Heat stress directly impairs gut integrity and recruits distinct immune cell populations into the bovine intestine

Significance Consequences of heat stress, particularly for the immune system and the intestinal health of mammals, are a topic of increasing global relevance due to rising temperatures and potential health impairments. Specific climate effects, however, are often difficult to discriminate from indirect consequences, e.g. reduced feed intake. Our study in dairy cattle, which are particularly sensitive to heat, identifies the infiltration of the small intestinal epithelium by a previously unobserved distinct cell population with macrophage-like phenotype in response to moderate heat stress. By using a pair-feeding design, we attributed these effects as direct consequences of heat stress via impaired intestinal barrier function. Therefore, an appropriate gut function is an important component in combating the negative consequences of heat stress.


Animals and experimental design
Ten German Holstein cows were genotyped for HSP70.1 5´UTR 895 (reference sequence M98823.1; (1)). Three different genotypes were identified and equally assigned to heatstressed (HS, n = 5) or pair-feeding (PF, n = 5) group. Pair feeding served as a control to consider the main effect to heat stress, because HS cows showed reduced feeding behavior (2). All cows were in established 2 nd lactation (HS: week 28 ± 8; PF: week 39 ± 16; P > 0. 3). Animals were adapted to the climate chamber at thermoneutral conditions (15°C) for six days and received a total mixed ration twice daily at 0700 h and 1500 h (3). As described recently (4), five HS cows were subsequently exposed for four days to 28 °C with 52 ± 2 % relative humidity (RH) resulting in a temperature-humidity index (THI) of 76 with ad libitum feeding. Cows had free access to water, feed and water tempered to 28 °C. The reduction of daily ad libitum intake of HS cows was calculated as percentage of the mean daily feed intake to provide the same amount of feed energy to PF cows under thermoneutral conditions. The five PF cows were exposed for four days to 15 °C with 63 ± 1 % RH and THI of 60. After four days, rectal temperature of HS or PF amounted to 40.2 and 38.4 °C, respectively. Animals were sacrificed in the institutional slaughterhouse, and jejunum samples and jejunum mucosa scrapings were taken, frozen in liquid nitrogen and stored at -80 °C until later analysis. For histological analyses, tissue samples were punched out from the jejunum wall, embedded in Tissue-Tek (Sakura Finetek, Zoeterwoude, Netherlands) and stored at -80 °C. For whole transcriptome expression analyses, we included a jejunum whole mucosa sample from a 80 days old calf essentially as described in Hammon et al. (5), as well as jejunum samples from four age-matched lactating control cows in their second lactation, fed ad libitum a total mixed ration and kept at thermoneutrality (details in Weikard et al. (6)). All procedures were approved by the ethics committee of the State Government in Mecklenburg-West Pomerania, Germany (LALLF M-V/TSD/7221.3-1.1-074/12, LALLF M-V/TSD/7221.3-2.1-010/03) and the local department for animal welfare affairs (Landesuntersuchungsamt, Koblenz, Germany (23 177-07/G 13-20-069).

Intestinal histology and image analysis
Embedded frozen tissue was cut (4 µm thick) on a cryostat microtome (Leica CM3050S, Bensheim, Germany). Three sections from each animal were stained with hematoxylin and eosin (H/E; Hematoxylin, Dako, Germany; Eosin, Chroma Gesellschaft, Münster, Germany) according to standard procedures and visualized by using an Olympus BX43 microscope (Olympus, Hamburg, Germany) equipped with a UC30 color camera and Cell^D imaging software (OSIS, Münster, Germany). Ten villi heights and ten crypt depths per sample section (magnification 10x) were measured with the interactive measurement module of Cell^D software comparable to the procedures described by Pearce et al. (7). The average of 30 villi heights and crypt depths were calculated and reported as one number per cow. The closed polygon function within the interactive measurement module of Cell^D software was used to measure the area of submucosa and infiltrating cells in all available H/E stained sections. The area fraction of infiltrating cells was calculated. The measurement of infiltrating cells in mucosa, Muscularis mucosa and submucosa was performed in unstained sections. A macro program was developed to obtain repeatable results. After loading the image, the green channel was extracted and a region of interest (ROI) was defined, comprising the submucosa, Muscularis mucosa, and mucosa within the image. Then, a threshold was set to detect infiltrating cells. False detected particles could be deleted interactively. The area of the ROI and all detected cells were measured in ten images per sample. The area fraction was calculated from the total area of ROIs measured and the area occupied by infiltrating cells.

