The present study was part of a larger project described earlier (33). Briefly, 30 primiparous, non-pregnant German Holstein cows (169 ± 48 days in milk (DIM)) were evenly allocated to heat-stressed (HS, n = 10), control (CON, n = 10), or pair-fed group (PF, n = 10). All cows were adapted to the climate chamber at thermoneutral conditions (permanent 16°C and temperature–humidity index (THI) = 60) for 6 days and received a total mixed ratio ad libitum twice daily at 0730 h and 1730 h (adaptation phase) (33). In the experimental phase, HS cows were kept for seven days at 28°C with 51% ± 2% relative humidity (RH), resulting in a THI of 76 with ad libitum feeding and water access. The CON group was exposed for seven days to 16°C with 69% ± 2% RH and a THI of 60 with ad libitum feeding. The PF cows, as the second control group to mimic metabolic stress, were exposed for seven days to 16°C with 69% ± 2% RH and a THI of 60. In order to ensure isoenergetic and isonutritive feeding of HS and PF cows, the feed intake of HS cows was calculated as a percentage of the daily mean feed intake per kilogram of body weight (33). In the climate chambers, the day–night rhythm was given by a light cycle ranging from 0600 to 1900 h. The cows were milked at 0700 h and 1730 h daily. The animal studies were approved by the ethics committee of the State Government in Mecklenburg-West Pomerania, Germany (LALLF M-V/TSD/7221.3-1.1-60/19). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent was obtained from the owners for the participation of their animals in this study. All methods were in compliance with the ARRIVE guidelines (34).

At 1 day before the start of the experimental phase, the cows were equipped with an indwelling jugular catheter (Cavafix, B. Braun, Melsungen, Germany). Blood samples were taken after feeding in the morning according to the following scheme: 0 h, 6 h, day 1, day 2, day 3, day 4, day 5, day 6, and day 7, each into 9-mL monovettes (Sarstedt, Nümbrecht, Germany) containing EDTA or into 9-mL serum tubes and immediately cooled on ice. After collection, the EDTA blood samples were centrifuged at 0.7 × g for 20 min at 4°C. The serum samples were kept at 4°C overnight and centrifuged using the same conditions. The plasma and serum samples were aliquoted and stored at -80°C until further analysis. On day 1 and day 6, additional 4-mL blood samples were subjected to PBMC isolation utilizing a Histopaque 1077 gradient (Sigma-Aldrich, Schnelldorf, Germany) (35). The isolated PBMC were frozen in liquid nitrogen and stored at -80°C before further analysis.

Whole-blood samples from EDTA-coated tubes were instantaneously subjected to a complete blood count using a hematology analyzer VetScan HM5 (Scil Animal Care Company, Vierheim, Germany). The total white blood cell count (WBC), WBC differential counts for neutrophils, eosinophils, basophils, lymphocytes, and monocytes, as well as the red blood count (RBC), hematocrit value (HCT), hemoglobin concentration (HGB), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), and RBC distribution width (RDW) were analyzed. In addition, platelet indices were analyzed, including the platelet count, mean platelet volume (MPV), platelet distribution width (PDW), and plateletcrit (PCT).

In serum samples, total iron concentrations were measured spectrophotometrically and potentiometrically (ABX Pentra C400 clinical chemistry analyzer; HORIBA Europe GmbH, Oberursel, Germany) using a commercial kit (A11A01-637; HORIBA ABX SAS, Montpellier, France). Serum LPS-binding protein (LBP) concentrations were measured by a commercially available ELISA (HK503, Hyltec Biotech). The intra- and inter-assay coefficients of variation were 7.2% and 15.5%, respectively.

