Multi-tissue transcriptomic profiling reveals the internal physiological landscape of laying hens in cage and cage-free systems

Abstract

Welfare-friendly housing systems for laying hens, such as cage-free, have become prevalent. However, the physiological effects of housing systems on the laying hens remain poorly understood. Here, we compared behavioral characteristics and transcriptomic profiles from 90 multi-tissue samples among three housing systems: battery cage (BC), barn (BR), and free-range (FR). BR and FR housing promoted behavioral diversity compared to cages. Transcriptome analysis of central tissues (diencephalon and cerebral hemisphere) and peripheral tissues involved in egg production (liver, ovary, oviductal segments of magnum, and uterine) revealed significant enrichment of insulin resistance-related pathways in both diencephalon and liver of BC hens, and enhanced norepinephrine signaling in the cerebrum of BR and FR hens. To validate these findings, we performed a glucose tolerance test to assess insulin sensitivity and quantified the cerebral norepinephrine concentrations by ECD-HPLC. The results showed that BR and FR hens tended to exhibit higher insulin sensitivity and enhanced norepinephrine signaling compared with BC hens. Taken together, our findings suggest that housing conditions markedly shape the internal physiological landscape of laying hens, and also that environment-enriched cage-free contributes to improving metabolic and neurophysiological signaling.

Introduction

Animal welfare has become a major global trend and embodied from a conceptual framework into legally binding regulations and guidelines across various countries. The World Organization for Animal Health (WOAH) established international welfare standards for farm animals in 2002. The draft WOAH standards for laying hens emphasize the importance of allowing highly motivated behaviors and the recommendation of equipment of resources such as perches and nest boxes (WOAH, 2019). The European Union (EU) has adopted more stringent standards than WOAH and implemented them through enforceable legislation. Battery cages (conventional cages) have been prohibited since 2012 in EU, and the use of all cage systems is gradually being phased out. In the United States of America (USA), ten states have passed legislation banning all cage systems, and in some of these states, the sale of eggs from cage systems is also prohibited (USDA, 2022). Additionally in the USA, numerous multinational corporations have pledged to transition to cage-free eggs by 2020–2025, aligning with the “Global Investor Statement on Farm Animal Welfare” endorsed by supporting investors (BBFAW, 2022). Therefore, cage-free has become the predominant housing system in both the EU and the USA.

Cage-free and enriched cages for laying hens are designed to promote normal behaviors and behavioral diversity by providing various environmental enrichments such as nest boxes, litter area, and perches (Wood-Gush and Gilbert, 1969; Wood-Gush and Gentle., 1978; Blokhuis, 1984; Baxter, 1994; Olsson and Keeling, 2000; Cooper and Appleby, 2003; Weeks and Nicol, 2006; Jong et al., 2007; Nicol et al., 2011; EFSA, 2015; Hemsworth and Edwards, 2021). Actually, hens in cage-free perform the highly motivated behaviors more frequently and exhibit more behavioral repertoire compared to cage systems (Blokhuis et al., 2007). On the other hand, cage-free has disadvantages in some welfare indicators, including mortality due to cannibalism, feather pecking, keel bone break or deformation, and bumble foot (Tauson, 2005; Blokhuis et al., 2007; Fossum et al., 2009; Shimmura et al., 2010; Hester, 2014; Blatchford et al., 2016; Heerkens et al., 2016; Saraiva et al., 2019; Nenadovic et al., 2022). Therefore, cage and cage-free have advantages and disadvantages in each welfare indicator, which have been well documented based on phenotypic indicators such as behavior, physical condition, and egg production. However, the molecular mechanism underlying the internal physical states —driving the integrated these phenotypic traits— remain poorly understood.

RNA sequencing (RNA-seq) is a widely used method for quantifying gene expression changes across multiple mammalian tissues under different environmental conditions. This approach captures the molecular alterations induced by environmental variation (Hadadi et al., 2022; Deota et al., 2023; Yan et al., 2024) and reveals tissue-specific gene expression patterns in mammals (Breschi et al., 2020; Sonawane et al., 2017; Wang et al., 2019). In avian studies, transcriptome analyses have primarily focused on individual chicken tissues from different genetic backgrounds, providing important insights into gene regulatory pathways associated with production traits such as body weight gain, meat quality, and egg production (Kong et al., 2017; Wang and Ma, 2019; Ma et al., 2021). However, RNA-seq studies exploring internal physiological responses to environmental variation remain limited in laying hens.