Laser capture microdissection (LCM)
Infiltrating cells in the jejunum of HS cows were collected with LCM for subsequent identification according to the procedure basically as described by Albrecht et al. (8). Cryo-sections, 12 µm thick, were cut (CM 3050S, Leica), transferred to PALM membrane slides (PALM, Bernried, Germany) and air-dried for 10 s. Sections were dehydrated by placing the slides in different concentrations of ethanol (1 x 70 %, 1 x 96 %, 2 x 100 %) for 30 s each and in xylene for 5 min. Finally, the sections were airdried under a fume hood for 5 min. Cells were cut and collected in adhesive caps (PALM) with a PALM MicroBeam laser capture microdissection device. The total cut areas were 145.6 µm², 128.0 µm², 23.1 µm², and 96.2 µm² for samples LCM1, LCM2, LCM6, and LCM7 (Datasets S1-S3), respectively. Assuming an average cell size of 150 µm² (measured in immunofluorescence images), the estimated cell number in each sample corresponded to 970, 850, 150, and 640 cells, respectively. The actual number was approximately 10 % lower because of loss of particles during catapulting into the adhesive caps.

RNA isolation
RNA isolation from the whole jejunal mucosa scrapings was performed essentially as described by Hammon et al. (5). RNA concentration was determined via Qubit analysis (Thermo Fisher Scientific, Berlin, Germany). RNA integrity was checked via Bioanalyzer 2100 (Agilent Technologies, Waldbronn, Germany) and potential DNA contamination according to Weikard and colleagues (9). Total RNA of the infiltrating cells isolated by LCM was prepared and purified using either the RNeasy Micro Kit (Qiagen, Hilden, Germany) or the Nucleo Spin RNA XS Kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's specifications and stored at -80 °C until further processing.

Library preparation for LCM-based RNAseq and for global jejunal mucosa RNAseq
Four RNAseq libraries were prepared from the infiltrating cells isolated by LCM using NuGen's Ovation SoLo RNA-Seq Kit (NuGen Technologies, San Carlos, CA, USA). According to manufacturers' recommendations, a qPCR step was performed on the LightCycler 96 (Roche Diagnostics, Mannheim, Germany): prior to the final library amplification, the number of PCR cycles optimal for amplification was determined. Finally, all libraries were amplified using 19 cycles before being sequenced. In addition to the four LCM-based RNAseq libraries of infiltrating cells, a RNAseq library from 100 pg calf whole jejunal mucosa total RNA was prepared with the same protocol used for the dissected infiltrating cells to serve as positive control for low RNA input and for subsequent comparison of gene expression between dissected infiltrating cells and jejunal mucosa. The final five RNAseq libraries were analyzed for correct size distribution with an Agilent Bioanalyzer 2100 (Agilent Technologies, Waldbronn, Germany), quantified using a Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA) and sequenced in a multiplex design on the Illumina HiSeq2500 platform using 2x100 bp paired-end sequencing cycles and standard Illumina primers. Comparing the LMC samples and the whole jejunum calf mucosa sample processed with the same library preparation method, 215 genes showed an expression above 5 FPKM in all LCM samples and an average expression across all LCM samples exceeding 10-fold the expression level in the control mucosa (Dataset S2). Among the 215 genes identified in the initial numerical analysis, 155 were also contained within the 3,076 significantly (q < 0.05) differentially expressed genes identified when comparing the LCM samples and the whole jejunum mucosa samples of the age-matched cows demonstrating a very good agreement between analyses. Analogous to the RNAseq libraries generated for the infiltrating cells (NuGen's Ovation SoLo RNAseq), stranded libraries from the entire jejunum mucosal RNA were prepared from four control cows age-matched to the HS cows and kept at thermal neutrality to establish a reference profile of regular tissue from unchallenged animals. Libraries were prepared using the TruSeq Stranded Total RNA Ribo-Zero H/M/R Gold Kit (Illumina, San Diego, USA). This protocol also included a ribo-depletion step, but did not select for polyA-tailed molecules. Quality control and sequencing of the resulting libraries was performed as described above for libraries generated using NuGen's Ovation SoLo protocol.