Circulating Gram-negative bacterial endotoxin concentrations were determined by a limulus amebocyte lysate (LAL) chromogenic endpoint assay using serum samples (HIT302, Hycult Biotech, Uden, Netherlands) according to manufacturer’s instructions. The assay was performed using certified endotoxin-free and depryogenated consumables in a sterile bench. The standards and samples were diluted in glass tubes incubated for 4 h at 200°C. The control standard endotoxin was diluted 1:10 with endotoxin-free water for serial dilutions, with final standard concentrations of 10, 4, 1.6, 0.64, 0.26, 0.10, and 0.04 EU/mL and blank. The samples were diluted 1:10 and assessed in duplicates on a 96-well plate. A reconstituted LAL reagent was prepared according to the individual kit certificate. The samples and standard (50 µL) were mixed with 50 µL LAL reagent, sealed, and incubated at room temperature for 20 min. Subsequently, the absorption was measured every 5 min at 405 nm until the difference between 10 and 4 EU/mL was ≤10%. Finally, 50 µL of stop solution was added, and the sample mixture was measured at 405 nm in a plate reader (Tecan infinite 200, Männedorf, Switzerland). The final sample’s endotoxin concentration was calculated against the absorbance of the standards in a linear regression model including the elimination of the background signal from the blank. Serum samples of the same experimental day were measured on the same plate. The intra- and inter-assay coefficients of variation were 6.0% and 17.3%., respectively.

The protocol for measuring the presence of NF-κB in the nucleus was adapted from Selkirk et al. (36). For each sampling time point (except for 6 h), duplicates of 50 µL of whole blood were incubated with 500 µL VersaLyse (Beckman Coulter, Krefeld, Germany) for 20 min. The samples were centrifuged for 5 min at 250 × g, and the supernatant was discarded. The cell pellets were washed with 500 µL PBS (Gibco Life Technologies, Carlsbad, CA, USA) and centrifuged for 5 min at 150 × g, and the supernatant was discarded. For nuclei isolation, BD CycleTest Plus DNA reagent kit was used according to the manufacturer’s instructions (BD Bioscience, Franklin Lakes, NJ, USA). In brief, the cell pellet was re-suspended in 500 µL of citrate buffer and centrifuged for 5 min at 300 × g, the supernatant was discarded, and the washing with citrate buffer was repeated. Thereafter, 125 µL of solution A was added and incubated 10 min in the dark, following the incubation with 100 µL of solution B for 10 min in the dark. Isolated nuclei were stained with 1.1 µL NF-κB p65 antibody (C22B4, rabbit monoclonal antibody, Cell Signaling Technology, Denvers, MA, USA), and the duplicate treated without primary antibody served as the negative control. After 30 min of incubation, 2.2 µL goat anti-rabbit AlexaFlour488 (1:10; Invitrogen, Thermo Fisher Scientific) was added and incubated for 30 min at 4°C. Solution C (100 µL) was added and incubated for 10 min at 4°C. The samples were quantified on a Gallios flow cytometer (Beckman Coulter) and analyzed using Kaluza software (Beckman Coulter). For the intra-nuclear NF-κB analysis, double-gating on a forward vs. side scatter dot plot and FL2 PI staining were used to analyze 10,000 cells. Flow cytometry immunofluorescence acquisition was performed within 1 h of staining. To calculate the NF-κB p65 fluorescence intensity, the fluorescence intensity of the negative control was subtracted from each NF-κB p65-stained sample.

For each sampling time point, 100 µL of whole blood was incubated for 20 min with 1,000 µL VersaLyse (Beckman Coulter). The samples were centrifuged for 5 min at 250 × g, and the supernatant was discarded. The cell pellets were washed twice with 700 µL 1× PBS (Gibco Life Technologies). Subsequently, the cell pellet was dissolved in 100 µL PBS and frozen in ice-cold methanol (VWR International, Darmstadt, Germany) for further analysis. Methanol-fixed cells were washed twice with 1× PBS. For staining, duplicates of each time point were prepared to either stain with HSP70 (SMQ-SMC-100B, mouse monoclonal antibody, Biozol, Sontheim an der Brenz, Germany) or without antibody as negative control and incubated overnight at 4°C. On the next day, the cells were washed with 1× PBS, and duplicates were stained with rabbit anti-mouse AlexaFlour488 antibody (1:1,000, Thermo Fisher Scientific, Waltham, MA, USA) with incubation for 1 h at 4°C. Thereafter, the cells were washed with 1× PBS, resuspended in 500 µL 1× PBS containing Hoechst 33342 (0.5 µM final concentration, Invitrogen, ThermoFisher), and incubated for 20 min at 4°C. The samples were quantified on a Gallios flow cytometer (Beckman Coulter) and analyzed using Kaluza software (Beckman Coulter). The mean fluorescence intensity of the negative control was subtracted from each HSP70-stained.