In this study, we first characterized phenotypic traits of laying hens in cage, barn, and free-range, and then performed transcriptome analyses of six tissues to further elucidate the physiological effects of different housing systems. Specifically, we compared gene expression profiles among the three housing systems in six tissues—cerebral hemisphere, diencephalon, liver, ovary, and oviductal segments of magnum and uterine—all of which are associated with behavior or egg production. To validate the transcriptomic findings, we also measured monoamine concentrations in the cerebral and performed a glucose tolerance test to assess insulin responsiveness. Although some housing systems used in this study were not fully commercial (e.g, small-scale cage-free systems) and this study may be considered a pilot investigation under controlled conditions, the limited number of laying hens allowed precise behavioral measurements, individual-level transcriptomic analyses, and a more direct integration of phenotypic and molecular data.

Materials and methods

Ethics statement

Animal experiments were conducted in accordance with the guidelines set by the Animal Experiment Committee of Azabu University (Certification No. 200612-3 and 230613-7).

Animals and housing arrangement

We conducted two experiments: phenomics and transcriptomic analysis (Experiment 1) to formulate a hypothesis, followed by a second experiment (Experiment 2) to test it.

In total, 78 brown layers (Momiji; Goto Hatchery Inc., Gifu, Japan) were used. They received infrared beak treatment at day 0 and were reared in cages. At the age of 17 weeks, the hens were randomly introduced into one of three housing systems: battery cage (BC), barn (BR), or free-range (FR). Fifteen birds were housed in BC, with five birds per cage, and in BR and FR, with twelve birds each in each experiment. All housing systems were equipped within the environmentally controlled room with room temperature of 23 ± 1 °C, exposed to 14 h of light/day from 5:00 to 19:00. The temperature of the outdoor area depended on ambient conditions and averaged approximately 19.6 ± 7 °C during the experimental period. Lighting was provided by fluorescent lights, adjusted to give an intensity of 30 lux at the food troughs. The hens had ad libitum access to water and feed. The feed contained approximately 170 g CP and 11.92 MF ME per kg, 2.7 % calcium, and 0.45 % available phosphorus. Feeding and other routine work, such as collecting eggs, was done at 8:00 and 17:00. Management conditions, including environmental temperature, lighting intensity, and diet formulation, were maintained within the ranges recommended in previous guidelines (Manser, 1996; Hy-Line International, 2016, 2021; Bryden et al., 2021).

The design and equipment of all systems fulfilled the UEP Certified Guidelines (UEP, 2023, 2024). All systems were not large commercial scale but small for individual behavior observation and sample collection. The battery cage (BC) was a laying cage 96 × 40 × 42 cm (wide × deep × high) at the rear. The BC provided 755 cm² with 21 cm feeder and drinker space per hen. The barn (BR) was floor rearing and a large enclosure 4.6 × 2.5 m, providing a total floor area of 0.95 m² per hen. The floor of BR was supplied with wood-shavings. The nest box (764 cm²/hen) was provided on the floor, and two wooden perches (32.8 cm/hen) were placed 1.24 m above the floor. The feeder and drinker space were 17.5 cm per hen. The free-range (FR) was an BR with an outdoor area 3.55 × 5.10 m, providing an additional floor area of 1.5 m² per hen. A passage hole (1 × 1 m) was provided between indoor and outdoor areas with a slope that hens could readily go outside. The outdoor area was enclosed by wire-mesh to allow sunlight exposure, and the floor was made of concrete to prevent differences in ingested materials across housing systems. Additional five perches (62.8 cm/hen) and three drinkers (8.16 cm/hen) were placed in the outdoor area. Hens were given access to the outdoor area from 8:00 to 17:00.

Behavioral observation

Behavioral observations were conducted from 30 to 33 weeks of age in Experiment 1 and from 54 to 56 weeks of age in Experiment 2.

Direct visual observations of behavior of all hens using continuous sampling were conducted to record the behavior of the birds in each housing system for a total of 4 h/d, 2 h each in the morning (10:00 to 12:00 h) and afternoon (13:00 to 15:00 h) (2 d/housing system). During the observations, the following comfort behaviors were recorded: head scratching, wing and leg stretching, ruffing, wing raise, wing flapping, and tail wagging (Nicol, 2015). The same observers collected all data. Observation of FR hens was carried out at the same time by the two observers, with one observer observing hens in the indoor area and the other in the outdoor area.