RNAseq bioinformatic analyses
The resulting raw reads from NuGen's Ovation SoLo RNAseq and whole jejunum mucosa Illumina Ribo-Zero RNAseq were demultiplexed, converted into FASTQ file format with Illumina's bcl2fastq2 conversion software (v2.17.1.14) and checked for quality using the FastQC tool (v0.11.5; (10)). Prior to the alignment, we removed adapter sequences and for the NuGen's Ovation SoLo generated reads also the first 5 bp of the forward reads as recommended in NuGen's manual using cutadapt (v1.13; (11)). In addition, subsequent quality trimming applying qualtrim removed low quality reads (https://bitbucket.org/arobinson/qualitytrim). Finally, the processed reads from the infiltrating cells as well as for the jejunum whole mucosa were mapped to the NCBI bovine reference genome assembly UMD3.1 using HISAT2 (v2.0.3; (12)) in the downstream transcript assembly mode (dta) with the Ensembl Bos taurus genome annotation v87 together with the strand-specific option and visualized using the Integrative Genomics Viewer (v2.4.7; (13,14)). Gene expression was quantified from the generated bam files by the featureCounts routine from the Subread package (v1.5.2; (15)) applying the Ensembl Bos taurus genome annotation v87 and by subsequent in-house R scripts. As a first step comparing gene expression between samples processed with the same NuGen's Ovation SoLo protocol, genes were filtered for FPKM values of at least 5 and a 10-fold higher average expression in all infiltrating cell samples compared to the low input control mucosa. In addition to the initial tentative expression analysis between infiltrating cell samples and a jejunal whole mucosa low input control sample, all processed with the identical library preparation method, a second control data set was included for final differential expression analysis. This second control data set comprised the global transcriptomic ribo-depleted non-polyA selected reads from jejunal whole mucosa of four cows age-matched with the HS cows from which the infiltrating cells had been sampled by LCM. We used raw read counts as determined by featureCounts with the edgeR package (16) to determine potential differences in gene expression between infiltrating jejunal cells of HS cows and jejunum whole mucosa of age-matched control cows. The Biomart platform (https://www.ensembl.org/info/data/biomart/index.html) transferred Ensembl identifiers to gene names. In addition, Ensembl identifiers without gene name assignment in Biomart were manually assigned gene names according to the NCBI Bos taurus UMD3.1.1 annotation (accession 2018-04-28). Lists of genes significantly higher expressed in the infiltrating cell samples and genes significantly higher expressed in the cow jejunum mucosa were separately submitted to enrichment analyses via Ingenuity Pathway Analysis (IPA, Qiagen Inc., https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis, (17)) to obtain information about overrepresented pathways. High purity of the collected cells was confirmed by RNAseq data demonstrating that markers of epithelial cells in the small intestine e.g. CD13/ANPEP (18) or genes known to be highly and specifically expressed in small intestinal mucosa (FABP6, https://gtexportal.org/home/gene/FABP6, (19)) displayed no expression in isolated cells (Dataset S1). However the respective genes were substantially transcribed in whole mucosa tissue (Dataset S3).

Immunohistochemistry
For immunohistochemical analysis, CD3, CD21, CD163 and CD172a were selected as markers for specific immune cell populations. Antibodies directed against these markers had been demonstrated to recognize the proteins of bovine origin (20)(21)(22)(23). Tissue sections were cut 10 µm thick, fixed for 15 min with 4 % paraformaldehyde in PBS and washed. Sections were incubated with either Alexa Fluor 488 labeled rat anti-human CD3 (1:50, AbD Serotec) or FITC labeled mouse anti-human CD21 antibody (1:10, AbD Serotec) at RT for 1 h in a humidity chamber. Slides were washed 3 x 10 min in PBS and nuclei were counterstained with Hoechst 33258 (1:10,000 in PBS; Sigma-Aldrich, St. Louis, MO, USA). After washing, slides were covered using ProLong Diamond Antifade Mountant (Thermo Scientific) and appropriate cover-slips. Fluorescence was visualized with a Nikon Microphot SA fluorescence microscope (Nikon, Tokio, Japan) and an image analysis system equipped with CELL^F software and a CC-12 high resolution color camera (OSIS). The number of CD3 + and CD21 + cells was counted in ten randomly selected villi and the respective villus areas were measured. The number of positive cells divided by the area of the villus (cells/mm 2 ) was determined for each animal. Additionally, CD163 and CD172a antibodies were utilized to characterize infiltrating cells. Tissue sections were cut 10 µm thick with a CM3050S cryostat microtome (Leica), air dried for 15 min, fixed with 4 % paraformaldehyde in PBS, and washed 3 x 5 min in PBS. Unspecific binding of the secondary antibody was blocked using 10 % goat serum in PBS for 15 min. Slides were incubated for 1 h at room temperature (RT) with either mouse anti-human CD163 (1:50, AbD Serotec, Puchheim, Germany) or CD 172a (1:10; (21)), and subsequently washed 3 x 10 min with PBS. Specific binding of the primary antibodies was detected with the respective Alexa Fluor 488 goat anti-mouse IgG (Thermo Fisher Scientific, Waltham, MS, USA) or MFP 488 goat anti-mouse IgG (Mobitec, Göttingen, Germany) secondary antibody (1:500, 45 min at RT in the dark). Slides were washed 3 x 10 min in PBS and nuclei were counterstained with Hoechst 33258 (1:10,000 in PBS; Sigma-Aldrich). After washing 2 x 5 min with PBS and 1 x 5 min with distilled water, slides were covered using ProLong Diamond Antifade Mountant (Thermo Scientific) and appropriate cover-slips. Tight junction proteins were detected using antibodies against ZO-1 (1:50, GTX108592 Genetex) and ZO-2 (1:50, sc-515115 SantaCruz) following the protocol described before, except that both antibodies were incubated at 4°C overnight. Specific binding of the primary antibodies was detected with an Alexa Fluor 488 goat anti-rabbit IgG (Thermo Fisher Scientific) and a MFP 488 goat anti-mouse IgG (Mobitec) secondary antibody (1:500, 45 min at RT in the dark), respectively.

