Estimation of Individual Glucose Reserves in High-Yielding Dairy Cows

Abstract

Glucose plays a central role in numerous physiological processes in dairy cows related to immune defence and milk production. A lack of glucose impairs both objectives, although to different degrees. A method for the estimation of glucose balance (GB) in dairy cows was developed to assess glucose reserves in the intermediary metabolism. Digestive fluxes of glucogenic carbon were individually estimated via the Systool Web application based on data on body weight (BW), dry matter intake (DMI), and chemical analyses of feedstuffs. Fluxes of endogenous precursors glycerol, alanine and L-lactate and the glucose demand imposed by major glucose-consuming organs were deduced from BW, lactose yield and lactation stage. GB was calculated for 201 lactations (1 to 105 DIM) of 157 cows fed isoenergetic rations. Individual DMI, BW and milk yield were assessed on a daily basis. The results showed that the GB varied greatly between cows and lactation stages. In the first week of lactation, average daily GB reached levels close to zero (3.2 ± 13.5 mol C) and increased as lactation progressed. Most cows risk substantial shortages of glucose for maintenance during the first weeks of lactation. In face of the specific role of glucose for the functional capability of the immune function, the assessment of glucose reserves is a promising measure for the identification of cows at risk of impaired immunocompetence.

1. Introduction

In contrast to monogastric animals, ruminants cover large proportions of their glucose demand through hepatic gluconeogenesis, while only small amounts are absorbed directly from the gut [1]. Major digestive precursors for gluconeogenesis include propionate, L-lactate and glucogenic amino acids [2,3]. In the event of shortage, precursors such as glycerol and alanine from the mobilization of adipose (lipolysis) and muscle tissue (proteolysis) as well as the increased recycling of lactate from muscle tissue can provide additional glucogenic carbon (C). In lactating animals, high amounts of glucose are required as a precursor for milk lactose [4] as well as for nicotinamide adenine dinucleotide phosphate reduction during the synthesis of milk fatty acids and milk proteins [5,6]. Furthermore, glucose is an essential fuel for self-sustaining neurological and immune functions. In the event of an infection, the immune system is a top consumer of glucose, inter alia required for the accelerated processes of immune cell proliferation and differentiation, phagocytosis and the production of reactive oxygen species [7]. Accordingly, a strong activation of immune functions is associated with a drop in the concentration of plasma glucose, e.g., following the injection of endotoxins [8] or immunostimulants [9]. Kvidera et al. [8] showed that during the 12 h following the injection of lipopolysaccharide (LPS), the glucose demand on the immune system of dairy cows may increase up to 1.5 kg. This demand is similar to the amount of glucose required for the synthesis of ~40 kg of milk per day. With regard to the metabolic priority for milk production during early lactation [10], shortages in the availability of glucose in the intermediary metabolism may thus compromise not only milk production but also immune cell function, such as phagocytosis and the synthesis of immunometabolites, and may be related to the occurrence of production diseases [11].

In the past, various models of glucose metabolism in dairy cows have been developed to study metabolic adaptations during the periparturient period [12] as well as the interrelationships between glucose metabolism and health parameters, such as fertility [13] and ketogenesis [14]. In those models, digestive processes leading to glucogenic metabolites absorbed from the gut were not considered in detail. However, detailed aspects of ruminant digestion, such as ruminal propionate production as well as starch and protein digestion, have been modelled since then, both in connection with ruminant feed formulation software for dairy cows [15,16,17] and as a part of research models including regulatory subsystems [18,19]. Following the claim “from joules to moles”, the calculation of digestive nutrient fluxes from nutrient composition of feedstuffs and rations fed was recently brought forward in the INRA feeding system for ruminants [20]. This enables the cow- and ration-specific analysis of digestive nutrient fluxes. Moreover, regression equations for the prediction of the net portal appearances (NPA) of glucose and glucose precursors from these nutrient fluxes have been developed [21].

Further research on quantitative glucose metabolism encompassed tracer studies on the rates of disappearance of glucose in peripheral tissues [22,23]. In protein tissue (PT), the ratio of glucose being oxidized or converted to L-lactate, a precursor for gluconeogenesis [2], is affected by the stage of lactation. Moreover, the dynamics of adipose (AT) and PT mobilisation and their shares in providing the glucose precursors glycerol and alanine [24] as well as the demands caused by major glucose-consuming organs (mammary gland, muscle tissue, red blood cells and the brain) can be deduced from animal data such as the lactose yield and changes in BW [25,26,27].

While comprehensive research has been conducted in the different subareas of intermediary metabolism, knowledge has so far not been operationalized by integrating input and output values into a coherent whole-body model for a glucose balance sheet on a daily basis. Therefore, the aim of this study was to develop a herd-data-based method of assessing the glucose balance (GB) of dairy cows during early lactation. In order to identify the extent as well as the inter- and intraindividual variations in GB over the course of early lactation (1 to 105 days in milk), key parameters such as the level and timepoint of the lowest GB (nadir), the contribution of precursors to the overall glucose supply and the contribution of glucose-consuming tissues to overall glucose demand have been assessed and evaluated for a sample of 157 dairy cows (201 lactations).

2. Materials and Methods

2.1. Model

2.1.1. General/Overview

The mechanistic, conceptually driven whole-body model of dairy cow GB described in this paper was initially a research model, but it has prospects for practical use in dairy farming. The main input variables are milk yield (MY) and milk composition, BW, DMI, ration composition and the chemical analyses of feedstuffs. The structure of the calculations integrates the main pathways of the absorption of glucose and other glucogenic C (propionate, L-lactate, amino acids) from the rumen and the duodenum, the supply of the endogenous precursors glycerol and alanine (during periods when body reserves are mobilized), the recycling of glucogenic C through L-lactate, and the demands of the main glucose-consuming tissues. The input sub-system (GLU_IN), i.e., the calculation of glucose supply, is based on data-driven models that predict the portal flows of glucose and glucogenic precursors according to Loncke et al. [21] and Martineau et al. [28] as well as estimates of the release of glucogenic C from AT and PT [24,29]. Further, the estimates of the proportions of glucose oxidation and L-lactate production in PT, and thus the recycling of L-lactate from PT for hepatic gluconeogenesis, are included according to the research by Larsen and Kristensen [2]. The output sub-system (GLU_OUT) encompasses the estimation of the daily glucose demand of major glucose-consuming tissues, including the mammary gland (MG), PT, the brain and red blood cells (RBC), derived from previously published research in the field [22,27,30,31,32,33,34]. “Mol of glucogenic C per day” (mol C/d) was used as a general unit for the calculation of in- and output fluxes and GB. The main pathways included in the model are illustrated in Figure 1.

2.1.2. Digestive and Portal Flows of Glucogenic Nutrients

The major part of circulating glucose derives from hepatic gluconeogenesis. Metabolites from both the feed and the intermediary metabolism serve as precursors. Propionate, which is produced from long-chain carbohydrates by rumen bacteria, is the main glucogenic precursor, supplying between 40 and 90% of total C for gluconeogenesis [1,35]. Other precursors that derive from the digestion of feedstuffs encompass glucogenic amino-acids absorbed from the gut as well as L-lactate, which is synthesized by rumen microbes, particularly when a large amount of starch is fed [1]. However, the availability of these exogenous precursors of gluconeogenesis depends on the rates of transfer from the ruminal or duodenal lumen to the portal vein, as propionate is used or modified by first-pass metabolism within the rumen epithelium [36,37]. In contrast, amino acids serve as an energy source throughout the digestive tract [38,39]. In contrast with monogastric animals, only small amounts of total glucose availability in ruminants originate from glucose absorbed directly from the gut. Even in dairy cows fed high-concentrate diets, glucose absorption is limited, primarily by the rate of starch hydrolysis in the gut [40,41,42]. In addition to the utilization of luminal glucose within the duodenal epithelium, the portal-drained viscera may use 0.3 to 2.2 g/d/kgBW of glucose from arterial supply [43,44].

Based on a series of scientific papers reporting regression equations for the main digestive and portal flows of glucogenic precursors [21,28,37,45], the net portal appearances (NPAs) of propionate (PRO), glucose (GLU), L-lactate (LAC) and glucogenic amino acids (gAA) were estimated based on the concentrations of ruminal fermentable organic matter (RfOM), rumen propionate production, truly digestible protein in the intestines (PDI), digestible starch in the intestines (STdI) and organic matter digestibility (OMd). The estimations are based on the work of Loncke and colleagues [21,28,37,45], who established response equations through a meta-analysis of studies involving multicatheterized ruminants. These equations were compiled in the FLORA database and include a diversity of experimental factors (species, physiological state, nutritional regimes). The digestive flows of RfOM, rumen propionate production, PDI, STdI and Omd, as well as contents of ruminal digestible NDF (RdNDF) are estimated from the nutrient compositions of feedstuffs calculated with the Systool Web application [46].

For the equations applied in this study, contents of three moles of glucogenic C per mole propionate as well as per mole L-lactate, four moles glucogenic C per mole of nitrogen (N; according to the average C/N ratio found in glucogenic amino acids), a proportion of 45% glucogenic amino acids of total amino acids and molar weights of 73.1, 89.1, 14.01 and 180.2 g/mol for propionate, L-lactate, N and glucose, respectively, were assumed (Table 1, Equation (2)).

Table 1. Equations for the calculation of glucose balance applied in this study.