Physical conditions

Physical conditions were conducted from 30 to 33 weeks of age in Experiment 1 and 54 to 56 weeks of age in Experiment 2.

Physical condition measurements, consisting of body weight, claw length, comb color, comb condition, feather condition, and foot condition, of all hens were conducted. The center claw length from the claw root to the tip of the right foot was measured using a digital vernier caliper. The comb color at the center of the comb was measured using a spectrophotometer (CR-200b, Konica Minolta, Tokyo, Japan). The spectrophotometer shows the color using 3 color parameters: l-value indicates lightness, a-value red chromaticity, and b-value yellow chromaticity. Comb, feather, and foot conditions were assessed according to the Welfare Quality® protocol (Welfare Quality®, 2019). Comb condition was scored from 1 to 3 (3, no comb pecking wounds; 2, < three wounds; 1, ≥ three wounds). Feather condition at 6 parts of the body (neck, breast, back, belly, wing, and tail) was scored from 1 (denuded) to 4 (no damage). Foot condition was scored from 1 to 3 (3, no inflammation; 2, chronic inflammation not dorsally visible; 1, inflammation of both footpads at dorsally visible). The assessment of comb, feather, and foot conditions was carried out by two observers working together.

Locomotor activity

To precisely compare changes of activity in a day, locomotor activity was automatically recorded using the nano tag® (Kissei comtec Co. Ltd., Matsumoto, Japan) for 48 h from 0:00 of the day in Experiment 2. At 55 weeks of age, hens had nano tags attached to the back feather with plastic tape and were habituated for at least five hours (Fijn et al., 2012). The nano tag® is a small three-axis accelerometer device (14.2 × 18.8 × 7.1 mm) and a non-invasive telemetry system. The nano tag® records a composite waveform calculated from three-axis (X, Y, and Z) acceleration values at 12 s intervals and counts the number of times that the composite waveform passes a predefined threshold. Based on preliminary experiments, the threshold was set at 485 g² to capture large movements such as wing flapping and moving, whereas avoiding the detection of smaller actions such as feeding and preening. We finally obtained 9, 10, and 8 valid data from BC, BR, and FR, respectively. Because hens are typically active before lights-on and after lights-off (Tanaka and Hurnik, 1992), we measured activity during an extended light period from 4:00 to 20:00, which includes one hour before and after the scheduled light-on period in the room (from 5:00 to 19:00) (Fig. 1b).

Fig. 1.

Phenotypic analysis and diurnal change of locomotor activity.

(a) Plot for PC1 and PC2 of phenotypic variation (Supplemental Table 1). (b) Diurnal changes in locomotor activity level for two days. The blue (BC), red (BR), and green (FR) dashed lines represent the average (± S.E.) locomotor activity in each housing system. The dark and light gray backgrounds indicate the lights-off period; dark gray indicates shortened dark period (20:00 to 4:00 the next day), and the light gray shows the extended light period within the lights-off period (19:00-20:00 and 4:00-5:00). The orange background indicates the time for feeding and other routine work. (c, d) Total locomotor activity levels during the extended light period (c) and the shortened dark period (d) (means ± S.E.). Different letters indicate statistically significant differences (P < 0.05).

Production

The number of eggs laid, including cracked eggs and floor eggs, and mortality were recorded daily by 36 weeks of age in Experiment 1. Egg quality traits, including egg weight, Haugh unit, yolk color, albumen height, egg shell thickness, deformation, and color were measured for eggs collected from each housing system at ages of 34 weeks for three consecutive days. Egg production was shown on a hen-d basis (total number of eggs/total number of hens × 100), and egg mass (g egg/hen per d) were calculated using mean the egg weight (g/egg) for the 3-day measurement period. Egg shell thickness (C1012XBS, Mitutoyo Crop., Kawasaki, Japan), egg shell color (CR-200b, Konica Minolta, Tokyo, Japan), and other egg quality parameters (DET 6500, Nabel Co., Ltd., Tokyo, Japan) were measured.

Transcriptomic analysis

Five hens per system were randomly selected, and six tissues were collected from each hen at 36 weeks of age in Experiment 1, resulting in a total of 90 tissue samples. The six tissues included two central nervous system tissues associated with behavior (left cerebral hemispheres and diencephalon) and four peripheral tissues involved in egg production (liver, ovary, oviductal segments of the magnum and uterine) (Martín, 2019). The samples were immediately dissected after euthanasia, flash-frozen on dry ice, and stored at −80 °C until RNA extraction. Total RNA was prepared using the RNeasy® Lipid Tissue Mini Kit (QIAGEN, Hilden, Germany). The quantity and quality of the RNA samples were evaluated using an Agilent 2100 Bioanalyzer.

