TNFα Altered Inflammatory Responses, Impaired Health and Productivity, but Did Not Affect Glucose or Lipid Metabolism in Early-Lactation Dairy Cows

Inflammation may be a major contributing factor to peripartum metabolic disorders in dairy cattle. We tested whether administering an inflammatory cytokine, recombinant bovine tumor necrosis factor-α (rbTNFα), affects milk production, metabolism, and health during this period. Thirty-three Holstein cows (9 primiparous and 24 multiparous) were randomly assigned to 1 of 3 treatments at parturition. Treatments were 0 (Control), 1.5, or 3.0 µg/kg body weight rbTNFα, which were administered once daily by subcutaneous injection for the first 7 days of lactation. Statistical contrasts were used to evaluate the treatment and dose effects of rbTNFα administration. Plasma TNFα concentrations at 16 h post-administration tended to be increased (P<0.10) by rbTNFα administration, but no dose effect (P>0.10) was detected; rbTNFα treatments increased (P<0.01) concentrations of plasma haptoglobin. Most plasma eicosanoids were not affected (P>0.10) by rbTNFα administration, but 6 out of 16 measured eicosanoids changed (P<0.05) over the first week of lactation, reflecting elevated inflammatory mediators in the days immediately following parturition. Dry matter and water intake, milk yield, and milk fat and protein yields were all decreased (P<0.05) by rbTNFα treatments by 15 to 18%. Concentrations of plasma glucose, insulin, β-hydroxybutyrate, non-esterified fatty acids, triglyceride, 3-methylhistidine, and liver triglyceride were unaffected (P>0.10) by rbTNFα treatment. Glucose turnover rate was unaffected (P = 0.18) by rbTNFα administration. The higher dose of rbTNFα tended to increase the risk of cows developing one or more health disorders (P = 0.08). Taken together, these results indicate that administration of rbTNFα daily for the first 7 days of lactation altered inflammatory responses, impaired milk production and health, but did not significantly affect liver triglyceride accumulation or nutrient metabolism in dairy cows.


Introduction
The periparturient period in dairy cows is characterized by substantial metabolic stress, endocrine changes, depressed feed intake, and negative energy balance [1]. In response to these changes, dairy cows mobilize adipose tissue triglyceride (TG), leading to elevated non-esterified fatty acid (NEFA) concentrations in the blood. The high flux of blood NEFA to the liver often exceeds the capacity of the liver to completely oxidize NEFA to CO 2 , resulting in partial oxidation to form ketones or esterification to form TG within hepatocytes [2]. Excessive production of ketones often leads to ketosis, which is characterized by elevated blood concentrations of b-hydroxybutyrate (BHBA), depressed appetite, and decreased milk production [3]. Furthermore, because ruminants are inefficient at exporting TG from the liver, excessive TG accumulation or fatty liver occurs. Ketosis and fatty liver affect up to 50% of dairy cows, compromising production, health, and reproduction [3,4]. Gluconeogenesis is of great importance at all times in ruminants, providing up to 90% of the necessary glucose [5]. Meeting glucose needs can be a tremendous metabolic challenge for early-lactation dairy cows, which often experience inadequate feed intake. Although lipid and glucose metabolism in periparturient cows has been extensively studied, the mechanisms underlying the development of these metabolic disorders are not fully understood.
Inflammatory mediators play critical roles in immunity and metabolism, and recent research has suggested that inflammation is involved in metabolic disorders as well. For example, obesity is associated with a chronic low-grade inflammatory state in multiple metabolic tissues including adipose, liver, muscle, pancreas, and brain [6]. Compared with lean controls, obese mice had increased levels of adipose tissue tumor necrosis factor-a (TNFa), a potent cytokine capable of triggering inflammatory responses [7]. In dairy cows, fatty liver was associated with increased plasma inflammation markers, including haptoglobin and serum amyloid A [8]. Although a recent study reported that plasma concentrations of TNFa in periparturient cows were decreased postpartum compared with prepartum [9], elevated activity of serum TNFa was observed in cows with moderate to severe fatty liver [10]. In light of these findings, we hypothesized that chronic low-grade inflammation may be a mechanism underlying bovine fatty liver. In fact, we found that low-level administration of the recombinant bovine TNFa (rbTNFa) for 7 d promoted liver inflammation and TG accumulation [11]. However, late-lactation animals were used in that work, and it is unclear if this model is applicable to cows in the first week of lactation, when fatty liver naturally occurs. Beyond direct promotion of fatty liver, hepatic inflammation may also affect glucose production. Bradford et al. [11] reported that rbTNFa administration reduced liver gluconeogenic gene expression in late-lactation cows. If inflammatory cytokines impair glucose production in earlylactation animals, the resulting hypoglycemia would likely increase adipose TG mobilization and metabolic disorders.