No	Variable	Equation	Reference
(1)	Glucose balance (GB; mol C)	GS–GD
(2)	Supply of glucogenic carbon (GS; mol C/d)	(PRO/1000 × 3) + (GLU/1000 × 24 × 6) + (LACdiet/89.1 × 3) + (gAA/1000 × 14.01 × 4) + (GLY/92.1 × 3) + (ALA/89.1 × 3) + LACendo
(3)	NPA of propionate (PRO, mmol/d/kgBW)	3.8 + 0.72 × RU_Pro; Sy.x = 1.6	[45]
(4)	Propionate in the rumen (RU_Pro, mmol/d/kgBW)	PROD_VFA (mol/kg DM) × DMI × Pro %/BW × 1000
(5)	NPA of glucogenic amino acids (gAA, mgN/d/kgBW)	NPA_tAA × 0.45	[47]
(6)	NPA of α-amino-N (NPA_αAA, mgN/d/kgBW)	63 + 0.74 × PDI (mgN/d/kgBW); RMSE = 60	[28]
(7)	NPA of total amino acids (NPA_tAA, mgN/d/kgBW)	NPA_ αAA × 1.3958	[48]
(8)	NPA of L-lactate (LACdiet, g/d/kgBW)	= 0.098 + 0.0098 × RfOM (g/d/kgBW) RMSE = 0.022	[21]
(9)	NPA of glucose (GLU, mmol/h/kgBW)	−0.103 + 0.0913 × StdI (g/d/kgBW); RMSE = 0.035	[21]
(10)	Mobilized glycerol (GLY; mol/d)	ΔLIP × 0.105 × 1000	[24]
(11)	Mobilized alanine (ALA; mol/d)	ΔPROT × 0.044 × 1000	[29]
(12)	Adipose tissue mobilized (ΔLIP; kg/d)	d_L × ΔEBW (kg) × −1
(13)	Protein tissue mobilized (ΔPROT; kg/d)	d_P × ΔEBW (kg) × −1
(14)	Empty body weight (EBW; kg)	BW (kg) − TDC	[20]
(15)	Total digestive contents (TDC, kg)	TRC/0.7	[20]
(16)	Total rumen contents (TRC, kg)	RL × 1.15	[20]
(17)	Rumen liquid (RL; kg)	3.78 × (NDF %BW − 1) + 12)/100) × BW (kg)	[20]
(18)	Reference level of adipose tissue mobilization (ΔLIPref; kg)	−1315 × exp(−0.0329 × DIM) + 0.124 × exp(0.0015 × DIM)	[20,49]
(19)	Reference level of protein tissue mobilization (ΔPROTref; kg)	−0.413 + exp(−0.0488 × DIM) + 0.0268 × exp(0.00047 × DIM)	[20,49]
(20)	Ratio of adipose tissue mobilized (d_L)	ΔLIPref/(ΔLIPref + ΔPROTref)
(21)	Ratio of protein tissue mobilized (d_P)	ΔPROTref/(ΔLIPref + ΔPROTref)
(22)	Endogenous L-lactate (LAC; mol C/d)	Lac % × GDPT	[19]
(23)	Lac %	1 − 0.5 × (ILR/ILRref)
(24)	Irreversible loss rate of glucose in peripheral tissue (ILR; mmol/h/kg0.75)	−0.35720 × 10−4 × DIM3 + 0.00386 × DIM2 − 0.08329 × DIM + 2.42587	derived from [2]
(25)	Glucose demand (GD; molC/d)	GDMG + GDPT + GDRBC + GDBrain
(26)	Glucose demand of mammary gland (GDMG; mol C/d)	lactose yield (g/d)/0.80/180.16 × 6	[30,31]
(27)	Glucose demand of protein tissue (GDPT; molC/d)	PT (kg) × 0.0288 mol/d × 6	[22]
(28)	Protein tissue mass (PT; kg)	(EBW − AT) * 0.215	[20]
(29)	Adipose tissue mass (AT; kg)	0.0377 × ((BCS − 0.5) × 8/4) × EBW	[50]
(30)	Glucose demand of red blood cells (GDRBC; molC/d)	4.75 × BW × 10−4 × 6	[27,34]
(31)	Glucose demand of brain (GDBrain; molC/d)	3.58 × BW × 10−4 × 6	[32,33]
(32)	Energy balance (EB; MJ)	(DMI (kg) × NEL(MJ NEL/kg)) − (ED_M + ED_P)
(33)	Energy demand for production (ED_P; MJ of NEL)	MY × 1.05 + 0.38 × F % + 0.21 × P %	[51]
(34)	Energy-corrected milk (ECM; kg)	E_P/3.28 MJ of NEL/kg	[51]
(35)	Energy demand for maintenance (E_M; MJ of NEL)	0.293 MJ/kg × BW0.75 (kg)	[51]

BCS: body condition score; BW: individual body weight; DIM: days in milk; DMI: dry matter intake (kg dry matter (DM) per day); NDF %BW: level of NDF intake (% of individual body weight); ILRref: reference level of irreversible loss rate of glucose; MY: milk yield; NEL: content of net energy for lactation in the ration (MJ/kg DM); NPA: net portal appearance; PDI: protein truly digestible in the intestines; Pro %: percentage of propionate production of total rumen VFA production; PROD_VFA: total rumen VFA production; StdI: starch digestible in the intestines.

Table 1. Equations for the calculation of glucose balance applied in this study.

No	Variable	Equation	Reference
(1)	Glucose balance (GB; mol C)	GS–GD
(2)	Supply of glucogenic carbon (GS; mol C/d)	(PRO/1000 × 3) + (GLU/1000 × 24 × 6) + (LACdiet/89.1 × 3) + (gAA/1000 × 14.01 × 4) + (GLY/92.1 × 3) + (ALA/89.1 × 3) + LACendo
(3)	NPA of propionate (PRO, mmol/d/kgBW)	3.8 + 0.72 × RU_Pro; Sy.x = 1.6	[45]
(4)	Propionate in the rumen (RU_Pro, mmol/d/kgBW)	PROD_VFA (mol/kg DM) × DMI × Pro %/BW × 1000
(5)	NPA of glucogenic amino acids (gAA, mgN/d/kgBW)	NPA_tAA × 0.45	[47]
(6)	NPA of α-amino-N (NPA_αAA, mgN/d/kgBW)	63 + 0.74 × PDI (mgN/d/kgBW); RMSE = 60	[28]
(7)	NPA of total amino acids (NPA_tAA, mgN/d/kgBW)	NPA_ αAA × 1.3958	[48]
(8)	NPA of L-lactate (LACdiet, g/d/kgBW)	= 0.098 + 0.0098 × RfOM (g/d/kgBW) RMSE = 0.022	[21]
(9)	NPA of glucose (GLU, mmol/h/kgBW)	−0.103 + 0.0913 × StdI (g/d/kgBW); RMSE = 0.035	[21]
(10)	Mobilized glycerol (GLY; mol/d)	ΔLIP × 0.105 × 1000	[24]
(11)	Mobilized alanine (ALA; mol/d)	ΔPROT × 0.044 × 1000	[29]
(12)	Adipose tissue mobilized (ΔLIP; kg/d)	d_L × ΔEBW (kg) × −1
(13)	Protein tissue mobilized (ΔPROT; kg/d)	d_P × ΔEBW (kg) × −1
(14)	Empty body weight (EBW; kg)	BW (kg) − TDC	[20]
(15)	Total digestive contents (TDC, kg)	TRC/0.7	[20]
(16)	Total rumen contents (TRC, kg)	RL × 1.15	[20]
(17)	Rumen liquid (RL; kg)	3.78 × (NDF %BW − 1) + 12)/100) × BW (kg)	[20]
(18)	Reference level of adipose tissue mobilization (ΔLIPref; kg)	−1315 × exp(−0.0329 × DIM) + 0.124 × exp(0.0015 × DIM)	[20,49]
(19)	Reference level of protein tissue mobilization (ΔPROTref; kg)	−0.413 + exp(−0.0488 × DIM) + 0.0268 × exp(0.00047 × DIM)	[20,49]
(20)	Ratio of adipose tissue mobilized (d_L)	ΔLIPref/(ΔLIPref + ΔPROTref)
(21)	Ratio of protein tissue mobilized (d_P)	ΔPROTref/(ΔLIPref + ΔPROTref)
(22)	Endogenous L-lactate (LAC; mol C/d)	Lac % × GDPT	[19]
(23)	Lac %	1 − 0.5 × (ILR/ILRref)
(24)	Irreversible loss rate of glucose in peripheral tissue (ILR; mmol/h/kg0.75)	−0.35720 × 10−4 × DIM3 + 0.00386 × DIM2 − 0.08329 × DIM + 2.42587	derived from [2]
(25)	Glucose demand (GD; molC/d)	GDMG + GDPT + GDRBC + GDBrain
(26)	Glucose demand of mammary gland (GDMG; mol C/d)	lactose yield (g/d)/0.80/180.16 × 6	[30,31]
(27)	Glucose demand of protein tissue (GDPT; molC/d)	PT (kg) × 0.0288 mol/d × 6	[22]
(28)	Protein tissue mass (PT; kg)	(EBW − AT) * 0.215	[20]
(29)	Adipose tissue mass (AT; kg)	0.0377 × ((BCS − 0.5) × 8/4) × EBW	[50]
(30)	Glucose demand of red blood cells (GDRBC; molC/d)	4.75 × BW × 10−4 × 6	[27,34]
(31)	Glucose demand of brain (GDBrain; molC/d)	3.58 × BW × 10−4 × 6	[32,33]
(32)	Energy balance (EB; MJ)	(DMI (kg) × NEL(MJ NEL/kg)) − (ED_M + ED_P)
(33)	Energy demand for production (ED_P; MJ of NEL)	MY × 1.05 + 0.38 × F % + 0.21 × P %	[51]
(34)	Energy-corrected milk (ECM; kg)	E_P/3.28 MJ of NEL/kg	[51]
(35)	Energy demand for maintenance (E_M; MJ of NEL)	0.293 MJ/kg × BW0.75 (kg)	[51]

BCS: body condition score; BW: individual body weight; DIM: days in milk; DMI: dry matter intake (kg dry matter (DM) per day); NDF %BW: level of NDF intake (% of individual body weight); ILRref: reference level of irreversible loss rate of glucose; MY: milk yield; NEL: content of net energy for lactation in the ration (MJ/kg DM); NPA: net portal appearance; PDI: protein truly digestible in the intestines; Pro %: percentage of propionate production of total rumen VFA production; PROD_VFA: total rumen VFA production; StdI: starch digestible in the intestines.