The PrimeFlow^® RNA assay (Invitrogen, Thermo Fisher Scientific) was performed daily on blood samples following the manufacturer’s protocol divided over 2 days. All buffers were included in the PrimeFlow™ RNA assay kit, and specific target probe sets for bTNFA:A488 and bIL1B:A647 (target set 1) and bIL6:A488 and bIFNG:A647 (target set 2) were designed and purchased from Thermo Fisher Scientific. Cell surface staining was performed in duplicate each by incubating 100 µL blood with 10 µL CD14:AlexaFluor700 (MCA1568A700, BioRad, Hercules, CA, USA) for 30 min at 4°C. Subsequently, the cells were fixed for 30 min at 4°C. After permeabilization, the cells were incubated for a second time with fixation buffer for 1 h at room temperature in the dark. Then, the cells were incubated for the hybridization with the appropriate target probe sets for 2 h at 40°C. The samples were kept overnight at 4°C in the dark. On the following day, pre-amplification and amplification of the hybridization were performed twice for 1.5 h at 40°C with the pre-amplification mix and subsequently the amplification mix. Next, the cells were incubated with the label probe set for 1 h at 40°C and quantified thereafter on a flow cytometer (Gallios, Beckman Coulter). The analysis was performed with Kaluza software (Beckman Coulter) based on total cell count and percentage of CD14⁺ cell surface staining.

Isolated PBMC obtained on days 1 and 6 were utilized to extract RNA by utilizing the NucleoSpin RNA mini kit (Macherey-Nagel, Düren, Germany) with an additional DNase digestion step (37). The extracted RNA was tested for genomic DNA contamination by performing PCR (38). RNA concentration was measured with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), and quality was verified by using a Qubit 2.0 fluorometer (Thermo Fisher Scientific). RNA integrity was evaluated by utilizing the Bioanalyzer 2100 (Agilent Technologies, Böblingen, Germany) with mean RIN values of 8.36 (min 7.6 and max 9.3). Subsequently, a stranded library preparation protocol for RNA sequencing was applied (TruSeq mRNA sample prep kit, Illumina, San Diego, CA, USA) using indices for multiplexing and polyA selection to focus on polyadenylated RNA (in the majority mRNA). The RNAseq libraries were checked for quality on the Bioanalyzer 2100 and subjected to paired-end sequencing (2 × 100 bp) using the Illumina HiSeq 2500 system (Illumina).

The CASAVA (Illumina) software was used for demultiplexing of reads. Scripts written in Linux [including command from the samtools suite (39)] and R were applied for data processing. For quality control, read alignment, and transcript quantification, we used the nf-core/rnaseq -r 3.4 pipeline. This included removing adapters and low-quality bases with Cutadapt version 3.4. The reads were aligned to the bovine reference genome ARC-UCD 1.2 with Ensembl reference annotation version 105 using STAR version 2.6.1d. In the nf-core/rnaseq pipeline, the option to use Salmon version 1.5.2 was selected to establish expression counts at the gene level. Differential expression analysis was performed with DESeq2 version 1.26.0 (40), with a threshold for significance of adjusted P (Padj) <0.05.

The Database for Annotation, Visualization, and Integrated Discovery (DAVID, version 6.8 with updates from December 2021) was utilized to investigate enriched biological themes and cluster analysis (41). The unique list of differentially expressed genes with ensemble gene ID (q < 0.05) was submitted as gene list and Bos taurus database as background. The cutoff value for multiple enrichment testing determined via the Benjamin–Hochberg algorithm was 0.05. Only the results from Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were selected for functional annotation categories (https://david.ncifcrf.gov/tools.jsp).