We hypothesized that metabolic stress in early-lactation cows is exacerbated by inflammatory challenge, thereby adversely affecting production and health. Therefore, the objective of this study was to determine whether exogenous rbTNFa administration promotes inflammation and TG accumulation, impairs gluconeogenesis, alters lipid metabolism, and affects milk production and health status in early-lactation dairy cows. A greater understanding of the pathological impact of inflammatory pathways in these animals may improve our ability to prevent metabolic disorders and increase production.

Materials and Methods
The Kansas State University Institutional Animal Care and Use Committee approved all experimental procedures.

Design and Treatments
A recombinant expression vector encoding the secreted form of bovine TNFa (Entrez Protein Accession AAB84086.1, region 77-233) was expressed in Escherichia coli and purified by a commercial laboratory (GenScript Corp., Piscataway, NJ). Purity of the isolated rbTNFa was verified by SDS-PAGE, and endotoxin was removed using polymyxin B until contamination was less than,1 EU/mg protein. Thirty-three Holstein cows (9 primiparous and 24 multiparous; body condition score: 3.2660.31, body weight 741683 kg; mean 6 SD) were randomly assigned to 1 of 3 treatments at parturition. Treatments were balanced within parity, and dystocia did not differ by treatment (calving difficulty scores: 1.36, 1.18, and 1.1860.5 for Control, 1.5 mg/kg and 3.0 mg/kg TNF, respectively; 1 to 5 scale, with 1 = no difficulty). Treatments (dose per d) were no rbTNFa (Control; 10% glycerol in saline), 1.5 mg rbTNFa/kg body weight in saline with 10% glycerol, and 3.0 mg rbTNFa/kg body weight in saline with 10% glycerol. These doses are slightly less and slightly more, respectively, than the 2.0 mg/kg dose that was previously shown to alter hepatic nutrient metabolism in lactating cows [10]. Glycerol was included in the solution to improve solubility of the rbTNFa protein. Cows were injected subcutaneously once daily (1600 h) for the first 7 d of lactation. Cows were milked 3 times daily (0400, 1200, and 2000 h) in a milking parlor and fed twice daily (0800 and 1600 h) for ad libitum intake of a diet formulated to meet National Research Council (NRC) [12] nutrient requirements. Ingredient and nutrient composition of the diet are shown in Table 1.

Sample and Data Collection and Analysis
Feed Intake, Milk Production, Energy Balance, and Health Monitoring. During the 7-d experimental period, intake of feed and water and milk yield were recorded daily. Milk samples were collected from d 4 to 7 of treatment at each of the 3 milkings for analysis of milk components. Samples were analyzed for concentrations of fat, true protein, lactose (B-2000 Infrared Analyzer; Bentley Instruments Inc., Chaska, MN), urea nitrogen (MUN spectrophotometer, Bentley Instruments Inc.), and somatic cells (SCC 500, Bentley Instruments Inc., Heart of America DHIA, Manhattan, KS). Somatic cell linear score was calculated as described by Shook [13]: log 2 (somatic cell count/100) +3. Energy balance was calculated for each cow using the following equation from NRC [12]: energy balance = net energy intake 2 (net energy of maintenance + net energy of lactation). Net energy intake = dry matter intake 6 net energy density of the diet [12]; net energy of maintenance = 0.086 body weight 0.75 ; net energy of lactation = [(0.09296 fat%)+(0.05476protein%)+(0.03956 lac-tose%)] 6 milk yield. In all equations, energy is expressed in Mcal and mass in kg.