2.1.3. The Mobilization of Glucogenic Precursors from Lipolysis and Proteolysis and L-Lactate Production in Muscle Tissues

The daily rate of the mobilization of glucogenic precursors from lipolysis and proteolysis was estimated from the daily changes in empty body weight (ΔEBW, kg), which can be obtained from individual BW and the level of NDF intake (NDF %BW; kg DM/kg BW) according to INRA (Table 1, Equations (14)–(17)). If ΔEBW was negative (BW loss), the amounts of daily lipid (ΔLIP) and protein mobilisation (ΔPROT) were obtained by multiplying individual daily ΔEBW with the day-specific shares of lipids (d_L) and proteins (d_P) in the daily ΔEBW (Table 1, Equations (12) and (13)). The ratios for d_L and d_P were calculated as the share of daily lipid and protein mobilization (ΔLIPref. and ΔPROTref.; Table 1, Equations (20) and (21)) in the sum of ΔLIPref. and ΔPROTref. The reference values for daily lipid and protein mobilization (Table 1, Equations (18) and (19)) originate from the intra-experiment adjustment of experiments on body lipid and body protein kinetics in dairy cows reported in the INRA feeding system for ruminants [20,49]. It is assumed that all circulating glycerol derives from lipolysis and enters gluconeogenesis. Based on the average molar masses of fatty acids (887 g/mol) and the molar mass of glycerol (92 g/mol), the share of glycerol in the amount of mobilized adipose tissue is 10.5%. Per mol of glycerol, three moles C were assumed (Table 1, Equation (10)).

Alanine is the only amino acid involved in the transfer of glucogenic C between muscle and liver [2,52]. Its share in the amount of mobilized PT was reported to be 4.4% [29], with a molar weight of 89.1 g and three moles C per mole of alanine (Table 1, Equation (11)).

Additionally, gluconeogenesis is supported by a substantial inter-organ transfer of L-lactate, emerging from L-lactate production in PT (Cori cycling) [2]. In this study, the individual daily amounts of protein (PT) and adipose tissue (AT) were calculated from BCS and EBW (Table 1, Equations (28) and (29)) [20,50]. Based on the work of Lindsay et al. [53], Rhodes et al. [54] and Martin and Sauvant [19], 50% of glucose being oxidized and 50% of glucose being recycled through the L-lactate production in PT were assumed as initial values. However, the whole-body irreversible loss rate of glucose excluding the loss of glucose in lactose was found to be reduced during early lactation [35], indicating the reduced oxidation of glucose in peripheral tissues. Based on the data of Bennink et al. [55], Bruckental et al. [56] and Baird et al. [57] as well as unpublished data from Larsen and Kristensen [35], who measured the glucose rate of disappearance during the transition period of dairy cows, a 3^rd-degree polynomial regression equation was derived to calculate the amount of glucose being oxidized and the amount of glucose being released as L-lactate from muscle according to DIM (Figure 2A; Table 1, Equations (22)–(24)). The curve obtained showed a drop around parturition followed by an increase up to 3.55 mmol/h/kg^0.75 (DIM 55). Thereafter, a constant rate of L-lactate recycling (Lac %) was presumed until DIM 105 (Figure 2B) together with a complete recycling of endogenous L-lactate through gluconeogenesis and three moles C per mole of L-lactate.

2.1.4. The Sum of Precursors for Gluconeogenesis

Due to uncertainties concerning the hepatic capacity for the conversion of glucogenic C to glucose (see discussion), the sum of fluxes of glucogenic C is chosen as the most consistent and comparable level of aggregation for glucose supply. Based on the estimations of the supplies of propionate, glucogenic amino acids, L-lactate and glucose from the portal-drained viscera and the estimations of glycerol, alanine and L-lactate released by protein and adipose tissue, the sum of glucogenic C from these digestive and endogenous precursors is used as the final value for glucose supply in this study.

2.1.5. Glucose Demand

In dairy cows, Bickerstaffe et al. [58] noted that mammary glucose uptake accounts for almost all of glucose turnover, while only small amounts are left for maintenance functions. As mentioned above, the quantification of the glucose demand imposed by the immune system is far from being clear, mainly due to a lack of robust and practical indicators of immunoactivation. However, the amount of glucose consumed by other major glucose-consuming organs, including the mammary gland, protein tissue, the brain and red blood cells, does not rely to a similar extent on the effect of environmental stressors. Therefore, the glucose demands of these tissues serve as a proxy for the overall glucose demand of a dairy cow and were subtracted from overall glucose supply to calculate the amount of glucose reserves available for immunoactivation. These demands were estimated from the levels of MY and L-lactate yield, the metabolic BW and assumptions made on the carcass compositions of individual cows according to the Equations given in Table 1.

Splanchnic Tissues/Portal-Drained Viscera and Liver

The glucose demand of the digestive processes in the duodenum as well as the glucose consumed by other processes in the portal-drained viscera is represented with the estimates of NPA described above. Hence, the glucose demand of these tissues, which accounts for an average of 22% of the whole-body rate of appearance of glucose [44], is integrated in the input part of the model.

Mammary Gland

According to studies on catheterized animals and radioisotope labelling [30,58], the average of 70% of glucose taken up by the mammary gland for lactose synthesis is often used for the calculation of the mammary glucose requirements in lactating dairy cows. However, the ratio varies greatly between and within studies measuring both glucose arteriovenous differences across the mammary gland and lactose yield [59,60,61]. Lemosquet et al. [62] found that the ratio shows a parabolic curve when related to the supply of glucogenic nutrients (rumen propionate + duodenal starch) in the range of 2.12 to 3.95 mmol C/h. In high-yielding dairy cows during early lactation, however, Galindo et al. [31] found up to 88% of mammary glucose flux being used for lactose synthesis during early lactation in control cows and 92% in cows being abomasally infused with casein + amino acids. Based on the latter study, and assuming that the mammary gland will prefer to use energetic nutrients other than glucose for oxidative processes in the mammary gland during early lactation hypoglycaemia, a proportion of 80% of mammary glucose uptake being used for lactose synthesis was assumed in our study (Table 1, Equation (26)).

Muscle/Protein Tissue

Galindo et al. [22] presented data on the whole-body rate of disappearance of glucose and the fluxes across mammary and splanchnic tissues in cannulated Holstein cows (77 ± 13 DIM; ~30 kg ECM/d) and estimated that the remaining residuals (12% of WbRa) would be left for non-splanchnic, non-mammary tissues, mainly the brain (<3%) and muscle tissues (~10%), which resulted in an average of 28.8 mmol glucose/d/kg PT. By multiplying this value by individual daily protein tissue mass (Table 1, Equation (27) and (28)), daily glucose demand of PT was calculated.

Red Blood Cells

According to Basarab et al. [27] and Harvey [34], the glucose consumption of red blood cells (RBC) is approximately 0.014 mmol/d/mL RBC. In the bovine, blood volume averages 10% of BW with a haematocrit of 33% [63]. Thus, the glucose demand of RBC was calculated from individual body weight (Table 1, Equation (30)).

Brain

As mentioned above, Galindo et al. [22] estimated that less than 3% of the whole-body rate of disappearance might be attributed to the brain metabolism. The glucose consumption of the human brain averages 5.6 mg/min per 100 g of brain tissue according to Mergenthaler et al. [32]. For the calculations applied in our study, the mean weight of the brain of dairy cows was estimated at 0.08% of BW according to Ballarin et al. [33] (Table 1, Equation (31)).

The whole-body glucose demand excluding the glucose demand of the immune cells (GLU_OUT, mol C/d) was calculated as the sum of the demand of the mammary gland, protein tissue, red blood cells and the brain, while daily GB (mol C/d) was calculated as the difference between glucose supply and glucose demand.

2.2. Data

2.2.1. Animals

The data were recorded at the Educational and Research Centre for Animal Husbandry, Hofgut Neumuehle, from January 2015 to November 2016 (23 months) for the project OptiKuh (for details see [64]), and the data were made available by the persons responsible. Animals were kept in a loose pen. The data on 157 Holstein cows (201 lactations) from lactations 1 to 11 were analysed. The data sampling encompassed the daily recording of MY (kg), daily recording of individual feed intake, daily BW measurement, weekly measurements of milk ingredients and chemical analysis of feedstuffs (Figure 3,

Table A1. Number of lactations enrolled in the study (N) and means of cow data according to week of lactation.

