduced BHB ( Figure 4A) concentrations were observed in the HC groups.
During the a.p. period, the pro-lipogenic effect of the dietetically designed energetic oversupply was manifested in the positive ER tissue ( Figure 1C) and RQ metabolic values of 0.92 ( Figure 1B). In principle, lipid deposition in AT would result in RQ metabolic values above 1.0 [44] but RQ metabolic reflects the net oxidation rates of a mixture of substrates irrespective of the metabolic interconversions of the substrate [44]. Hence, flowing transitions between oxidation and de novo synthesis of lipids were assumed, which became apparent in the steady decrease in ER tissue since the beginning of the trial in spite of the energetic oversupply ( Figure 1C). The present energy-deficient transition from gestation to lactation was accompanied by significant metabolic adaptations (Figure 4). The decreased RQ metabolic corresponded to the negative ER tissue and accumulation of serum NEFA ( Figure 5A), collectively indicating excessive fat oxidation from AT resulting in increased O 2 consumption (NEFA vs. O 2 consumption (g/d); r = 0.18; p < 0.001; N = 511) and ketogenesis from acetyl-CoA and NEFA (NEFA vs. BHB; r = 0.42; p < 0.001; N = 532). The observed increased circulating BHB ( Figure 4A) likely originated from an oxaloacetate deficiency [10] and a concomitant hepatic overload to completely oxidize the excessively flooding NEFA (Figure 4B), released by lipolysis in AT (Table A1 and Table 4), into ATP and CO 2 [11]. It can be summarized that the decrease in RQ metabolic could be partially explained by the increased O 2 consumption due to lipolysis in AT, whereby CO 2 did not increase due to the aforementioned incomplete metabolization of NEFA into BHB but not into CO 2 . Correspondingly, the RQ metabolic did not behave in the same manner as the RQ total because the latter also reflected CO 2 production arising from rumen fermentation. Hence, in the early-lactation period, increased fermentative CO 2 production from high-forage diets led to higher RQ total values, whereby RQ metabolic decreased due to the abovementioned increased but incomplete fat oxidation resulting in less intermediary CO 2 formation when NEFA were converted to BHB, rather than CO 2 and ATP. Accordingly, as previously published for the present experiment, the CO 2 yield (g CO 2 production/kg DMI) was significantly higher in LC when compared to the HC groups, but the opposite was the case when it came to total CO 2 production (g CO 2 /d) over the complete experimental period [15].
In the present experiment, the tendency for an increased ER total in the NOPHC group (Table 3; Figure 1C) confirmed similar results reported by van Gastelen et al. [13]. The increased ER tissue , ER residual and RQ metabolic of 1.01 in the NOPHC group ( Figure 1B; Table 2) could be explained by an improved energy budget in that group. Hence, decreased NEFA levels were associated with increased ruminal propionate concentrations, and both were inversely related to CH 4 production ( Figure 5B). Recently, it was observed that supplementing 3-NOP combined with high CFP in the ration shifted rumen fermentation pathways to hydrogen-consuming glucogenic propionate and decreased loss of CH 4 energy in a synergistic manner (Tables 2 and 3) (details in Schilde et al. [15]). Correspondingly, lower serum NEFA concentrations were observed in the 3-NOP cows ( Figure 4B) although neither 3-NOP nor CFP affected ER fat depot and lipomobilization in AT depots (Tables 3 and 4). This could indicate that the increased glucogenic propionate proportions in the 3-NOP groups improved the intramitochondrial oxaloacetate availability and, therefore, the hepatic capacity for NEFA oxidation. Interestingly, neither blood glucose (NEFA oxidation and conversion of elevated propionate levels to glucose and CO 2 ) nor TAG (re-esterification of NEFA) and BHB (reduced incomplete NEFA oxidation) were affected by 3-NOP, which confirms previous findings [20]. This opens the question as to whether the direct extrapolation of NEFA concentrations to circulating BHB levels is appropriate in the present CH 4 mitigation experiment. Accordingly, in the companion study, Schilde et al. [15] observed that butyrate formation was preferred to that of acetate in the 3-NOP-treated cows, which can be explained by the reduced hydrogen release when carbohydrates are degraded into butyrate and not into acetate [54]. Butyrate also serves as a carbon source for ketone body synthesis in the rumen epithelium [55]; therefore, increased circulating BHB originating from enhanced intraepithelial ketogenesis could have masked the assumed causal relationship that decreased serum NEFA concentrations in the 3-NOP-treated cows, which would necessarily have led to reduced BHB in the blood stream. Besides the intraepithelial butyrate metabolization, propionate can be metabolized to lactate in the rumen epithelium, which could also have reduced the propionate flux to the liver, thereby eliminating the energetic advantage of the 3-NOP-mediated increased propionate formation in the rumen. The observed accumulation of ketoacids (BHB) and the blood lactate peak at d 1 p.p. (Figure 3A) could have increased the risk for metabolic acidosis [44]. Indeed, blood pH was observed to slightly drop from 7.41 to 7.38 at parturition contemporaneously to the lactate peak at d 1 p.p. (Figure 3C; TIME; p < 0.001). In this context, the temporal decrease in the 3-NOP groups ( Figure 3C; 3-NOP × TIME; p = 0.014) is, however, difficult to explain as blood lactate and BHB were not affected by 3-NOP treatment. The pH decrease at d 1 p.p. possibly caused buffering reactions via the largest CO 2 body pool, hydrogen carbonate, which could have led to an overestimation of HP and RQ metabolic [44]. Indirect calorimetry is stated to be accurate as long as body pool sizes of energy-related metabolites (ketone bodies, lactate) and intermediary products (O 2 and CO 2 , N U ) remain stable [44]. However, the potential effects of intermediary pool sizes on HP from nutrient oxidation and RQ metabolic were considered negligible because the bicarbonate and pH values remained within their physiological area [56,57] (Figure 3B,C). In general, the CO 2 pool size is supposed to be subjected to greater fluctuations when compared to the O 2 body pool [44]. Accordingly, blood concentrations of haemoglobin, the main O 2 body pool, remained stable within the physiological range [57] although a slight divergence was observed between the LC and HC groups at the end of the experiment ( Figure 3D; CFP × TIME; p < 0.001).

Conclusions
The present study revealed that using the GF system as an indirect calorimetry chamber for the assessment of cows' energy metabolisms is a promising approach, although further validations of the O 2 sensor and algorithm principles are needed. The ER tissue determined by indirect calorimetry coincided with that calculated from GfE [24], except for the antepartal period. The hypothesis that feeding 3-NOP in combination with high CFP synergistically improves the cows' energy budgets was partially confirmed because effects were not apparent in all of the examined parameters. 3-NOP combined with high CFP increased RQ metabolic and ER tissue and decreased serum NEFA. In contrast, lipomobilization from fat depots and blood lactate were neither affected by 3-NOP nor CFP and 3-NOP did not affect blood glucose, TAG and BHB levels. Blood pH and bicarbonate remained within their physiological range and metabolic adaptations to energy-related changes via the CO 2 body pool were not observed. High CFP decreased BHB but increased blood glucose and, at the end of the trial, haemoglobin levels, which possibly indicates that the cows adapted differently to metabolic changes. Future research will be focused on the relationship between the 3-NOP-induced changes in the rumen VFA profile and gene expression in the liver.

Data Availability Statement:
The data presented in this study are available in the present article and in the previously published manuscript of the comprehensive experiment by Schilde et al. [15] (article number: 1877986).