Cows were monitored daily for health status. Ketosis was recorded when urine acetoacetate exceeded 80 mg/dL on any day or 40 mg/dL for 2 consecutive days (Ketostix; Bayer Corp. Diagnostics Division, Elkhart, IN). Fever was diagnosed when a cow had a rectal temperature greater than 39.4uC. Other disorders or diseases, including mastitis, metritis, and milk fever, were diagnosed according to the guidelines by Kelton et al. [14]. Plasma Metabolites and Hormones. Blood samples were collected from coccygeal vessels daily (0800 h) in 2 tubes, one containing potassium EDTA and the other containing potassium oxalate with sodium fluoride as a glycolytic inhibitor (Vacutainer; Becton Dickinson, Franklin Lakes, NJ). Blood was centrifuged at 2,0006 g for 10 min immediately after sample collection, and plasma was frozen at 220uC until further analysis. Plasma samples collected daily during the 7-d treatment period were analyzed for NEFA (NEFA-HR; Wako Chemicals USA Inc., Richmond, VA), glucose (kit #439-90901; Wako Chemicals USA Inc.), BHBA (kit #H7587-58; Pointe Scientific Inc., Canton, MI), and TG (#10010303; Cayman Chemical, Ann Arbor, MI). Plasma samples collected on d 0, 3, 5, and 7 of treatment were analyzed for TNFa (an ELISA method described by Farney et al. [15]), insulin (ELISA kit #10-1201-01; Mercodia AB, Uppsala, Sweden), and haptoglobin (ELISA kit #2410-7; Life Diagnostics, West Chester, PA). Plasma samples collected on d 7 of treatment were analyzed for 3-methylhistidine. Samples were prepared by mixing 500 mL of plasma with 500 mL of Seraprep (Pickering Labs, Mountain View, CA) and frozen overnight. The deproteinized plasma was then thawed, vortexed, and centrifuged at 17,0006g for 10 min. The resulting supernatant was analyzed for 3-methylhistidine using Li + cation exchange chromatography and detection by fluorimetry following post column derivitization with o-phthalaldehyde. The HPLC column and all reagents and eluents were purchased from Pickering Labs (Mountain View, CA). Column flow rate was 0.375 mL/min and column temperature was maintained at 36uC. Injection volumn was 10 mL. Eluent 1 (Li357) was run for 5 min then eluent 2 (Li750) for 15 min and eluent 3 (RG003) for 5 min. The column was then re-equilibrated back to eluent 1 for 20 min. Post column derivitization was performed using a post column mixing T at a flow rate of 0.375 mL/min using 950 mL OPA diluent (OD104) containing 0.3 g o-phthalaldehyde, 2.0 g Thiofluor, and 3 mL 30% Brij 35. The fluorimeter excitation was set to 330 nm and emission to 465 nm.
Quantitation of eicosanoid concentrations was performed with Waters Empower 2 software (Waters Corp.). A linear calibration curve with 5 points (R 2 .0.99) was generated for each eicosanoid with standards and internal standards purchased from Cayman Chemical Co. (Ann Arbor, MI). The curves ranged from 0.0024 to 2.38 ng/mL. Empower 2 identified the sample peak by matching its retention time with the standard. A response was calculated for each matched peak by dividing the sample peak's response by its internal standard's response. This response was multiplied by the concentration of the internal standard for each analyte. The concentration of each analyte was calculated using the response peak and injection volume.
Glucose Turnover Rate. After daily blood sampling on d 5 of treatment, jugular vein catheters were placed and at least 18 h of recovery were allowed prior to sample collection through the catheters. On treatment d 7, cows were given a glucose bolus containing U-13 C-glucose (99 atom%, Sigma-Aldrich Co., St. Louis, MO) through jugular catheters [17]. Jugular blood samples were collected 10 min prior to infusion and at 10, 20, 30, 40, 50, 60, 90, and 120 min post-infusion. Each cow received approximately 1 g of U-13 C-glucose in 50 mL of sterile saline; syringes were weighed immediately before and after infusion to determine the exact amount administered. Catheters were flushed with a sterile solution of 3.5% sodium citrate after the labeled glucose bolus (20 mL) and after each blood sample collection (5 mL). Plasma glucose concentration was quantified to verify that steadystate conditions were met.
Plasma samples collected for the glucose turnover assay were analyzed for U-13 C-glucose enrichment (Metabolic Solutions, Inc., Nashua, NH). Glucose was extracted and converted to aldonitrile pentaacetate derivative [18], and negative chemical ionization GC/MS (Hewlett-Packard 5890; Agilent Technologies, Santa Clara, CA) was used to analyze derivatized samples. The isotopic composition of the glucose was determined by monitoring unlabeled (M+0: m/z = 328) versus U-13 C-labeled (M+6: m/ z = 334) glucose derivatives. This approach, as opposed to oxidation of glucose and measurement of CO 2 enrichment, ensures that results are not biased by carbon recycling via the Cori cycle. Turnover rate of plasma glucose was calculated from the disappearance curve for U-13 C-glucose. Enrichment of plasma glucose for each animal was fitted to an exponential decay curve according to the following equation: E t = E 0 6 e 2kt , where t = time relative to infusion (min), E t = enrichment of plasma glucose (U-13 C-glucose: unlabeled glucose ratio) at time t, E 0 = enrichment at time t = 0, and k = rate constant (min 21 ). After using the best-fit equations to determine k and E 0 , the total glucose pool was calculated by the following equation: G = M4E 0 , where G = total glucose pool (g) and M = mass of tracer infused (g). Plasma glucose turnover rate (GTR, g/min) was calculated according to the equation GTR = G6k. Samples collected 10 min prior to infusion of U-13 C-glucose were also analyzed to verify the lack of natural occurrence of the M+6 isotopomer.