Week	N	MY	ECM	F %	P %	L %	DMI	BW	EBW	AT	PT	BWloss	EB
1	187	29.0 ± 6.4	27.2 ± 9.2	3.8 ± 0.6	3.2 ± 0.4	4.6 ± 0.2	11.3 ± 2.0	658 ± 79	576 ± 70	124 ± 27	97 ± 12	−2.5 ± 1.9	−49.8 ± 20.9
2	199	34.5 ± 7.4	33.2 ± 8.3	3.8 ± 0.7	3.2 ± 0.4	4.8 ± 0.2	14.4 ± 2.8	646 ± 80	565 ± 69	118 ± 24	96 ± 12	−1.8 ± 1.4	−45.9 ± 23.7
3	198	37.1 ± 7.8	35.7 ± 8.8	3.8 ± 0.7	3.2 ± 0.4	4.8 ± 0.2	17.1 ± 3.3	637 ± 77	556 ± 68	112 ± 23	95 ± 11	−1.2 ± 1.2	−35.0 ± 24.1
4	194	38.6 ± 8.2	37.3 ± 8.9	3.8 ± 0.7	3.1 ± 0.3	4.8 ± 0.1	18.0 ± 3.2	631 ± 76	551 ± 66	108 ± 22	95 ± 11	−0.8 ± 0.9	−32.3 ± 24.7
5	188	40.1 ± 7.6	38.4 ± 8.6	3.8 ± 0.7	3.1 ± 0.4	4.8 ± 0.1	18.9 ± 3.1	627 ± 76	547 ± 66	105 ± 22	95 ± 11	−0.5 ± 0.8	−30.5 ± 20.8
6	186	40.5 ± 7.4	38.8 ± 8.2	3.7 ± 0.6	3.1 ± 0.4	4.8 ± 0.2	19.8 ± 2.9	626 ± 75	547 ± 65	103 ± 21	95 ± 11	−0.4 ± 0.6	−24.4 ± 20.2
7	182	40.6 ± 7.5	38.7 ± 8.1	3.7 ± 0.6	3.1 ± 0.4	4.8 ± 0.2	20.5 ± 3.0	627 ± 74	547 ± 65	102 ± 21	96 ± 11	−0.3 ± 0.5	−19.5 ± 21.3
8	178	40.5 ± 7.2	38.7 ± 8.2	3.7 ± 0.6	3.1 ± 0.4	4.8 ± 0.2	20.8 ± 3.0	628 ± 73	548 ± 64	101 ± 20	96 ± 11	−0.2 ± 0.5	−18.6 ± 20.2
9	176	40.1 ± 6.9	38.3 ± 8.0	3.7 ± 0.7	3.1 ± 0.4	4.8 ± 0.2	21.1 ± 3.0	629 ± 72	549 ± 63	100 ± 20	96 ± 11	−0.2 ± 0.7	−14.1 ± 19.7
10	170	39.4 ± 6.9	37.3 ± 7.9	3.7 ± 0.7	3.1 ± 0.4	4.8 ± 0.1	21.0 ± 3.3	629 ± 72	549 ± 63	100 ± 20	97 ± 11	−0.2 ± 0.4	−13.1 ± 21.7
11	164	39.0 ± 6.8	37.0 ± 7.6	3.7 ± 0.7	3.1 ± 0.3	4.8 ± 0.1	21.6 ± 3.0	631 ± 72	551 ± 63	101 ± 20	97 ± 10	−0.2 ± 0.5	−7.9 ± 20.2
12	160	38.6 ± 7.1	36.4 ± 7.6	3.6 ± 0.6	3.1 ± 0.4	4.8 ± 0.1	21.7 ± 3.0	632 ± 72	552 ± 63	101 ± 21	97 ± 10	−0.2 ± 0.5	−5.4 ± 17.7
13	157	37.9 ± 6.6	35.8 ± 7.5	3.7 ± 0.6	3.1 ± 0.4	4.8 ± 0.1	21.8 ± 3.1	635 ± 72	554 ± 63	101 ± 21	97 ± 11	−0.2 ± 0.5	−2.7 ± 18.6
14	152	37.9 ± 6.6	36.0 ± 7.4	3.7 ± 0.7	3.2 ± 0.4	4.8 ± 0.1	22.1 ± 3.1	636 ± 73	555 ± 64	102 ± 21	98 ± 11	−0.2 ± 0.5	−1.4 ± 19.4
15	136	37.3 ± 7.0	36.0 ± 8.0	3.8 ± 0.6	3.2 ± 0.4	4.8 ± 0.2	21.9 ± 3.2	640 ± 75	559 ± 66	103 ± 21	98 ± 11	−0.3 ± 0.6	−1.3 ± 18.9

AT: the mass of adipose (kg); BW: body weight (kg); BWloss: daily loss of body weight (kg/d); DMI:, dry matter intake (kg/d); EB: daily energy balance (MJ/d); EBW: empty body weight (kg); ECM: energy-corrected milk (kg); F %: percentages of milk fat; L %: percentage of milk lactose; MY: milk yield (kg); P %: percentage of milk protein; PT: the mass of protein tissue (kg).

). Animal data were analysed for the first 105 days in milk. Individual feed intake was recorded daily using feeders equipped with a weighing unit and automatic cow identification (Roughage Intake Control, Insentec B.V., Marknesse, The Netherlands). Cows were milked twice daily using a combination of a herringbone and a side-by-side milking parlour manufactured by GEA Farm Technologies (located in Bönen, Germany). BW was measured automatically after every milking via walk-over scale, and daily values were derived by averaging morning and evening BW. The data from the milking parlour were recorded via the herd management system Dairy Plan C21 (GEA Farm Technologies, Bönen, Germany). Due to different housing during the first DIM, the collection of the cows’ DMI, BW and MY data started between days 1 and 8. Daily MY was recorded at morning and evening milking, and milk was analysed for fat, protein and lactose via infrared analyser (MilkoScan FT-6000, Foss Analytical A/S, Hillerod, Denmark; LKV Rheinland-Pfalz-Saar e.V., Bad Kreuznach, Germany). BCS was assessed once every two weeks.

Measurements of milk ingredients and BCS, as well as values for MY and DMI missing due to technical errors were inter-/extrapolated linearly up to three consecutive days. If more values were missing, the lactations were excluded from the analyses. After the removal of values differing by more than 10 kg EBW (see below for the calculation) from the previous or succeeding day (or up to 50 kg within 5 days), BW was smoothed across the 105 DIM-period for each cow using a cow-specific 5th-degree polynomial function.

For further analyses, daily energy demand for production (ED_P; MJ of NEL/d) and daily energy demand for maintenance (ED_M; MJ of NEL) were calculated according to GfE [51]. Energy intake was calculated from the energy content of the diet (MJ NEL) and individual DMI. The sum of the energy demands was subtracted from the energy intake (MJ NEL = energy content of the diet (MJ NEL/kg) × individual DMI (kg)) to obtain individual energy balance (EB) (Equations (32)–(35); Table 1).

2.2.2. Feeding Rations

All cows were fed a total mixed ration (TMR), consisting of grass silage, corn silage, pressed sugar beet pulp silage, hay, straw, vegetable oils, extruded rape seed, extruded soybean, corn, barley, soybean hulls, molasses, minerals, urea, salt, vegetable oils and calcium carbonate (

Table 2. The means of the proportions of feedstuffs and the contents of energy and nutrients of the total mixed rations fed during the study period.

Feed	DM	DM	ME	OM	CP	CL	CF	ST	S	ADFom	aNDFom
	%	g/kg FM	MJ/kg DM	g/kg DM	g/kg DM	g/kg DM	g/kg DM	g/kg DM	g/kg DM	g/kg DM	g/kg DM
Grass silage, wilted, with additive	25.4 ± 3.5	404.5 ± 13.6	10.6 ± 0.1	905.1 ± 2.2	157.9 ± 1.3	40.5 ± 1.5	247.6 ± 3.7	0.0 (0.0)	14.0 ± 1.6	286.7 ± 8.7	445.7 ± 4.1
Hay, permanent grassland	4.3 ± 1.1	916.0 ± 0.0	7.3 ± 0.0	922.0 ± 0.0	100.0 ± 0.0	18.0 ± 0.0	306.0 ± 0.0	0.0 (0.0)	69.0 ± 0.0	370.0 ± 0.0	233.0 ± 27.8
Corn silage, whole crop	20.9 ± 3.6	330.9 ± 3.5	11.3 ± 0.1	958.7 ± 1.7	78.8 ± 2.1	36.9 ± 0.6	163.2 ± 6.6	345.3 ± 3.5	0.0 ± 0.0	196.5 ± 8.2	656.0 0.0
Pressed sugar beet pulp silage	12.2 ± 4.1	295.2 ± 10.3	11.7 ± 0.2	920.5 ± 23.4	84.4 ± 5.9	13.5 ± 1.7	169.8 ± 8.0	0.0 ± 0.0	12.9 ± 1.0	203.0 ± 7.9	337.0 ± 5.9
Vegetable oils	0.8 ± 0.0	990.0 ± 0.0	28.7 ± 0.0	875.0 ± 0.0	0.0 ± 0.0	840.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	419.1 ± 12.8
Barley straw	1.6 ± 0.7	860.0 ± 0.0	5.9 ± 0.0	954.0 ± 0.0	34.0 ± 0.0	12.0 ± 0.0	380.0 ± 0.0	0.0 ± 0.0	8.0 ± 0.0	440.0 ± 0.0	0.0 ± 0.0
Concentrates *	34.8 ± 0.7	883.1 ± 2.8	13.0 ± 0.2	938.8 ± 12.3	268.0 ± 17.5	38.7 ± 3.8	101.0 ± 25.2	253.6 ± 17.2	69.1 ± 6.0	141.5 ± 34.5	668.0 ± 0.0
Total mixed ration	100.0 ± 0.0	420.3 ± 18.2	11.5 ± 0.2	930.9 ± 4.3	160.7 ± 1.7	33.5 ± 1.1	160.1 ± 2.7	222.1 ± 6.2	58.9 ± 5.9	219.7 ± 8.0	334.6 ± 4.8

ADFom: ash-free acid detergent fibre; aNDFom: ash-free neutral detergent fibre; CF: crude fibre; CL: crude lipids; CP: crude protein; DM: dry matter; ME: metabolizable energy; OM: organic matter; S: sugar; ST: starch; * extruded rape seed, extruded soybean, corn, barley, soybean hulls, molasses, minerals, urea, salt, vegetable oils, calcium carbonate.

). Over the period of two years, the ration composition was adapted on a monthly basis to ensure equal levels of metabolizable energy intake (11.5 ± 0.2 MJ ME/kg DM). Cows were fed ad libitum. TMR and feedstuffs were characterized monthly and/or if ration composition changed by Weende and van Soest analysis for dry matter, organic matter, crude nutrients (protein, fat, fibre, starch and sugar) as well as ash-free acid detergent fibre and ash-free neutral detergent fibre (Table 2). The energy contents of the diets were calculated according to GfE (2001). Subsequently, information on ration composition was assigned to each day and cow.

Further feed values required for the calculation of ruminal and duodenal flows of nutrients with the Systool web application were taken from the INRA feed tables (INRA, 2018,

Table 3. The tabulated values of digestibility and further nutrients required for the analyses in Systool Web according to the INRA Feeding system for ruminants (INRA, 2018).

Feed	OMd	Ed	ED6_N	ED6_S	FA	PF	FLref
	%	%	%	%	g/kg DM	g/kg DM	% BW
Grass silage, wilted, with additive	70	66	77	-	22	85	1.39
Hay, permanent grassland	62	59	68	-	16	0	-
Corn silage, whole crop	73	70	75	70	28	80	1.53
Pressed sugar beet pulp silage	86	82	66	-	5	140	1.28
Vegetable oils	89	100	100	-	840	0	1.63
Barley straw	44	40	68	-	6	0	-
Concentrates *	80	86	52	85	37	0	2.47

Ed: apparent digestibility of gross energy; ED6_N: effective degradability of nitrogen; ED6_S: effective degradability of starch; FA: fatty acid content; FLref: reference feeding level; OMd: organic matter digestibility; PF: fermentation products in silages; * extruded rape seed, extruded soybean, corn, barley, soybean hulls, molasses, minerals, urea, salt, vegetable oils, calcium carbonate.