Daily measurements on the same animal were analyzed by repeated-measurement ANOVA using the MIXED procedure of SAS (Version 9.4, SAS Institute Inc., Cary, NC, USA). Based on Akaike’s Information Criteria (AIC), the autoregressive type or compound symmetry for the block diagonal residual covariance matrix was chosen. The model contained the fixed effects of treatment (HS, CON, PF), time (experimental day), the interaction (treatment × time), and days in milk which served as covariate. Least squares means (LSM) and their standard errors (SE) were computed for each fixed effect in the ANOVA model. Additionally, differences of these LSM were tested using the Tukey–Kramer procedure. The SLICE statement of PROC MIXED was used to perform a partitioned analysis of the LSM for the interaction treatment × time. Results were considered as statistically significant at P <0.05 and tendencies at 0.05 < P <0.09.

As shown in Table 1 , the number of erythrocytes was significantly lower in HS than PF cows on day 7 (P < 0.05). The hemoglobin concentration (P = 0.07) and hematocrit (P = 0.09) parameters tended to be lower in HS than PF cows on day 7. Furthermore, there was a significant treatment × day interaction for the number of erythrocytes (P < 0.01), hemoglobin concentration (P < 0.01), and hematocrit (P < 0.001). In addition, the RDW was significantly lower in HS than PF cows on day 0, 6 h, and day 2 (P < 0.05, respectively) and significantly lower in HS and CON than PF cows between days 4 to 7 (P < 0.05, respectively). No group differences were observed for the MCV, MCH, and MCHC values.

For the platelet parameters, neither the platelet count, percentage of platelet counts, MPV, or plateletcrit reached the overall significance level for treatment ( Table 1 ). However, the percentage of platelet counts and plateletcrit tended to be lower in HS than PF cows on day 6 (P < 0.08), yet the MPV was (P < 0.05) or tended to be (P = 0.08) lower in HS than PF cows on days 6 and 7, respectively. In addition, the PDW was significantly lower in HS than PF cows on days 4 and 6 (P < 0.05, respectively) and was significantly higher in PF than CON cows on day 0, 6 h, and days 6 and 7 (P < 0.05).

The number of blood lymphocytes in HS cows was (P < 0.05) or tended to be (P = 0.08) lower than in PF cows between days 3 and 7 ( Table 2 ). During the same time, the number of lymphocytes was or tended to be higher in PF compared to CON cows (P < 0.05, respectively). In addition, there was a significant treatment × day interaction for lymphocytes (P < 0.05) and a tendency for the eosinophils (P = 0.08) and the percentage of neutrophils (P = 0.094). However, no differences were found among the groups for the percentage and the number of leukocytes, monocytes, neutrophils, eosinophils, and basophils.

The HS cows tended to have lower serum iron concentrations than PF or CON cows (P < 0.1), but on day 7, the serum iron concentration was significantly lower in HS than CON and PF cows (P < 0.05; Table 1 ). The serum endotoxin concentration was not significantly different between groups but higher in HS than PF cows on day 1 and tending to be higher in HS than CON cows on days 2 and 7 (P < 0.09, respectively; Figure 1A ). Furthermore, PF cows showed higher endotoxin concentrations in comparison to CON cows on days 6 and 7. However, the concentration of the acute-phase protein LBP was not affected by treatment ( Figure 1B ).

To evaluate if the temporary higher endotoxin concentrations in HS cows triggered NF-κB p65 translocation into the nucleus of leukocytes, fluorescence double-labeling and subsequent flow cytometry analyses were performed ( Figure 2A ). The portion of NF-κB p65 positively stained nuclei did not differ by treatment or time ( Figure 2B ). However, the mean fluorescence intensity (FI) of NF-κB p65 labeling was significantly lower in HS than CON and PF cows on day 5 (P < 0.05), whereas on day 6, it was significantly higher in HS than CON and PF cows (P < 0.05, Figure 2C ). Interestingly, heat stress did not affect the leukocytic HSP70 protein expression as determined by fluorescence intensity ( Supplementary Figure S2 ).