Liver TG. At the end of treatment d 7, liver samples were collected by percutaneous biopsy as described by Morey et al. [19]. For analysis of TG content, approximately 20 mg of liver was placed into 500 mL of chilled phosphate buffered saline (pH 7.4) and homogenized. The homogenate was centrifuged at 2,0006 g for 10 min at 4uC, and 100 mL of the supernatant was then removed for free glycerol and total protein analyses. Triglyceride content was measured using a method adapted from Starke et al. [20]. The remaining liver homogenate was incubated with 100 mL of lipase (porcine pancreatic lipase; MP Biomedicals, Solon, OH) for 16 h at 37uC, and glycerol content was then determined by an enzymatic glycerol phosphate oxidase method (#F6428, Sigma-Aldrich Co.). Triglyceride content was calculated based on the difference between glycerol concentrations before and after lipase digestion. Total protein content of the original homogenate was analyzed by a Coomassie blue [21] colorimetric method (kit #23236; Thermo Scientific Pierce, Rockford, IL). To avoid potential bias introduced by differences in moisture content of liver samples, liver TG concentration was normalized by protein concentration, which is unaltered in fatty liver [22].
Transcript Abundance. Total RNA was extracted from liver tissue using a commercial kit (RNeasy Lipid Tissue Mini Kit; Qiagen Inc., Valencia, CA) according to the manufacturer's instructions. Two micrograms of total RNA was used as the template for the reverse transcriptase reaction using random primers (High-Capacity cDNA RT Kit; Applied Biosystems, Foster City, CA). Quantitative real-time PCR was performed in duplicate on 96-well plates with 5% of the cDNA product in the presence of 200 nM gene-specific forward and reverse primers with real-time SYBR green fluorescent detection using SYBR Green Premix reagent (7500 Fast Real-Time PCR System, Applied Biosystems). Primers were designed (www.ncbi.nlm.nih. gov/tools/primer-blast/) using GenBank sequences ( Table 2). Data were recorded and analyzed with Sequence Detector software (Applied Biosystems). All sample values were normalized against ribosomal protein subunit 9 (RPS9) values [23], and relative gene expression was quantified by using the 2 2DCt method. Treatment did not influence the Ct value for RPS9 (P = 0.56), suggesting that it served as a valid control gene.
Western Blot. Relative protein abundance of IkBa, c-Jun, phosphorylated c-Jun, and TNFa in liver samples was determined by Western blot. Liver samples (,20 mg) were homogenized at 4uC in RIPA lysis buffer containing a broad-spectrum protease inhibitor cocktail (Protease Inhibitor Cocktail I; EMD Millipore, Billerica, MA) and a phosphatase inhibitor (PhosphoStop; Santa Cruz Biotechnology, Santa Cruz, CA). The homogenate was centrifuged at 15,0006 g for 10 min at 4uC, and total protein concentration of the supernatant was measured [21]. Forty micrograms of total protein from liver tissue was heated at 90uC for 5 min, vortexed, and loaded onto a 4 to 12% Tris-HCl gel for electrophoresis. Samples were separated by SDS-PAGE and drytransferred onto nitrocellulose membranes (iBlot; Invitrogen, Carlsbad, CA). Membranes were blocked for 2 h in blocking buffer (5% dry milk in Tris-HCl buffer, pH 7.5, with 0.05% Tween 20). After incubation with blocking buffer, the membranes were washed 3 times for 5 min each with washing buffer (phosphate-buffered saline, pH 7.5, containing 0.05% Tween 20). Membranes were then incubated with primary antibodies (diluted 1:250) from Santa Cruz Biotechnology against IkBa (catalog #sc-847), c-Jun (#sc-44), and phosphorylated c-Jun (#sc-822) overnight at 4uC. After washing, secondary antibody diluted 1:10,000 was incubated for 1 h at room temperature. Antibody incubation and detection for TNFa was carried out as previously described [11]. Immunodetection was performed by chemiluminescence (West-Dura; Thermo Scientific, Waltham, MA), and bands were quantified by scanning densitometry (ChemiDoc-It Imaging System; UVP Inc., Upland, CA).