). The FA content of forages was predicted from their CP content (INRA, 2018).

The Systool web application (Version 1.2, 2017; for details see [46]), which is based on a series of regression equations described in the INRA feeding system for ruminants [20], was used to calculate nutrient fluxes for a total of 26 rations. The digestive processes modelled in the application are modified by species (bovine), feeding level, proportion of concentrate in the diet and rumen protein balance. Therefore, the digestive fluxes were calculated for all rations at six different levels of DMI and five different levels of BW (

Table 4. The categories of body weight (BW) and dry matter intake (DMI) for the calculation of nutritional values via Systool Web for the 26 total mixed rations fed during the study period as well as the numbers and means of the daily measurements in each category.

BW Level (kg)	n	Mean BW (kg)
<500	357	485
500 to 600	6015	561
600 to 700	8007	644
700 to 800	3416	738
>800	300	824
DMI Level (kg)	n	Mean DMI (kg)
<10	650	8.6
10 to 15	2621	12.9
15 to 20	6138	17.7
20 to 25	12,859	20.1
25 to 30	1880	26.5
>30	136	31.4

). Subsequently, the digestive fluxes were attributed to each day of lactation according to individual BW, DMI and the ration fed.

The main pathways of the digestion of glucose and glucogenic metabolites that were modelled in the application included: organic matter degradability, the amount of ruminal fermentable organic matter, the amount of neutral detergent fibre (NDF) digestible in the rumen, the whole-tract non-digestible NDF, the production of volatile fatty acids (VFA) in the rumen, the proportion of propionate in the total VFA production, the degradation of starch in the rumen, duodenum and large intestine, the amount of starch truly digestible in the intestine, the degradation of proteins, the synthesis of microbial proteins and the total flow of proteins digestible in the intestines (

Table 5. The means (SD) of nutritional values of a total of 780 calculations (26 rations for six levels of DMI and five levels of BW, see Table 4) performed with the Systool Web application Version 1.2 (2017). In addition to the levels of DMI and BW, digestive interactions are based on ration composition, the type of the animal (dairy cow) and the existence of urea in the ration.

Intake Level (kgDM/kgBW)	OMD %	RfOM g/kgDM	RdNDF g/kgDM	VFA Prod mol/kgDM	Pro % mol/100 mol	PDI g/kgDM	STdI g/kgDM
<2	73.0 ±1.1	506.5 ±11.5	186.1 ±12.8	4.3 ±0.1	19.4 ±0.7	96.2 ±2.8	24.3 ±2.8
2–3	70.2 ±1.1	469.8 ±11.3	167.3 ±12.7	4.0 ±0.1	21.8 ±0.7	99.0 ±3.0	28.4 ±3.2
3–4	67.2 ±1.1	431.6 ±11.3	147.8 ±12.6	3.6 ±0.1	24.5 ±0.7	102.0 ±3.3	32.6 ±3.6
4–5	64.3 ±1.2	395.2 ±12.8	129.4 ±12.5	3.3 ±0.1	27.0 ±0.8	104.8 ±3.5	36.6 ±4.1
>5	60.1 ±1.5	344.9 ±17.0	105.8 ±10.8	2.9 ±0.1	30.7 ±1.2	109.1 ±3.9	42.5 ±4.9

OMD: organic matter digestibility; PDI: protein truly digestible in the intestines; Pro %: the percentage of propionate production of the total VFA production in the rumen; RdNDF: neutral detergent fibre digestible in the rumen; RfOM: organic matter fermentable in the rumen; STdI: starch digestible in the intestines; VFAProd: volatile fatty acid production.

). Detailed explanations on the development and validation of the databases and equations involved are described by Nozière et al. (2013) [17] and in the INRA feeding system for ruminants (2018) [20].

2.3. Statistical Analyses

The calculation of daily and weekly supplies, demands, GB and EB was performed with Microsoft Excel^®. Weekly means of individual lactations were calculated if three or more measurements were available in the respective week. Zero-order, partial and semi-partial correlations between GB and animal data as well as between GB and ration composition were calculated using the procedure “linear regression” in IBM^® SPSS^®. Separate analyses of these correlations were performed for the first week of lactation (DIM 1 to 7) to evaluate the effects of animal and ration characteristics on GB during the period of the highest metabolic stress.

3. Results

The average weekly MY, ECM, percentages of milk fat (F %), milk protein (P %), milk lactose (L %), DMI, BW, EBW, and mass of AT and PT as well as the BW-loss were within the expected ranges for high-producing dairy cows and the lactation stage investigated (Table A1). Across the 105-day period, MY, ECM yield, F %, P % and L % averaged 38.1 ± 7.2 kg, 36.3 ± 8.1 kg, 3.7 ± 0.6%, 3.1 ± 0.4% and 4.8 ± 0.2%, respectively. DMI, BW, EBW, AT, PT and BW-loss averaged 19.5 ± 3.0 kg, 634 ± 75 kg, 554 ± 65 kg, 105 ± 22 kg, 96 ± 11 kg and −0.6 ± 0.8 kg/d, respectively. The average EB ranged from −49.8 ± 20.9 MJ NEL/d during the first week of lactation (DIM 1 to 7) to −1.3 ± 18.9 at the end of the 105-day period.

The mean GB reached levels close to zero (3.2 ±13.5 mol C per day) during the first week of lactation, followed by a steep increase during the second and third weeks of lactation, and a moderate but continuous increase until 105 DIM (Figure 4,

Table A2. Means of daily glucose balance and daily precursor supplies (mol C/d) according to week of lactation.

		GLUIN from Digestive Precursors					GLUIN from Endogenous Precursors
Week	GB	Total	PRO	gAA	LACdiet	GLU	Total	LACendo	GLY	ALA
1	3.2 ± 13.5	59.0 ± 12.1	38.8 ± 7.0	21.5 ± 4.3	4.7 ± 0.6	−5.9 ± 1.3	19.3 ± 5.9	11.7 ± 1.6	6.8 ± 5.0	0.9 ± 0.6
2	8.8 ± 15.1	78.8 ± 17.6	49.7 ± 10.0	28.0 ± 6.1	5.4 ± 0.8	−4.3 ± 1.5	17.3 ± 4.5	12.0 ± 1.7	5.1 ± 3.9	0.6 ± 0.4
3	19.4 ± 16.3	97.0 ± 20.6	59.7 ± 11.7	34.0 ± 7.1	6.0 ± 0.9	−2.8 ± 1.6	15.4 ± 4.1	11.8 ± 1.6	3.4 ± 3.3	0.3 ± 0.3
4	21.2 ± 17.6	103.8 ± 20.9	63.6 ± 12.0	36.1 ± 7.1	6.2 ± 0.9	−2.1 ± 1.7	13.7 ± 3.4	11.3 ± 1.4	2.4 ± 2.7	0.2 ± 0.2
5	22.1 ± 15.3	109.3 ± 20.1	66.6 ± 11.6	37.9 ± 6.8	6.4 ± 0.9	−1.7 ± 1.5	12.2 ± 2.8	10.6 ± 1.2	1.6 ± 2.3	0.1 ± 0.2
6	26.7 ± 14.8	116.0 ± 18.9	70.5 ± 11.1	40.1 ± 6.3	6.6 ± 0.8	−1.1 ± 1.5	10.9 ± 2.4	9.7 ± 1.1	1.2 ± 2.0	0.1 ± 0.1
7	30.2 ± 16.3	121.0 ± 20.2	73.4 ± 11.9	41.6 ± 6.6	6.8 ± 0.8	−0.7 ± 1.7	9.8 ± 2.0	8.9 ± 1.2	1.0 ± 1.8	0 ± 0
8	31.2 ± 14.8	122.4 ± 19.7	74.1 ± 11.7	42.0 ± 6.5	6.8 ± 0.9	−0.6 ± 1.6	9.1 ± 1.8	8.3 ± 1.1	0.8 ± 1.7	0 ± 0
9	34.8 ± 15.2	125.2 ± 20.1	75.8 ± 11.8	42.9 ± 6.6	6.9 ± 0.8	−0.4 ± 1.7	9.1 ± 2.4	8.2 ± 1.2	0.9 ± 2.4	0 ± 0
10	34.6 ± 18.1	123.8 ± 22.5	75.0 ± 13.1	42.5 ± 7.4	6.9 ± 0.9	−0.5 ± 1.9	8.9 ± 1.8	8.3 ± 1.1	0.7 ± 1.5	0 ± 0
11	40.3 ± 17.2	128.7 ± 21.0	77.9 ± 12.4	44.0 ± 6.8	7.0 ± 0.8	−0.2 ± 1.8	9.0 ± 1.9	8.3 ± 1.1	0.7 ± 1.8	0 ± 0
12	41.8 ± 15.5	129.3 ± 20.6	78.3 ± 12.2	44.2 ± 6.7	7.0 ± 0.8	−0.2 ± 1.8	9.1 ± 1.8	8.4 ± 0.9	0.7 ± 1.8	0 ± 0
13	43.7 ± 15.9	129.7 ± 21.2	78.5 ± 12.5	44.3 ± 6.9	7.1 ± 0.8	−0.2 ± 1.8	9.1 ± 1.8	8.4 ± 0.9	0.7 ± 1.7	0 ± 0
14	46.4 ± 16.8	131.9 ± 21.3	79.8 ± 12.6	45.0 ± 7.0	7.1 ± 0.8	0.0 ± 1.9	9.3 ± 1.9	8.4 ± 0.9	0.9 ± 1.8	0 ± 0
15	46.7 ± 17.4	130.8 ± 22.3	79.3 ± 13.1	44.7 ± 7.3	7.1 ± 0.9	−0.3 ± 1.9	9.6 ± 2.0	8.5 ± 1.0	1.2 ± 1.9	0 ± 0

ALA: alanine; gAA: glucogenic amino acids; GLU: glucose (negative values for GLU indicate a net consumption of duodenal glucose by portal-drained viscera); GLY: glycerol; LACdiet: L-lactate from feed; LACendo: endogenous L-lactate from muscle catabolism; PRO: propionate.