Cytokine ELISA
Potential effects of heat stress on the local immune status were investigated by quantifying interleukin-1β (IL-1β) and interleukin-4 (IL-4) in the jejunal mucosa. Sixty mg mucosa was homogenized in ice-cold phosphate buffered saline (PBS) containing: 0.4 M NaCl, 0.05 % Tween 20, 0.5 % BSA, 0.1 mM phenylmethylsulphonyl fluoride, 0.1 mM benzethonium chloride, 10 mM EDTA and 20 KI/ml aprotinin (Sigma-Aldrich). The homogenates were centrifuged at 10,000 rpm for 20 min at 4 °C and protein concentrations of the supernatant were determined using the Bradford method. The concentration of IL-1β and IL-4 in jejunum mucosa extracts was determined by capture ELISA performed on 96-well microtiter plates according to manufacturer's instructions (Bovine IL-1β ELISA, ESS0027, Bovine IL-4 ELISA, ESS0031; Thermo Fisher Scientific). Cytokine concentrations were calculated from a standard curve and expressed as pg/ml (Tecan SUNRISE 96-well microplate reader, Tecan, Männedorf, Switzerland). Data were normalized to total protein concentration (pg/mg) of the homogenate.

Jejunal inflammation and stress markers
Mucosa samples were analyzed for lysozyme activity using EnzChek Lysozyme Assay kit (Thermo Fisher Scientific) as previously described (7). The assay detection range was 0 to 500 U/ml and measurements were taken after 60 min of incubation. Tissue alkaline phosphatase was measured with a commercial kit (QuantiChrom Alkaline Phosphatase, BioAssay Systems, Hayward, CA, USA) according to the manufacturer´s instructions. Catalase activity measurement was performed as previously described (26) using a Catalase Assay kit (Cayman Chemicals, Ann Arbor, MI, USA). All concentrations determined were normalized to the total protein concentration (Bradford, Thermo Fisher Scientific) of the tissue homogenates.    Additional data tables (separate files) Dataset S1. Analysis of FPKM values for all genes listed in the Ensembl annotation v87 in infiltrating cells of heat-stressed cows. Dataset S2. Comparison between the 215 differently expressed genes within NuGen Ovation SoLo libraries (infiltrating cells of heat-stressed cows vs. whole jejunum control mucosa) and the edgeR-based calculation of differential expression between infiltrating cells in HS cows and the whole jejunum mucosa of age-matched cows. Dataset S3. Differential transcriptome analysis by edgeR between infiltrating cells in jejunum in response to heat stress and whole jejunal mucosa from age-matched control cows. Dataset S4. Full list of canonical pathways significantly enriched with genes significantly higher expressed in infiltrating cells in heat-stressed cows than in the whole jejunum mucosa of age-matched control cows. Dataset S5. Full list of upstream regulators with significant overlap with genes significantly higher expressed in infiltrating cells in heat-stressed cows than in the whole jejunum mucosa of age-matched control cows.