Calculations and Statistical Analyses
One cow that received 3 mg/kg rbTNFa developed a severe fever on d 6 of treatment, and the d 7 treatment was not given. Consequently, the liver biopsy and glucose turnover test were not conducted for this cow, but all other data were retained through d 6. Plasma glucose data indicated that 2 cows did not meet steadystate conditions during the glucose turnover test; these cows were not used for turnover rate analysis. In total, data from 33, 32, and 30 cows were used for production responses and plasma analyses, liver tissue analyses, and glucose turnover analysis, respectively. Data were analyzed using the MIXED Procedure of SAS (version 9.2; SAS Institute Inc., Cary, NC) to assess the fixed effects of treatment, parity, time, and 2-and 3-way interactions; cow was included as a random effect. Data were log-transformed for analysis when necessary (including plasma concentrations of

Inflammatory Signals
Compared with Control, plasma TNFa concentrations tended to be increased (P = 0.09) by rbTNFa treatments, but 1.5 mg/kg did not differ (P = 0.19) from 3 mg/kg rbTNF ( Figure 1A). Plasma haptoglobin concentrations were greater (P,0.01, Figure 1B) in rbTNFa treatments than Control, but did not differ between 1.5 and 3 mg/kg rbTNF (P = 0.68). Across the 3 groups, haptoglobin levels were comparable (P = 0.94) at parturition, but increased during the first week of lactation in response to rbTNFa. On d 5 and 7 of treatment, rbTNFa treatments increased haptoglobin concentration by ,2.5-and ,3.5-fold, respectively, compared with Control.
Plasma eicosanoid results are shown in Table 3. Most eicosanoids were unaffected (P.0.10) by treatments; out of the 16 measured eicosanoids, only 9-oxoODE was decreased (P = 0.01) by rbTNFa administration. The pro-inflammatory class as a composite was unaffected by treatments; however, the antiinflammatory class tended (P = 0.08) to be decreased by the 3 mg/ kg compared with 1.5 mg/kg rbTNFa dose (Figure 2), though rbTNFa treatment did not differ from Control (P = 0.22).
Plasma eicosanoids with significant (P,0.05) day effects are presented in Table 4. Plasma concentrations of TXB 2 , 15-HETE, and 9-HODE were elevated around parturition and decreased during the first week of lactation. In contrast, resolvin D 2 and 7maresin 1 increased gradually after parturition. The pro-inflammatory class decreased (P = 0.02) during the first week of lactation, whereas the anti-inflammatory class (P,0.01) increased after parturition.
As shown in Figure 3A, hepatic transcript abundance of TNFa was increased (P = 0.02) by 3 mg/kg compared with 1.5 mg/kg rbTNFa, but overall, rbTNFa treatments did not differ from Control (P = 0.73). There were no treatment effects (P.0.10) for protein abundance of TNFa, c-Jun, or relative c-Jun phosphorylation ( Figure 3B). There was a tendency for a treatment 6parity interaction (P = 0.07) for IkBa, reflecting increased (P = 0.04) IkBa abundance by rbTNFa treatment in primiparous cows (data not shown). Compared with primiparous cows, IkBa was decreased (P = 0.02), TNFa was increased (P = 0.04), and relative c-Jun phosphorylation was decreased (P = 0.04) in multiparous cows ( Figure 3C).  The total anti-inflammatory eicosanoid concentration represents the sum of resolvin D 1 and D 2 , protectin, lipoxin A 4 , 7maresin 1, 9-oxoODE, and 13-oxoODE concentrations [27]. The 3.0 mg/ kg dose tended to differ from the 1.5 mg/kg rbTNFa dose (P = 0.08), but no overall rbTNFa was detected (P = 0.22). A tendency for a dose by time interaction was also observed (P = 0.06), with a significant dose contrast on day 5 (P,0.01). Values are least squares means 6 SEM, n = 10-11. doi:10.1371/journal.pone.0080316.g002 Energetics Dry matter and water intake were decreased (P,0.05) by rbTNFa treatments, but no dose effect (P.0.10) was observed ( Table 5). Cows were in negative energy balance during the experimental period, and the energy balance values were not affected (P.0.10) by treatments. Milk yield was decreased (P = 0.03) by rbTNFa treatments, with no dose effect (P = 0.76). Similar to milk yield, milk energy output was decreased (P,0.01) by rbTNFa administration. Milk fat percentage was decreased (P = 0.02) by 3.0 mg/kg rbTNFa compared with the 1.5 mg/kg dose, but overall, rbTNFa treatments did not differ from Control (P = 0.55). Nevertheless, milk fat yield was decreased (P = 0.01) in cows treated with rbTNFa, and there was a tendency for decreased (P = 0.07) milk fat yield for the 3.0 mg/kg rbTNF dose compared with 1.5 mg/kg. Milk protein and lactose content were unaffected by treatment (P.0.10), but yields of both were decreased (P,0.01) by rbTNFa administration. Milk urea nitrogen concentration and somatic cell linear score were not affected (P.0.10) by treatments.