Next, we analyzed the mRNA abundances of inflammatory cytokines and mediators in CD14⁺ monocytes and total leukocytes. The PrimeFlow analysis revealed no differences in the TNFA and IL1B ( Figures 3A–F ) or IL6 and INFG ( Supplementary Figures S1B–D ) abundances in CD14⁺ cells or total leukocytes between the groups (P > 0.05). However, PF cows tended to have more leukocytic IL1B mRNA than CON cows on day 7 (P = 0.07; Figure 3F ). Furthermore, the IL6 and IL1B (P < 0.05) and TNFA (P = 0.054) mRNA abundances in leukocytes changed over time ( Figures 3D, F , Supplementary Figure S1C ).

To study further the cellular pathway activations of the adaptive and innate immune response on the mRNA level, RNAseq analysis in PBMC (including T cells, B cells, monocytes, dendritic cells, and natural killer cells) sampled on days 1 and 6 was conducted. Demultiplexing and filtering yielded 80–100 million reads per PBMC sample. Approximately 90% were mapped to the bovine genome. The result of the Sparse Partial Least Squares-Discriminant Analysis (sPLS-DA) on day 6 revealed a discrimination between the gene expression of each group and explained 18% of the variation by component 1 and 16% of the variation by component 2 ( Figure 4 ). It is worth noting that on day 1, almost no difference on gene expression was found between cow groups, and only KLRF2 was 23-fold more highly expressed in HS than CON cows (q < 0.001; Supplementary Table S1 ). However, the transcriptome analysis on day 6 revealed 305 differentially expressed genes between HS and CON cows (q < 0.05, Supplementary Table S2 ). Among these, 63 genes were more highly expressed and 242 genes were expressed lower in HS compared to CON cows ( Supplementary Table S2 ). Comparing HS to PF cows, the transcriptomic analysis showed 879 differentially expressed genes (q < 0.05, Supplementary Table S3 ). In the HS group, 501 genes were more highly expressed and 378 genes were expressed lower than in the PF group (q < 0.05, Supplementary Table S3 ). The top five genes that were most significantly expressed higher in HS compared to CON cows were RBPMS2 (2.3-fold), GJA10 (2.0-fold), P2PX1 (1.9-fold), CDKN1A (1.8-fold), and CROT (1.7-fold), whereas COL5A3 (-3.2-fold), FMO2 (-2.6-fold), SCN11A (-1.6-fold), MS4A14 (-1.6-fold), and LRRC63 (-1.5-fold) were expressed significantly lower in HS than CON cows (q < 0.05; Supplementary Table S2 ). Comparing HS with PF cows, the five highest upregulated genes in HS cow were P2PX1 (2.4-fold), RBPMS2 (2.3-fold), CROT (1.9-fold), CDKN1A (1.8-fold), and MAP2 (1.6-fold), while HBA (-3.3-fold), MTUS1 (-2.5-fold), TENM4 (-2.2-fold), COL5A3 (-1.9-fold), and MYOM3 (-1.9-fold) were the genes with the lowest expression in HS than PF cows ( Supplementary Table S3 ).

The functional pathway analysis in DAVID showed an enrichment of genes with higher expression of HS compared to CON cows in the following pathways: platelet activation, regulation of actin cytoskeleton, leukocyte transendothelial migration, focal adhesion, tight junction, oxytocin signaling pathway, cGMP-PKG signaling pathway, reactive oxygen species, thermogenesis, and endocrine resistance, whereas pathways related to terpenoid backbone biosynthesis, lysine degradation, and other metabolic pathways were enriched with genes that were expressed lower in HS compared to CON cows ( Figure 5 , Supplementary Table S4 ). The comparison of HS vs. PF cows revealed pathways enriched for genes encoding for platelet activation, complement and coagulation cascades, focal adhesion, ECM receptor, fluid shear stress and atherosclerosis, Rap1 signaling, phospholipase D signaling, regulation of actin cytoskeleton, and adherents junction, whereas genes with a lower expression were enriched in pathways regulating cell cycle, DNA replication, base excision repair, intestinal immune network for IgA production, lysine degradation, homologous recombination, antigen processing and presentation, and T cell receptor signaling ( Figure 6 , Supplementary Table S5 ).