Libraries from each RNA sample were prepared using the Lasy-seq v1.1 protocol (Kamitani et al., 2019) and sequenced with 150 bp paired-end reads in an Illumina Hi-seqX sequencer. FASTQ files from RNA-seq were pre-processed by removing adapter sequences, selection by read length (≥ 30 bp), and low-quality bases (Q ≥ 20) using Trimmomatic-0.39 (Bolger et al., 2014). The quality of cleaned reads was assessed using fastqc-v0.11.9 (Brown et al., 2017). The per-processed sequences were mapped on the chicken reference genome (https://ftp.ensembl.org/pub/release-96/gtf/gallus_gallus/) using HISAT2 and quantified using RSEM-1.3.3 (Li and Dewey, 2011). The conversion of the output from RSEM to trimmed mean of M values (TMM) and the detection of differentially expressed genes (DEGs) were conducted by edgeR using R (> version 4.0.4). The DEGs were identified using a threshold of log₂ foldchange (FC) > 2 and P < 0.1. Among them, genes with P < 0.05 were further defined as highly differentially expressed genes (highly-DEGs).

A total of 18 lists of DEGs and highly-DEGs were generated from six tissues and three pairwise comparisons among housing systems (BC vs. BR, BC vs. FR, BR vs. FR). To identify the highly-DEGs associated with specific factors—environmental enrichment and sunlight—these 18 lists were grouped into three: (1) ALL DEGs, which include all DEGs from the three comparisons; (2) EE DEGs, defined as the common DEGs between BC vs. BR and BC vs. FR, representing the effect of environmental enrichment; and (3) SUN DEGs, defined as the common DEGs between BC vs. FR and BR vs. FR, representing the effect of sunlight (Fig. 2).

Fig. 2.

Transcriptomic profiling of multi-tissue.

The colored area in each venn diagram shows the range of DEGs, whose gene lists are shown in the right heatmaps. The blue area indicates ALL DEGs (a), the orange EE DEGs (b), and the green SUN DEGs (c). Each transcript of each multi-tissue and housing system is displayed in the row of heatmaps (ALL DEGs: Supplemental Table 4; EE DEGs: Supplemental Table 5; SUN DEGs: Supplemental Table 6). The z-score is shown as the standard deviation of the gene expression value from the mean after normalization. A transcript with positive and negative z-score is represented by green and red color, respectively. Cer; cerebral hemispheres, Die; diencephalon, Liv; liver, Ova; ovary, Mag; oviductal segments of the magnum, and Ute; oviductal segments of the magnum uterine.

To investigate the functional enrichment of DEGs, KEGG pathway enrichment analysis performed on the 18 DEGs lists using DAVID (> version 6.8, Dennis et al., 2003). KEGG pathways with a P < 0.1 were considered significantly enriched.

Monoamines and their metabolites

Fifteen right cerebral hemispheres samples (three housing systems × five replicates), corresponding to the contralateral hemispheres used for RNA-sequencing, were used to measure monoamine components. The following monoamines were measured using ECD-HPLC: Dopamine (DA), 3,4-Dihydroxyphenylacetic acid (DOPAC), Norepinephrine (NE), Homovanilic acid (HVA), Epinephrine (E), 3-Methoxy-4‑hydroxy-phenylglycolaldehyde (MHPG). The frozen brain samples were weighed and homogenized in 0.2 M perchloric acid in an ultrasonic homogenizer. The homogenates were deproteinized on ice for 30 min and centrifuged at 20,000 g for 15 min at 4 °C. After filtration for the supernatant (0.2 μm), the solution of 1 M sodium acetate was added to adjust the pH to approximately 3.0. The samples were infected into an high-performance liquid chromatography (HPLC) system (HTEC-510, Eicom, Kyoto, Japan). Monoamines and their metabolites were separated using an HPLC system at 30 °C on a reverse-phase column (Eicompak SC-50DS, 30 150 mm, Eicom) and were detected by an electrochemical detector. The mobile phase comprised 83% citrate-acetate buffer, 17% methanol with sodium octane sulfonate, and EDTA-2Na. Monoamine and their metabolites concentrations were calculated as pg/mg sample weight. Turnover rates of DA (DOPAC/DA, HVA/DA, and NE/DA), NE (MHPG/NE), and 5-HT (5-HIAA/5-HT), and the ratio of HVA/5-HIAA were also calculated.