Health Disorders
Incidence of health disorders is shown in Table 7. There was a tendency for increased (P = 0.08) risk of experiencing 1 or more health disorders for 3 mg/kg compared with 1.5 mg/kg rbTNFa, although the overall rbTNFa effect was not significant (P = 0.25). Ketosis incidence reported in Table 7 was based on detection of urine ketones; of the cows identified by this method, 5 cows (1 control, 2 on each of the rbTNFa treatments) had plasma BHBA concentrations that exceeded 3 mM on at least one day, identifying them as clinical cases, whereas the 2 others exceeded the 1.4 mM threshold for subclinical ketosis [3].

Discussion
Our previous research demonstrated that administration of 2 mg/kg body weight rbTNFa daily for 7 d was adequate to mimic a chronic low-grade inflammatory state without causing acute systemic inflammation in late-lactation cows [11]. Therefore, in the present study, we administered 1.5 and 3 mg/kg rbTNFa in an attempt to characterize a similar low-grade inflammation in earlylactation cows, which experience dramatic energy demands of milk production and tremendous metabolic stress compared with late-lactation animals. Although blood samples were collected 16 h after the previous injection, we still observed a tendency for elevated plasma TNFa concentrations in treated cows, indicating a sustained chronic inflammation status in cows receiving rbTNFa. Moreover, 3 mg/kg rbTNFa administration increased hepatic transcript abundance of TNFa, further supporting that treatment enhanced pro-inflammatory responses.
We also evaluated plasma concentrations of haptoglobin and eicosanoids, and the protein abundance of key inflammatory mediators in the liver. Haptoglobin is an acute phase protein primarily released by the liver in inflammation [24]. Our finding that rbTNFa increased haptoglobin agrees with a previous study that daily administration of 2.5 mg/kg rbTNFa for 7 d dramatically elevated plasma haptoglobin in lactating cows [25]. Interestingly, haptoglobin elevation has been associated with bovine fatty liver [4], suggesting that liver TG accumulation may be associated with inflammation. However, recent research indicated that excess NEFA influx to the liver, rather than hepatic TG accumulation per se, causes liver damage through promoting lipotoxicity, oxidative stress, and inflammatory reactions [26]. In the present study, neither NEFA nor liver TG were affected by treatments, suggesting that rbTNFa increased haptoglobin independent of altered lipid mobilization.
Eicosanoids are a family of signaling molecules produced by oxidation of 20-carbon essential fatty acids. They exert complex control over inflammation, and can either enhance (such as certain types of prostaglandins, thromboxanes, and leukotrienes) or resolve (such as resolvins, protectins, and maresins) inflammatory responses depending on their types and the timing of production [27]. The increased lipolysis and elevated plasma NEFA in early lactation may increase the supply of fatty acid substrates for eicosanoid production, thereby affecting the duration and magnitude of inflammation. In the present study, 6 out of the 16 measured plasma eicosanoids were significantly altered over time, but few treatment effects were detected for eicosanoids. One interesting finding was that the highest dose of rbTNFa tended to suppress the anti-inflammatory eicosanoid class, primarily reflecting decreased concentrations of 7-maresin 1, protectin, and 9-oxoODE. We are unaware of evidence in the literature of cytokines inhibiting release of anti-inflammatory eicosanoids, but such an effect is consistent with the pro-inflammatory effects of TNFa [7]. We cannot rule out the possibility that decreases in anti-inflammatory eicosanoids may have been driven by an unknown mediator rather than a direct effect of TNFa. Still, it is intriguing that rbTNFa administration apparently did more to suppress the production of resolving lipid mediators than to enhance the production of inflammatory lipids, and this result should inform choices about measures of interest in future studies.
We found that concentrations of pro-inflammatory eicosanoids, including certain types of prostaglandins and thromboxanes, were elevated around parturition and decreased during the first week of lactation, whereas pro-resolving metabolites, such as resolvin D 2 Table 5. Production responses in early-lactation dairy cows during the experimental period. Treatments: cows were given 0, 1.5, or 3.0 mg/kg body weight rbTNFa injections daily for the first 7 days of lactation, respectively. 2 No treatment by day interactions were significant except as noted. 3 Contrast between Control and rbTNFa treatments. 4 Contrast between 1.5 and 3.0 mg/kg rbTNFa treatments. *There was a treatment by day interaction (P = 0.04) for milk lactose%.