). Accordingly, the highest weekly averages were observed at the end of the monitoring period (14th week of lactation (DIM 92−99): 46.7 ± 17.4 mol C per day).

The results for GB varied greatly between cows (Figure 5). In the first week of lactation, average GB varied from −32.6 to 37.2 mol C/d (without outliers </> 1.5 × interquartile range (IQR)) between cows. Across all days and cows, the rolling mean (−3 days) of GB ranged from −63 to 145 mol C per day.

As an indicator of maximal glucose availability, maxGB was calculated assuming (1) glucose consumption in no tissues other than the mammary gland and (2) maximal NPA of precursors (+standard error, see Table 1). On average, maxGB surpassed GB by 29.9 ± 3.5 mol C per day.

The contributions of the precursors to the overall glucose supply varied according to the stage of lactation (Figure 6, Table A2). The contributions of digestive precursors averaged 54.2 ± 3.7%, 30.7 ± 2.4% and 5.2 ± 0.4% for propionate, amino acids and L-lactate, respectively. Weekly averages of the NPA of glucose were negative throughout the period. The contributions of endogenous precursors ranged from 24.7 ± 7.5% (during the first week of lactation) to a basic level of 6.8 ± 1.4% in periods with less (adipose) or no (protein) tissue mobilization (DIM 43–105; Figure 7). Accordingly, the contributions of endogenous L-lactate (glycerol and alanine) ranged from 13.9 ± 1.9% to 6.2 ± 0.8% (8.0 ± 5.9% to 0.8 ± 1.3% and 1.0 ± 0.7% to 0.0 ± 0.0%).

The mean glucose demand of the mammary gland ranged from 56.4 ± 13.1 mol C per day in the 1st week of lactation to 81.8 ± 14.4 mol C per day in the 7th week. The means (SD) demand imposed by protein tissue, red blood cells and the brain were constant over the 105-day period at 16.6 ± 0.2, 1.1 ± 0.0 and 1.4 ± 0.0 mol C per day, respectively. With an average of 76.1 ± 15.6 mol C/d, GD_MG far exceeded the glucose demand of other tissues (Figure 8).

Assuming the glucose demand in the case of an immune challenge by lipopolysaccharide that is not covered by reductions in milk production (17.7 mol C/d) according to Kvidera et al. [8], 86.3% (72.8, 51.8%) of the weekly mean GB observed in the 1st (2nd and 3rd) week of lactation were below this value. This means that cows would not have been able to meet the challenge without reducing glucose consumption of other tissues. On average, the weekly mean GB reached this value at 20 DIM. However, due to (1) the uncertainty regarding dairy cows’ ability to reduce glucose consumption in peripheral tissues in case of glucose shortage and due to (2) statistical variance in the response equations for the NPA of the precursors, maximum GB was calculated assuming no glucose demand in muscle tissue, RBC and the brain on one side and the maximal NPA of the precursors (+standard error) on the other. For maxGB, 9.8% (7.2 and 3.6%) of the 1st (2nd and 3rd) weeks of lactation were below this threshold.

Semi-partial correlations between GB and main input variables across the whole dataset revealed that DMI had the strongest effect on GB, followed by MY, BW loss, BW, L % and DIM, while lower correlation coefficients were obtained for the contents of metabolizable energy, crude fibre, sugars, ash-free neutral detergent fibre and the percentage of concentrate in the diet (

Table 6. The correlations between glucose balance (mol C), animal and ration characteristics.

	p	Correlation Coefficients DIM 1–105			p	Correlation Coefficients DIM 1 to 7
		Zero Order	Partial	Semi-Partial		Zero Order	Partial	Semi-Partial
DMI	0.000	0.819	0.987	0.849	0.000	0.332	0.956	0.716
MY	0.000	−0.004	−0.964	−0.491	0.000	−0.521	−0.944	−0.628
L %	0.000	−0.028	−0.552	−0.090	<0.001	−0.239	−0.473	−0.118
BW	0.000	0.064	−0.667	−0.121	<0.001	−0.055	−0.448	−0.110
BWLOSS	0.000	0.244	−0.698	−0.132	<0.001	−0.276	−0.839	−0.339
DIM	0.000	0.467	−0.284	−0.040	0.501	−0.285	−0.023	−0.005
LACT	<0.001	0.021	−0.054	−0.007	0.691	−0.009	−0.014	−0.003
ME	0.001	0.039	−0.024	−0.003	<0.001	−0.052	−0.141	−0.031
CP	0.396	−0.097	−0.006	−0.001	0.358	0.115	−0.031	−0.007
CF	0.005	−0.045	0.021	0.003	0.827	0.042	0.007	0.002
ST	0.973	−0.029	0.000	0.000	0.201	−0.036	−0.044	−0.010
CS	<0.001	0.110	−0.048	−0.007	0.489	−0.035	−0.024	−0.005
aNDFom	<0.001	0.090	0.037	0.005	<0.001	−0.055	0.117	0.026
CON %	<0.001	0.129	0.120	0.016	<0.001	0.100	0.226	0.051

aNDFom: amylase-treated, ash-free neutral detergent fibre (g/kg DM); BW: body weight (kg); BWLOSS: body weight loss (kg/d), CF: crude fibre (g/kg DM); CON %: proportion of concentrate in the diet (% of DM); CP: crude protein (g/kg DM); DIM: days in milk; DMI: dry matter intake (kg/d); L %: lactose percentage in milk (%); LACT: lactation number; ME: metabolizable energy content (MJ ME/kg DM); MY: milk yield (kg/d); S: sugar (g/kg DM); ST: starch (g/kg DM).

). The zero-order correlation coefficients between EB and GB varied from r = 0.53 to r = 0.99. between lactations. The median day of the lowest GB (nadir) was 13 with an interquartile range of 24 DIM and was identical with the nadir of EB for 51% of the lactations investigated.

4. Discussion

The aim of this study was to develop a method for the calculation of the GB of dairy cows from individual lactose yield, BW, DMI and nutrient composition of feedstuffs and to evaluate glucose reserves during early lactation. For this purpose, the amounts of digestive glucose, propionate, glucogenic amino acids and L-lactate as well as the amount of glycerol and alanine from adipose and protein tissues mobilization, the amount of endogenous L-lactate and the demand imposed by major glucose-consuming organs within the cow were estimated. Given that mean GB values of a sample of 201 dairy cows reached levels close to zero during the first week of lactation, glucose availability was severely compromised in the immediate postpartum period. However, GB varied considerably across dairy cows within and between lactation stages.

4.1. Estimation of the Supply with Glucogenic C in Cows during Early Lactation

4.1.1. Assessing Digestive and Portal Fluxes

The calculations within the digestive part are based on the estimation of the relevant proportions of nutritional fractions within the rumen and duodenum of the cow from the chemical analyses of feedstuffs, individual DMI and BW via the Systool Web application [46]. Subsequently, the NPA of the digestive precursors were assessed according to Loncke et al. [21] and Martineau et al. [28]. These models exhibited higher R² and lower standard deviation of the residuals compared with other published models of VFA production and absorption [65,66,67,68]. Nonetheless, regression-induced deviations, which over- or underestimate values in specific ranges of the data, have to be considered. In this regard, Loncke et al. [69] pointed out that the models of nutrient NPA applied in this study are based on data encompassing intakes up to 41 g DMI per kg BW per day, while the majority of their data are related to intake levels below 35 g DMI/kgBW/d. The median (IQR) intake level of the cows investigated in this study was 31.5 (9.4) DMI/kgBW/d. Values greater than 41 g DMI/kgBW/d (6% of the data) were not excluded from the calculations, assuming the linearity of the relationship between the ruminal and duodenal fluxes of nutrients and their net transfer to the portal vein. However, limitations in the absorptive capacity of PRO, gAA and GLU are related to increased rumen concentrations of VFA [70], increased duodenal concentrations of amino acids [71] and the capacity of starch hydrolysis [1], respectively. Thus, the flux of digestive precursors across splanchnic tissues is likely to be compromised in cows with high levels of DMI, suggesting in these cases an overestimation of GB. In general, deviations between calculated values and the amount of glucose available to an individual cow emerge from individual variations in digestive and metabolic capacities that are not depicted by the applied regression equations.

4.1.2. The Contribution of Digestive Precursors

Across all (digestive and endogenous) precursors, rumen propionate production was the main source of glucogenic C both in terms of absolute and proportional values. Compared to the data of lactating ruminants presented by Loncke et al. [69], NPA of propionate was somewhat higher (1.53 ± 0.41 vs. 1.16 ± 0.29 mmol/kg BW/h), while the share of propionate in the overall supply with glucogenic C (54.4 ± 9.2%) was slightly lower than maximal contributions to gluconeogenesis (60.9 ± 10.1%). According to the results of several studies [35,72,73,74,75,76,77,78], the proportional contribution of propionate to liver glucose release measured in multicatheterized cows between DIM 1 and 105 averaged 57%, with a tendency to increase as lactation progressed. Hence, the proportion of propionate to glucogenic C calculated in our study is in accordance with these previous results.

The methodological considerations of the approach presented here assume that alanine, glutamate and glycine are the only amino acids that contribute to hepatic gluconeogenesis. Although other amino acids are considered glucogenic, their availability for gluconeogenesis is thought to be severely restricted in lactating dairy cows, as great amounts are required for milk protein synthesis. In fact, it was proposed that the rate of liver uptake is low for amino acids other than alanine, glycine and serine [2]. Bergman and Heitmann [79] found that the rate of conversion of glucogenic C to glucose is low in amino acids other than alanine and glutamine. Young [80], who compiled evidence on the splanchnic amino acid metabolism from single injection and radioisotope tracer studies identified a substantial conversion of C from glutamine to glucose. However, it was postulated that increased fractional contributions of amino acids other than alanine to glucose synthesis are unlikely during the periparturient period [74]. Therefore, we used the value of 45% glucogenic AA in overall AA appearing in the portal blood proposed by Loncke et al. [24], which was derived from publications reporting the NPA of total AA as well as the NPA of alanine, glycine and glutamate. With average weekly contributions of 25.5 ± 5.1% (1st week) to 31.9 ± 5.0% (12th week), the calculated gAA still made up a large share of glucogenic supply. While the amount of amino acids maximally converted to glucose amounted to 37% in deprived cows in the study by Lomax and Baird [81], other values reviewed by Larsen and Kristensen [2] were found in the range between 7.1 to 21% for lactating dairy cows, with a tendency to decrease as lactation progresses.