Glucose tolerance test

Glucose tolerance test (GTT) was conducted at 57 weeks of age in Experiment 2. Eighteen hens (three housing systems × six replicates) were selected so that average body weight was similar across groups. The birds were fasted for 16 h prior to testing. A glucose solution (10 mL) containing 2 g of glucose per kg of body weight was orally administered using a flexible plastic feeding needle for birds (Simon and Rosselin, 1978). Blood samples were collected from the wing veins using 10 μl heparinized syringe at 0, 15, 30, 60, 90, 120, and 240 min after glucose administration and were immediately placed on ice. The samples were centrifuged at 3,000 × g for 15 min at 4 °C, and the plasma was collected and stored at −80 °C until glucose concentration analysis. The glucose concentration was colorimetrically quantified using a kit (Glucose CⅡ Test Wako, Wako Pure Chemical Industries Ltd., Osaka, Japan). The area under the curve (AUC) for the rate of change in blood glucose concentration after administration was calculated.

Statistical analysis

The score data (comb condition, feather condition, and foot condition) were normalized by transformations (square-root, arcsine square root, log) (Martin and Bateson, 1993). For the locomotor activity data, without normal distribution and homogeneity of variance, a GLM was used assuming a Poisson distribution. The effect of housing systems was evaluated using a Type Ⅱ analysis of deviance with likelihood ratio chi-square tests. For variables measured once per individual, with normal distribution, one-way ANOVA was used to evaluate the effects of the housing system. For repeated measurements, such as the glucose tolerance test, two-way ANOVA was performed to examine the effects of time and the housing system. The significances of the effect of the housing system were evaluated by multiple comparisons using the Tukey-Kramer test for parametric data. Principal component analysis (PCA) of phenotypic data was performed using the prcomp function in R after variable-wise normalization, excluding variables with zero variance across individuals. In Experiment 2, PCA was additionally conducted including locomotor activity data. Data were analyzed using the statistical software Statcel (Yanagii, 2007) and R (>version 4.0.4).

Results

Phenotypic characterization

Phenotypic characteristics, including behavior and physical conditions of hens in three different housing systems, were shown in Supplemental Table 1 (Experiment 1) and Supplemental Table 2 (Experiment 2). In Experiment 1, total comfort behaviors were observed more frequently in BR and FR compared to BC (P < 0.001). The claw length was longer in BC than in BR and FR (P < 0.001). The foot damage was greater in FR than in BC and BR (P < 0.01). For comb color, FR hens showed lower l-value (P < 0.001) and higher a-value (P < 0.05) compared to BC and BR.

Principal component analysis (PCA) score plot to visualize overall phenotypic differences among housing systems showed that the first two principal components (PC1 and PC2) explained 40.6 % of the total variance (23.6 % in PC1 and 17.0 % in PC2, respectively) and separated the three housing systems in Experiment 1 (Fig. 1a). PC1 distinguished between cage and cage-free, whereas PC2 differentiated the BC and BR from FR. The PCA of Experiment 2 also showed a similar separation among housing systems with the pattern observed in Experiment 1 (Supplemental Figure 1).

As expected, the activity data automatically recorded using the nano tag® showed that hens exhibited higher locomotor activity during the light period and were largely inactive during the dark period in all housing systems (Fig. 1b). However, during the extended light period, BR and FR hens showed significantly higher activity levels than BC hens (Fig. 1c; P < 0.0001). In contrast, during the shortened dark period, BC hens were more active than BR hens (Fig. 1d; P < 0.0001).

Transcriptomic profiling

A total of 2336 annotated genes were identified as DEGs in all comparisons, including 558 genes in the cerebral, 611 in the diencephalon, 173 in the liver, 801 in the ovary, 456 in the oviductal segments of magnum, and 351 in the oviductal segments of uterine (Supplemental Table 3). Based on highly-DEGs across comparisons, 1694 genes were identified as the ALL DEGs, including 364 in the cerebral, 382 in the diencephalon, 249 in the liver, 432 in the ovary, 303 in the oviductal segments of magnum, 229 in the oviductal segments of uterine (Fig. 2a, Supplemental Table 4). Based on comparison between BC and BR or FR, 127 genes were identified as EE DEGs (environmental enrichment-related DEGs), including 30 in the cerebral, 25 in the diencephalon, 23 in the liver, 24 in the ovary, 18 in the magnum, and 10 in the uterus (Fig. 2b, Supplemental Table 5). Based on comparison between BC or BR and FR, 114 genes were identified as the SUN DEGs (sunlights-related DEGs), with 23 in the cerebral, 25 in the diencephalon, 15 in the liver, 20 in the ovary, 16 in the magnum, and 15 in the uterus (Fig. 2c, Supplemental Table 6).