Values are least squares means 6 SEM, n = 10-11. doi:10.1371/journal.pone.0080316.t005 Table 6. Plasma metabolites and hormones, liver triglycerides, and glucose turnover rate in early-lactation dairy cows. Treatments: cows were given 0, 1.5, or 3.0 mg/kg body weight rbTNFa injections daily for the first 7 days of lactation, respectively. 2 No treatment by day interactions were significant. 3 Contrast between Control and rbTNFa treatments. 4 Contrast between 1.5 and 3.0 mg/kg rbTNFa treatments. Values are least squares means 6 SEM, n = 10-11. doi:10.1371/journal.pone.0080316.t006 and 7-maresin 1, increased after parturition. Consistent with the results of the above-mentioned eicosanoids, the composite proinflammatory class was elevated at parturition and decreased thereafter, whereas the anti-inflammatory class changed in the opposite direction. Although the inflammatory state of postpartum dairy cows has been well documented [28], these are the first data to document an increase in anti-inflammatory eicosanoids during the resolution of inflammation as lactation proceeds. The causative role of eicosanoids in this resolution phase should be a fruitful area for further investigation. Nuclear factor-kB (NF-kB) is a key transcription factor that plays crucial roles in inflammation [29]. In nonstimulated cells, NF-kB is bound to inhibitory IkB proteins and sequestered in the cytoplasm. Activation of NF-kB primarily occurs via the phosphorylation, ubiquitination, and degradation of IkB proteins; subsequently, the liberated NF-kB is translocated to the nucleus to drive expression of inflammatory genes [30]. One of the major IkB proteins in inhibiting NF-kB activation is IkBa, the abundance of which can be used to estimate anti-inflammatory status. In this study, hepatic IkBa abundance was significantly increased by rbTNFa in primiparous cows. This observation is unexpected, because TNFa is a potent activator of IkBa degradation [29], and suggests the involvement of compensatory anti-inflammatory adaptations in response to rbTNFa.
We also investigated the inflammatory mediator c-Jun, which is a key component of the pro-inflammatory transcription factor activator protein-1. The transcriptional activity of c-Jun is increased by phosphorylation of serine residues in response to various stimuli, including inflammation [31]. As for the NF-kB pathway, we found no evidence of hepatic c-Jun activation in response to rbTNFa. Collectively, rbTNFa treatment did not significantly alter the hepatic protein abundance of inflammatory mediators, but nevertheless greatly increased the plasma concentration of haptoglobin, an acute phase protein induced primarily by inflammatory signals in the liver [32]. It is possible that haptoglobin was primarily induced by transcription factors other than NF-kB or activator protein-1.
In this study, feed intake and milk production were decreased by rbTNFa treatments by 18 and 15%, respectively. TNFa is a key mediator causing anorexigenic responses in various diseases, and many of the effects of TNFa are dependent on its actions in the hypothalamus [33]. Interestingly, our previous research in latelactation cows showed that 2 mg/kg rbTNFa treatment daily for 7 d also decreased feed intake by 18%, but did not significantly alter milk production [11]. Kushibiki et al. [25] reported that daily administration of 2.5 mg/kg rbTNFa for 7 d dramatically decreased feed intake (by ,34%) and milk yield (by ,15%). As expected, cows were in negative energy balance during the immediate postpartum period, indicating that their energy expenditure, driven largely by lactation requirements, greatly exceeded their energy intake. However, energy balance values were not affected by rbTNFa administration because the magnitude of decreased milk production was comparable to that of decreased feed intake.
Administration of rbTNFa did not alter plasma markers of lipid metabolism or promote fatty liver. As a lipolytic mediator in adipose tissue [34], previous studies found that rbTNFa administration at 2.5 mg/kg significantly increased plasma NEFA  concentrations in mid-lactation cows [25] and heifers [35]. We also previously reported that 7-d rbTNFa administration at 2 mg/ kg daily increased liver TG by ,2-fold [11]. It is possible that because plasma NEFA and BHBA concentrations, as well as liver TG content, were already dramatically elevated in response to negative energy balance in the periparturient period, the doses of rbTNFa used in this study were insufficient to promote further alterations in lipid metabolism. On the transcriptional level, our previous work [11] in late-lactation cows showed that 2 mg/kg rbTNFa administration daily for 7 d increased the transcript abundance of AGPAT1 (which catalyzes the esterification of a fatty acyl-CoA to the sn-2 position of the glycerol backbone in TG synthesis), tended to decrease CPT1a (which catalyzes transport of fatty acids into mitochondria for oxidation), and did not affect ApoB (which is a key component of very low-density lipoprotein that mediates hepatic TG exportation). Here, we did not find treatment effects for AGPAT1 or ApoB, and found that rbTNFa increased CPT1a abundance dose-dependently. Collectively, rbTNFa did not promote the pathways favoring hepatic lipid accumulation in early-lactation cows.