The values calculated for the NPA of LACdiet (0.14 ± 0.02 mmol/kgBW/h) were lower than the averages for lactating ruminants (0.24 ± 0.08 mmol/kgBW/h) calculated by Loncke et al. [21] and lower than the net portal flux of L-lactate in cows during early or mid-lactation with indwelling catheters in the portal vein [31,82]. In addition to the different utilization rates within the PDV at different stages of lactation, differences between studies and datasets in the content of the predicting variable (ruminal fermentable organic matter) might play a relevant role. Generally, the measurement of NPA of L-L-lactate does not differentiate between the amounts of digestive L-lactate and endogenous L-lactate derived from glucose metabolism in gut tissues [83]. However, although glucose oxidation rates are reduced in muscle tissue during the periparturient period, Loncke et al. [21] did not observe any effect of the physiological status on the regression equation for the prediction of NPA of digestive L-lactate. This suggests a rather constant conversion rate of glucose to L-lactate in gut tissues, providing a continuous downhill gradient for glucose absorption.

Finally, the NPA of duodenal absorbed glucose, which is a result of the simultaneous processes of absorption from the lumen and utilization within the PDV, varied widely with low and high levels of bypass starch of, respectively, −10.2 to +11.2 mol C/d, which is in the range described by Galindo et al. [22,31]. The authors observed a positive NPA of glucose averaging 11.2 mol C/d in the control cows at 77 ± 13 DIM and a negative NPA of glucose during early lactation averaging at −6.3 mol C/d. Although the glucose rate of appearance was shown to increase with enhanced starch intake and glucose absorption from the gut in lactating dairy cows [25,84], limitations to the hepatic conversion of glucogenic C at high levels of bypass starch must be considered. It was supposed that these may emerge from an associated increase in propionate leading to a lower conversion rate of propionate [69] and/or effects associated with the insulin/glucagon ratio [85].

4.1.3. The Contribution of Endogenous Precursors

The amount of endogenous precursors from tissue mobilization and in particular, cori-cycling of L-lactate, was supposed to contribute considerably to the total afferent flux of glucogenic C to the liver [2]. With 19.3 ± 5.9 mol of glucogenic C, endogenous sources contributed more than 25% to the overall glucose supply in the first week of lactation in our study. When expressed in grams of glucose, this value is in close agreement with a rough estimate given by Drackley et al. [86], who calculated that glucose supply from feed intake may fall short of glucose demands by 500 g/d (16.7 mol C) in the immediate postpartum period in high-yielding dairy cows.

The shift in precursor supply is also reflected by a shift in mRNA-expression and in the activity of the enzymes related to gluconeogenesis. The increase in pyruvate carboxylase after parturition increases the entry of L-lactate and alanine via the Krebs cycle [1,2], while glycerol enters the glucogenic pathway through the action of glycerol kinase [87]. The averages of GLY (0.95 ± 1.65 mmol/kg BW/d) and ALA (0.07 ± 0.17 mmol/kg BW/d) calculated from the weight losses of cows between 1 and 105 DIM are in the same range, with averages (1.07 ± 0.64 and 0.06 ± 0.04 mmol/d/kg BW, respectively) calculated from lactating cows in NEB between DIM 11 to 240 by Loncke et al. [69], when expressed in the same unit. The proportional contribution of GLY and ALA to overall glucogenic supply was low, with the exception of weeks 1 to 3, where GLY contributed 8 ± 6%, 5 ± 4% and 3 ± 3%. Accordingly, the loss of BW correlated more strongly with GB during the first week of lactation compared with the whole 105-day period. In studies with multicatheterized cows, GLY contributed a maximum of 4.9% to hepatic glucose release in fed cows during their first week of lactation [2]. Galindo et al. [31] measured similar values (4.6%) at 5 DIM, while 8.1% (17 and 22%) were measured by Lomax and Baird [81] in cows deprived for several days. According to the model of Guo et al. [14], glycerol provided between 12 and 17% of the glucose demand.

The calculation of the proportion of L-lactate derived from anaerobic glycolysis within muscle tissues contributed between 13.9 ± 1.9% in the first week and 6.0 ± 0.7% (week 15) to overall glucogenic C. The dynamics of L-L-lactate assumed in our study are in accordance with Reynolds et al. [73], Benson et al. [76] and Larsen and Kristensen [2]. Similar to our results, the greatest contributions of alanine and L-lactate to the net hepatic glucose release were reported during the first week of lactation [35]. Physiologically, this is reflected in both a higher net hepatic uptake of L-L-lactate during this period [74,76] and a reduced irreversible loss rate of glucose (excluding loss in lactose) in the peripheral tissues [2]. However, the metabolic and nutritional influences on the ratio between glucose being oxidized to CO₂ and glucose being converted to L-lactate are worth being investigated in more detail to advance metabolic models of the nutrient partitioning in dairy cows.

4.1.4. Hepatic Turnover of Glucogenic Carbon

The calculation of GB presented in this study assumes that the NPAs of PRO, gAA, LACdiet and GLU represent the general gluconeogenic potential of the cows. It does not account for the capacity of the hepatic conversion of precursors to glucose. In fact, the hepatic conversion of digestive precursors to glucose was shown to not follow first-order kinetics but rather to follow a curvilinear course in ruminants [69]. The same authors concluded from the results of a meta-analysis that the total digestive precursors for gluconeogenesis accounted for only 63% of the net hepatic glucose release in dairy cattle. This estimation is in accordance with those of Larsen and Kristensen [2], who reviewed the maximal contributions of precursors to gluconeogenesis and summarized that in most studies, precursor supply was not sufficient to explain the hepatic glucose release at various stages of lactation. At first, it seems that these results are anything but self-evident, because the net flux and true flux of glucose across the liver were found to be nearly identical, and thus, a minimal hepatic glucose consumption from plasma must be assumed [31]. Yet the great flexibility of hepatic metabolism through the storage and release or conversion of glucose, the conversion of glucogenic C to glucose in other tissues and metabolites other than those investigated (see Section 4.1.5) might partially explain the lack of glucogenic C relative to liver glucose release. Due to the uncertainties associated with the individual hepatic capacity for gluconeogenesis, we assume that the supply with glucogenic C represents the most consistent measure for the calculation of GB. However, because high liver fat contents, as typically observed in dairy cows, particularly in early lactation, were shown to impede hepatic gluconeogenic capacity [88], glucose availability and therewith, glucose balance, as calculated in this study, may be overestimated.

4.1.5. Other Sources of Glucogenic C

In our study, weekly GB across cows averaged positive values throughout the period investigated, but GB was negative for an average of 10.8 ± 8.4 days per cow. If no glucose consumption was assumed in tissues other than the mammary gland and no maximal standard errors of the NPA of precursors (maxGB) were assumed, this value decreased to 0.9 ± 1.9 days per lactation. Physiological explanations for cows in negative GB according to the concept presented in this study include the depletion of glycogen stores, the withdrawal of glucose from the plasma pool or other sources of glucogenic C not accounted for. Regarding the plasma pool of glucose, the timepoint of the lowest average GB concurs with the drop in plasma glucose described for the immediate postpartum period in high-yielding dairy cows [89,90]. However, even in the high range of values, the decrease in plasma glucose would hardly surpass 2 mmol/L and thus would provide less than 1 mol C in dairy cows of 600 kg. Moreover, the depletion of glycogen stores in the liver and muscle of dairy cows was not included in our calculations, as it was assumed that these play a minor role in the provision of glucogenic C during early lactation. In contrast with monogastric animals, where the synthesis of glycogen is needed to overcome postprandial hyperglycaemia, a constant supply of glucogenic precursors from rumen fermentation compensates for discontinuous feed intake in ruminants. Accordingly, ruminants store only small amounts of glucose, mainly in the liver and in muscle tissue. This may play a role in maintaining the short-term homeostasis of blood glucose (liver glycogen) as well as for a rapid supply of energy for exercise (muscle glycogen). The hepatic glycogen pool comprises maximally 20 mol glucogenic C before calving [91,92]. Similarly, the amount of glycogen stored in muscle tissue (less than 0.4% of wet weight; [52,93]) would amount to ~20 mol C for maximal weight of protein tissue observed in our data (132 kg). However, both hepatic [12,94,95] and muscle glycogen [52] were shown to be depleted almost completely immediately after calving. Hammon et al. [89] showed significantly lower glycogen stores in the livers of cows with high vs. low fat content. Interestingly, Galvão et al. [96] also found reduced glycogen concentrations in polymorphonuclear neutrophils at calving, which are associated with the occurrence of subclinical endometritis and metritis. At reference levels of blood volume and neutrophil counts, the estimated glycogen content of neutrophils (~30 mg/10⁹ cells [97]) would be less than one mol C in cows of 600 kg and would not be sufficient to cover the glucose demand of an activated immune system [8]. Although these have been thought to be of minor quantitative importance, the pyruvate from catabolism of amino acids [75], the formation of D-Lactate in the gastrointestinal tract, the pyrimidine bases of deoxyribose and ribose from the breakdown of ruminal microbes and propionyl-CoA from the beta-oxidation of C15 and C17 fatty acids [2] must be considered to have a share in providing C for gluconeogenesis. Moreover, glucogenic C provided by the involution of the uterus might not be negligible, particularly in cows during early lactation [69].