In our study, heat stress had a stronger impact on the number of erythrocytes, hemoglobin, and hematocrit. The RBC, although still in the normal range with 6–8 × 10⁶ cells/µL, decreased after 7 days of heat stress and showed a trend toward lower hemoglobin and hematocrit in HS than in PF cows, suggesting the onset of normochromic normocytic anemia with unaltered MCH, MCHC, and MCV. Very limited information is available for heat stress effects on RBC-related parameters in mid-lactating cows. One study reported lower RBC, hematocrit, and RDW prepartum in summer-calving cows compared to winter-calving cows (11). In our study, lower serum iron concentrations were found in HS cows compared to both ambient temperature groups. These results are consistent with a previous study showing that lower hemoglobin concentrations and hematocrit were associated with lower serum iron concentrations in HS calves (42). In both studies, the decrease in erythrocytes, hemoglobin, and hematocrit was explained by a reduction in cellular oxygen demand and to reduce heat production due to oxidative metabolism (43). Different mechanisms are possible to explain the lower erythrocyte number occurring during heat stress. The lower RBC might be caused by a hemodilution effect, as more water is taken up and transported into the circulation to increase heat dissipation (11). Earlier findings from this heat stress trial reported a significant increase in water uptake per kilogram of dry matter intake after 2 days in HS compared to CON cows and after 5 days when compared to PF cows (33). Another mechanism might be that autoinflammation and subsequent hemolysis may reduce the lifespan of erythrocyte. The causes of autoinflammation are related to increased reactive radical formation and toxin activation (44). Furthermore, the accumulation of reactive oxygen species might damage the erythrocyte membrane and induce erythrocyte deformation, leading to erythrocyte dysfunction and removal from the blood circulation (45). Hypoxic episodes in the splanchnic area and diminished gut barrier function with ingress of toxic particles are likely to occur during heat stress (7, 46) and might enhance the hemolysis of erythrocytes. The lower erythrocyte number was found to be associated with lower serum iron concentration during heat stress, suggesting the disturbance of iron homeostasis accompanied with reduced iron availability for erythrocyte production (47). Under conditions of anemia in humans, iron is sequestered in ferritin within macrophages of the reticuloendothelial system, resulting in the reduced availability of iron-bound transferrin for erythropoiesis (48). However, this issue needs further research to elucidate the cellular mechanism in erythrocytes of heat-stressed dairy cows.

Interestingly, HS cows tended to have lower relative platelet and PCT and significantly lower MPV and PDW on day 6 compared to PF, but not CON cows. Different studies have reported that moderate and severe heat stress causes microvascular injuries of the epithelium, dominantly found in gut, kidney, and lungs (49, 50). Platelets and leukocytes adhere to the vascular endothelium and activate the coagulation pathway during severe heat stress (50). In our study with moderate heat stress, the temporary decline in the relative number of circulating platelets might be linked to the sealing of microvascular injuries in HS cows. However, it is not entirely clear why these differences did not occur between HS and CON cows, and thus it merits future research.

Our data showed that 7 days of persistent heat stress was associated with a dynamic lymphocyte response. Specifically, the absolute number of blood lymphocytes in HS cows decreased in a time-dependent manner and was lower than in PF cows but was not different from CON cows. This phenomenon started to occur after 3 days of heat stress and persisted until the end of the period, indicating that under conditions of long-, but not short-, term heat stress the number of lymphocytes declines. Previous results have shown a tendency for lower lymphocyte numbers in the prepartum and postpartum period in summer-calving cows to winter-calving cows (11). In buiatrics, a decline in the number of leukocytes is usually associated with stress, fear, exercise, or parturition, leading to lymphocytopenia (51). Possibly, the lower number of blood lymphocytes observed in HS compared to PF cows may be explained by heat-stress-induced lymphocytopenia, but this assumption needs to be proven in future investigations.