Lactating cows rely heavily on hepatic gluconeogenesis to meet glucose requirements because the majority of dietary carbohydrates are fermented in the rumen. Under steady-state conditions, the rate of glucose appearance equals the rate of its disappearance, and the calculated glucose turnover rate approximately equals gluconeogenesis in ruminants, because net portal appearance of glucose is negligible [5]. In this study, there were no treatment effects on plasma glucose or insulin, abundance of key gluconeogenic transcipts, or plasma glucose turnover rate, indicating that low-grade inflammation did not affect hepatic gluconeogensis in early-lactation cows. Previous data regarding TNFa effects on gluconeogensis have been inconsistent. Short-term recombinant human TNFa administration at 10 mg/kg, but not 3.5 mg/kg, significantly increased glucose production (likely from glycogenolysis) and metabolic clearance rate in dogs; coupled with decreased fatty acid flux, the authors concluded that TNFa given in a high dose causes a shift toward carbohydrate as an energy substrate [36]. In a subsequent study, Sakurai et al. [37] infused 2.5 mg/kg TNFa in dogs for 2 h and did not observe a change in glucose production.
The primary substrate for gluconeogenesis in ruminants is propionate, a short-chain fatty acid derived from ruminal fermentation; as a result, liver glucose production in cows is often a function of energy intake [38]. To assess whether gluconeogenesis was affected by substrate supply, we expressed glucose production relative to feed intake. Interestingly, this analysis revealed that the 3 mg/kg rbTNFa treatment increased the apparent efficiency of gluconeogenesis from dietary substrates, primarily because of the low feed intake for that treatment. Given that propionate is used for glucose production at.90% efficiency in ruminants [39], increased metabolic efficiency was not likely the reason for this response. When requirements are greater than nutrient intake, such as in early lactation, the contributions of lactate, glycerol, and amino acids from body tissues to gluconeogenesis increase markedly [40]. Glycerol is released primarily from adipose TG mobilization, which was likely unaffected by treatments (based on unaltered plasma NEFA). In an attempt to evaluate whether 3 mg/kg rbTNFa-treated cows had greater amino acid supply from protein mobilization, we measured concentrations of plasma 3-methylhistidine, a marker of muscle protein breakdown that has been shown to increase rapidly during the periparturient period and peak at 1 week after parturition [41]. Treatment effects on plasma 3-methylhistidine were not significant; however, the 3 mg/kg rbTNFa treatment mean was numerically 20% greater than the other treatments. It seems likely that a combination of slightly enhanced muscle proteolysis and decreased amino acid use for milk protein synthesis resulted in greater supply of amino acids for gluconeogenic substrate in the 3 mg/kg rbTNFa treatment.
Despite the lack of dramatic alterations in metabolism, the higher dose of rbTNFa tended to impair the health of postpartum cows. Health disorders affect considerable numbers of periparturient dairy cows, compromising animal welfare, milk production, and fertility. Ketosis is considered the most important metabolic disease in the U.S. dairy industry [3], and ketotic cows are also at increased risk for other disorders, such as displaced abomasum and metritis [42]. Ketosis may increase the risk of infectious disorders through negative effects of ketone bodies on immune cell function [43]. In this study, rbTNFa increased the incidence of ketosis by 3-fold in the first week of lactation, accounting for much of the increase in total disease incidence. There are numerous mechanisms by which rbTNFa may have promoted subclinical ketosis, but poor feed intake in early lactation is a key risk factor for ketosis [1], and it is likely that the suppression of intake by rbTNFa played a role in this study. Decreased feed intake did not decrease energy balance or increase lipolysis in this study, but low intake affects more than just energy status; decreased protein, mineral, vitamin, or lipid intake may have adversely affected health.

Conclusions
In conclusion, rbTNFa treatment during the first 7 d of lactation decreased feed intake, reduced milk production, and tended to impair health status, but did not significantly alter glucose or lipid metabolism. These data indicate that in response to low-grade inflammation, early-lactation cows compensate for the reduced intake by suppressing milk production rather than further compromising their energy balance or systemic metabolism. This coordinated physiological response to inflammatory mediators can be considered a biological tradeoff between lactational performance and survival [44]. However, because effects on feed intake and milk yield occurred simultaneously, it is unclear whether one or the other was a secondary response. Further studies are needed to investigate if inflammation directly alters mammary or neurological function. Nevertheless, this study indicates that preventing excessive inflammation in early lactation has the potential to improve productivity and health of dairy cows.