4.2. Glucose Demand

4.2.1. Quantitative Glucose Metabolism in Non-Mammary Tissues

Although the mammary gland consumes major quantities of glucose in high-producing dairy cows, the amount of glucose consumed by non-mammary tissues, i.e., the glucose requirement for maintenance functions, (~200 g/d) that was suggested earlier [26] is thought to be a “bare-bone minimum” [25]. Accordingly, Baldwin reported a turnover rate of glucose in ruminants under basal conditions ranging from 0.03 to 0.05 mol/d/kg BW^0.75, which is equal to 655 g to 1092 g of glucose for a cow of 600 kg [98]. Similar values were obtained for postpartum dairy cows within the model of Guo et al. [14], while the proportion of whole-body glucose flux not being used by the udder is in the range of 20 to 50% [22,31,84,99], indicating a substantial demand of glucose for tissues other than the mammary gland, even in high-producing dairy cows. After the estimation of the organ-specific glucose demands of individual cows according to their BW, EBW and body composition, the results obtained in our study suggest a substantial glucose demand by muscle tissue, while the glucose demands of red blood cells and the brain seemed to be fairly low (1.1 ± 0.2 and 1.4 ± 0.2 mol C/d, respectively). However, a constant supply of glucose to erythrocytes is crucial because glucose oxidation through the pentose phosphate pathway and glycolysis is usually the only source of energy for these cells [100]. The small variations in individual values reported in this study, being attributable only to differences in BW, is thus thought to reflect the physiological function of glucose supply to erythrocytes. Accordingly, the enzyme activities of glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase in erythrocytes did not vary according to physiological status in high-producing dairy cows [101]. Similarly, the low variation in glucose utilization by the brain, which was estimated from data presented by Ballarin et al. [33] can be assumed to be due to the importance of glucose for neurometabolism and low variations in the weight of the brain of dairy cows (<1%). The low brain weight of the domestic Bos taurus also leads to a much lower proportional glucose consumption compared with humans, where at least 20% of the body’s energy consumption is attributed to the brain [102].

The calculation of the glucose demands of protein tissue performed in this study is based on the work of Galindo et al. [22]. The authors assumed that the difference between the whole-body rate of appearance and the glucose consumed by the mammary and splanchnic tissues could be largely traced to muscle glucose metabolism. However, the calculated values are lower than the average values reported for the in vitro muscle tissue of steers [54] but higher than the values measured in the hind-limb muscle tissue of Merino ewes [103]. Glucose utilization by the muscle was shown to be tightly regulated through plasma insulin, insulin sensitivity and receptor expression [52]. Thus, reductions not only in the amount of glucose oxidation but also in the amount of glucose uptake to muscle cells must be assumed and should be investigated in more detail. Guo et al. [14] as well as Martin and Sauvant [18] presented interesting approaches in this respect. In fact, peripheral glucose sparing might be more exhaustive during early lactation. A 40% depression in glucose transporter (GLUT) 4 expression and protein levels within the first month of lactation was shown [52]. The estimation of glucose demand of protein tissue performed in our study is based on data on high-producing dairy cows at 77 ± 13 DIM [22]. Due to similar metabolic adaptations between lactation and inflammation [11], reductions in the glucose demand of protein tissue may thus also play a role in providing glucose for immunoactivation. Additional glucogenic C both in case of high mammary demand and for immune processes may be provided by splanchnic tissues, which on average extract 22% of whole-body utilization [44]. However, the glucose consumed by these tissues is integrated in the digestive part of the GB calculation presented in this study and could not be calculated separately. This is because the net portal appearance of glucose is a result of both glucose absorption and glucose consumption during digestive processes. While acetate is supposed to be the major source of C for lipogenesis in ruminants [104], glucose was shown to provide 1–10% of the acetyl units in subcutaneous adipose tissue and 50–75% in the intramuscular fat depots of Angus steers [105]. Due to these low values in adipose tissues, and because the cows we investigated were in early lactation, no glucose consumption was assumed for adipose tissues.

4.2.2. Quantitative Glucose Metabolism of the Mammary Gland

The glucose demand of the mammary gland of high-producing dairy cows overrides the glucose demand of other tissues by far. Linear correlations between glucose uptake and milk or lactose yield are described in the literature [60,106,107,108,109]. In contrast with previous work in the field, however, we assumed a ratio of 80% glucose uptake to the MG being used for lactose synthesis during the lactation stage we investigated (see method section). This higher value leading to lower overall values for mammary glucose demand was chosen because Galindo et al. [31] found a higher ratio of mammary lactose output to glucose uptake in cows during early lactation. Accordingly, homeorhetic changes occurring during postpartum hypoglycaemia are likely to favour the use of energetic nutrients other than glucose for oxidative processes in the mammary gland. In fact, the dynamics of mammary glucose uptake and consumption in dairy cows are tightly interrelated with the uptake and utilization of other energetic metabolites and precursors. A high degree of metabolic flexibility was observed by Amaral-Phillips et al. [26] and confirmed by Lemosquet et al. [110]. The latter showed that variations in the levels of milk solids are not explained by increases in whole-body or mammary glucose availability from plasma but are a result of the mammary partition of acetate, glucose and glucogenic C between oxidation, lactose, fat and protein synthesis. Thus, the partitioning of mammary nutrients is likely to adapt to the nutritional status of the animal [111]. GLUT1, which is a non-insulin dependent glucose transporter on the apical and basal membrane of bovine mammary epithelial cells [112] is a major factor contributing to a constant rate of glucose uptake to the mammary glands of dairy cows. GLUT1 expression is regulated through local concentrations of growth hormone (GH) releasing factor and local hypoxia in response to the high metabolic activity of the mammary gland [68,113]. Therefore, the rate of glucose uptake to the bovine mammary gland remains fairly constant across a wide range of plasma glucose concentrations [114].

4.3. Glucose Balance as a Measure of the Cow’s Ability to Respond to Immune Challenges

In the past, the quantification of the availability of energy and nutrients at the individual animal level to ensure both the health and the productivity of individual cows focused on overall energy balance. Negative energy balance (NEB) has been related to metabolic disbalances and the occurrence of disease in dairy cows [115]. However, questions remain about the differences between healthy and diseased cows at similar levels of NEB [116,117]. In fact, energy balance (EB) is a highly aggregated measure and information is lacking on their nutrient-specific trade-offs as well as on their compensation through mobilization or through metabolic flexibility in nutrient utilization. This is supported by the heterogeneity in the plasma concentrations of energy metabolites between cows suffering from similar levels of NEB [118,119]. Accordingly, the quantification of separate nutrient fluxes at the whole-animal level, i.e., “moving from joules to moles of molecules or groups of molecules” was claimed by Ortigues-Marty et al. [120] to advance nutritional concepts related to high-producing animals.

GB highlights the immunometabolic bottleneck associated with limited glucose availability during the postpartum period, when mammary and immune cells simultaneously impose great demands for milk production as well as uterine reorganization and other stressors associated with, e.g., calving and regrouping, while feed intake lags [11,116]. Because immune functionality is not only essential for pathogen elimination but also is part of the coordinated reaction of the organism to all kinds of stressors [121] and because glucose plays a pivotal role for immune functions, GB is supposed to specifically reflect the cow’s ability to adapt to immune challenges [11], compared with the estimation of EB only. However, EB and GB correlated well for most cows and days investigated in our study. This might be related to the homogeneity of the rations fed, which were fairly constant regarding the energy, fibre and protein contents and ration composition, resulting in few differences between the intake levels of glucogenic C and energy. Accordingly, the highest semi-partial correlations were observed between DMI and GB, while correlation coefficients for diet characteristics were low. Apart from variation related to the metabolic part of GB calculation, correlation between EB and GB is thus expected to vary between rations differing more than those fed in our study.

Given that the glucose receptors GLUT1 and GLUT3 of immune cells were found to be negatively correlated to lactose yield [122], metabolic conflicts are expected to arise in situations where both mammary and immune functions impose great demands while the supply of glucose is limited. This is in line with Kvidera et al. [8], who calculated a total glucose deficit of 1553 g during the 12 h following LPS-injection, which is equal to 103.5 mol C on a daily basis. The glucose deficit described by the authors is composed of 530 g (17.7 mol C/d) of glucose infusion required to maintain euglycemia, while the remaining glucose is derived from reductions in the cow’s milk production. The results of our study suggest a depletion of glucose reserves during the first weeks of lactation and thus indicate a limited availability of glucose for regulatory processes such as immune defence during this period. According to the dominant role of mammary glucose demand for whole-body glucose metabolism, particularly during early lactation, the correlations between MY and GB were found to be greatest in the immediate postpartum period (DIM 1 to 7, Table 6) in our study. This is in line with Gross et al. [90], who observed that cows were less able to cut down milk synthesis during negative energy balance in early lactation, compared with a similar challenge in later lactation phases. Although metabolic prioritization of the mammary gland is part of the physiological adaptation to lactation in all mammals, it was hypothesized that homeorhetic trade-offs between self-sustainment and the survival of the offspring are dysbalanced during early lactation in cows being bred intensively for high milk yields [117]. For instance, changes in the somatotropic axis related to decreased glucose availability for immune cells, such as hypoinsulinemia [123] and hepatic GH-resistance leading to decreased levels of IGF-1 and its stimulating effect on immune cells [124], are more profound in cows with high genetic merit for milk production [125,126,127].

5. Conclusions

The methodology presented in this article integrates previous research work on quantitative glucose metabolism in dairy cows, providing a consistent physiologic model for the estimation of glucose reserves. Due to the specific role of glucose for both mammary and immune functions, this parameter highlights a central trade-off of quantitative energy metabolism during the challenging postpartum period. GB might help to advance scientific knowledge as well as data-driven management measures associated with cow-specific glucose requirements to ensure sufficient availability for both regulatory and productive functions. The estimation of glucose reserves for 201 lactations in high-producing dairy cows fed energy-dense diets showed that most cows are facing glucose shortage in the immediate postpartum period, while it is uncertain to what degree reductions in glucose demands of other tissues (mammary and muscle) or other sources of glucogenic C provide additional glucogenic C in case of immunoactivation. Further investigations should focus on the relationships between GB and different feeding regimes, management practices and production outcomes, including the risk for the development of production diseases.