A lower leukocyte number in the blood circulation is associated with endotoxemia and inflammation (11). The presence of endotoxin in the systemic circulation might be the cause for a low-grade inflammation during mild heat stress as mentioned earlier. Daily endotoxin measurements revealed an initial increase of endotoxin after 1 day of heat stress, a decline between days 3 and 5, and a re-increase until 7 days of heat stress, indicating the dynamic adaptational processes of HS cows to maintain immune homeostasis. Interestingly, 7 days of pair-feeding increased the serum endotoxin concentrations in comparison to ad libitum feeding, suggesting that the reduction in feed intake is a sufficient criterion for the increase in endotoxin concentration. A previous study in pigs showed that both pair-feeding and heat stress increased the serum endotoxin levels after a short-term environmental challenge (12 h) compared to CON pigs (52). Pearce et al. (2015) related their results to a reduced gut integrity and/or liver capacity to detoxify endotoxin molecules (52). In dairy cows, a 60% reduction in feed intake for 5 days relative to ad libitum feeding increased the gut permeability, as indicated by increased plasma LBP and Cr concentrations upon oral Cr administration (53). An expected increase of the LBP, as observed after 14 days of heat stress in dairy cows (54), was not found in the present study, which might be due to individual differences among the animals.

To analyze how heat stress affects the response of the immune system, we studied the effect of chronic heat stress on the daily leukocytic NF-κB p65 and HSP70 presence potentially regulating the gene expression of pro-inflammatory cytokines. Our results showed a first drop of the mean NF-κB p65 FI on day 5 and a subsequent rise in HS cows compared to both control groups kept under thermoneutral conditions, indicating dynamic changes and a temporary higher NF-κB p65 presence in the nucleus, although the total number of NF-κB p65⁺ cells were maintained. Interestingly, HSP70 was not altered during heat stress, indicating that HSP70 does not appear to be involved in the regulation of NF-κB p65 translocation, which was not expected. The activation of the CD14/TLR2/4 pathways leads to the translocation of NF-κB p65 from the cytosol to the nucleus, a condition known to initiate, among others, the gene expression of pro-inflammatory cytokines (55). In this study, we measured in leukocytes and CD14⁺ cells concomitantly the gene expression of TNFA, IFNG, IL1B, and IL6. Unfortunately, the variation in gene expression among animals was enormous; therefore, we could not find significant differences between groups. The limited number of animals may be one reason for this. Thus, future experiments should be conducted with a larger sample size of animals to gain a more comprehensive understanding of the cytokine expression in CD14⁺ positive cells. However, in an earlier study using the same number of animal, we showed that heat stress induced an increase in plasma TNFα and IFNγ concentrations and TNFA mRNA abundance in the PBMC of HS cows (33), suggesting differences between the transcriptional and protein expression levels or a different origin of pro-inflammatory cytokines than from circulating leukocytes and CD14⁺ cells.

The transient activation of NF-κB p65 signaling was not found parallel in time with increased blood endotoxin concentration triggering the TLR2/4 pathways. However, other pathways, e.g., Notch, MAPK, and JAK-STAT, also influence NF-κB p65 activation and the transcription of cytokines, which may lead to higher plasma TNFα and IFNγ concentrations (56–59), particularly at the beginning of heat stress (33). Further research is required to gain detailed insight into the complex immune signaling system of heat-stressed dairy cows.

We examined the immunological response in PBMC using RNA sequencing to improve our understanding of the molecular regulatory mechanism after short-term (24 h) and long-term (6 days) heat stress. Surprisingly, 24 h of heat stress did not alter the gene expression pattern in comparison to both control groups. Significant differences were found between the treatment groups after 6 days of chronic heat stress. The most prominent signaling pathway upregulated in HS than in both control groups was the activation of platelets in conjunction with vascular smooth muscle contraction, complement and coagulation cascade, and focal adhesion ( Table 3 ). In line with this finding, earlier studies in different species described that heat stress leads to endothelial microvascular injuries, thrombosis, fibrinolysis, and inflammation, resulting in disseminated intravascular coagulation and compromised blood supply to organs (50, 60, 61). The microvascular injury might be caused by fluid sheer stress due to the increasing blood flow from the gut toward the skin as an attempt to minimize heat load (61). An activated endothelium attracts platelets, monocytes, and neutrophils which can initiate and amplify coagulation (50). For the activation of platelets, different sets of genes belonging to hemostasis were more highly expressed in HS cows (1): vasoconstriction (VASP) (2), platelet activation (PTGR1, GNAS, P2X1, ITGA2, and ITGA2B), and (3) coagulation cascade with clot formation (VWF, F2RL3, F2, GP5, and GP1BA). The required platelets for this process might also explain the tendency for the lower relative platelet count on day 6 of heat stress. An activated coagulation cascade including coagulation factors and fibrinogens was evidenced in the plasma proteome of lactating dairy cows sampled in summer compared to winter, which agrees with our findings (62). Furthermore, our enrichment analysis revealed the activation of leukocyte transendothelial migration during 6 days of heat stress. In PBMC, genes related to docking structure (VASP, VCL, and ACT) and F11 receptor (F11R) were found to be more highly expressed during heat stress. The F11 receptor is involved in T cell and neutrophil transmigration (63). The data imply leukocyte transendothelial migration via chemoattractants, e.g., pro-inflammatory cytokines (64) that may guide these immune cells to their destination. Recently published data from the same animals revealed higher plasma TNFα and IFNγ concentrations after 7 days of heat stress in comparison to CON and/or PF cows, suggesting an increased level of chemoattractants (33), which, in turn, supports our RNAseq data. However, future research needs to be intensified to prove altered coagulation by testing clotting time and platelet aggregation tests in HS cows under practical conditions.

Among the pathways less activated in HS cows in comparison to both control groups were metabolic pathways ( Table 3 ). Immune cells are able to sense metabolic stress and adapt adequately to new nutritional situations (65). Heat stress seems to affect the protein metabolism by the reduced expression of genes involved in lysine degradation (SUV39H1, ALDH3A2, EZH2, and HYKK). The altered mRNA expression of these genes could be caused by altered plasma lysine concentrations, which, unfortunately, were not analyzed in this study. Recently, we observed no difference in plasma lysine concentration between HS and PF cows (66), suggesting that plasma lysine concentrations did not affect the lysine degrading pathways in PBMC. Rather, the reduced expression of genes involved in lysine degradation seems to spare lysine for cellular protein synthesis.

Moreover, heat stress negatively affects the lipid metabolism of PBMC by reducing the mRNA expression of ACAT2, CROT, MAPK14, SCL25A17, and SCD. Several studies reported reduced plasma non-esterified fatty acids (NEFA) concentration under thermal stress (67, 68). There is increasing evidence that NEFA does not seem to be a desirable fuel for a heat-stressed organism, potentially to prevent metabolic heat production via β-oxidation in the mitochondria (66, 69).

Surprisingly, we did not find any enrichment of immune-response-related pathways in PBMC of HS cows in comparison to both control groups. However, only the comparison between HS and PF cows revealed a lower expression of genes involved in the T cell receptor signaling pathway (CD28, ICOS, CD8A, and LOC100296565 (T cell receptor alpha variable 18), LOC100300510 (T cell receptor beta-1 chain C region)), antigen processing and presentation (CD8A, LOC100296565, LOC100300510, HSPA8, and NFYA), and the intestinal immune network for IgA production (CD28, ICOS, LOC100296565, LOC100300510, and AICDA) in HS cows. The genes CD28 and Inducible T Cell Costimulator (ICOS) belong to the same T cell surface receptor family. ICOS encodes an important T cell enhancer and is responsible for differentiation, proliferation, and cytokine production, mediates the interaction between T and B cells, and promotes antigen secretion by B cells (70). It is known that the differentiation of T cells into memory and effector cells is supported by metabolic pathways (71). A small portion of T cells differentiates into memory T cells that is primed by fatty acid oxidation and mitochondrial metabolism to sustain the energetic requirements (71). The lower activity of the lipid metabolism could provide fewer fatty acids to promote T cell differentiation, which could partially explain the inactivated T cell receptor signaling pathway of PBMC containing T cells and immune suppression during heat stress. Zachut et al. (2020) showed that summer-calving cows had less CD8⁺ cytotoxic T cells and CD335⁺ natural killer cells than winter-calving cows (11), which could also be a cause of lower proliferation capacity during heat stress (15). As a limitation of our study, we did not differentiate the PBMC population into B cells, T cells, and cytotoxic or natural killer cells. More research is required to distinguish between the different cell types and their function during heat stress to identify potential targets for the development of new treatment strategies.