KEGG annotation and enrichment analysis

The DEGs in the cerebrum identified from comparisons of BC vs. BR, BC vs. FR, and BR vs. FR were mapped to 9, 7, and 7 KEGG pathways, respectively (Fig. 3a–c). The neuroactive ligand-receptor interaction pathway was significantly enriched in all three comparisons (P < 0.1). In addition, the calcium signaling pathway was enriched in both comparisons, BC vs. BR and BC vs. FR (P < 0.5). In the diencephalon, DEGs from BC vs. BR and BR vs. FR comparisons were mapped in 8 and 9 pathways (Fig. 3d, e, Supplemental Table 7). The insulin resistance pathway, related to insulin signaling, was significantly enriched (P < 0.1) in both comparisons BC vs BR and BR vs. FR. In the liver, DEGs from BC vs. BR comparison were mapped to 5 pathways (Fig. 3f). The adipocytokine signaling pathway, which regulates glucose and lipid metabolism and is linked to insulin signal, was enriched (P < 0.1).

Fig. 3.

KEGG pathway analysis.

In the bubble plot, each circle and horizontal axis indicate the size of factors affecting the pathway (bigger circle represent bigger impacts). Circle color indicates the P-value of enrichment analysis (darker colors represent higher statistical significance).

KEGG pathway analysis of the ovary, oviductal segments of magnum, and uterine were showed in Supplemental Table 7. In the ovary, DEGs from comparisons of BC vs. BR, BC vs. FR, and BR vs. FR were mapped to 13, 6, and 18 KEGG pathways, respectively (P < 0.05) (Supplemental Table 7). The neuroactive ligand-receptor interaction pathway was significantly enriched in all three comparisons (P < 0.05). The cytokine-cytokine receptor interaction was significantly enriched in both comparisons BC vs. BR and BC vs. FR (P < 0.05), and intestinal immune network for IgA production was significantly enriched in comparisons BC vs. BR (P < 0.01). In addition, the cell adhesion molecules and tight junction were enriched in comparison to BR vs. FR (P < 0.01 and P > 0.1, respectively). In the oviductal segments of magnum and uterine, DEGs from all comparisons were mapped 36 and 22 pathways, respectively. In both tissues, the neuroactive ligand-receptor interaction pathway and calcium signaling pathway were enriched in all three comparisons (P > 0.1).

Norepinephrine signaling in the cerebral

Since the transcriptomic analysis of the cerebral suggested alteration in neurotransmission, we focused on monoaminergic signaling pathways, which were also implicated in the KEGG pathway annotations for neuroactive ligand-receptor interaction and calcium signaling (Fig. 3a, b, and c). When we examined the expression of genes involved in tyrosine metabolism and norepinephrine receptors signaling in the RNA-seq data of the cerebral (Fig. 4a, b), PNMT (phenylethanolamine N-methyltransferase) was significantly down-regulated in FR than BC (|log₂ FC| = 1.64 and P < 0.001). In contrast, ADRB1 (β−1 adrenergic receptor) was significantly upregulated in both BR and FR than BC (BC vs. BR: |log₂ FC| = 2.68 and P < 0.05; BC vs. FR: |log₂ FC| = 2.81and P < 0.05). In addition, CRACR2B (calcium release activated channel regulator 2β) expression was significantly higher in cage-free hens (BC vs. BR: |log₂ FC| = 3.83 and P < 0.05; BC vs. FR: |log₂ FC| = 3.74 and P < 0.01) (Fig. 4a).

Fig. 4.

Norepinephrine-rerated gene expression and monoamine concentration in the cerebral.

(a) Gene expression of PNMT, ADRB1, and CRACR2B in RNA-seq of cerebral. (b, c) Norepinephrine (NE) concentration and norepinephrine turnover (MHPG/NE ratio) in the cerebral. *P < 0.05. (d) Possible mechanism for differences in norepinephrine signaling in cage and cage-free.

The upregulation of ADRB1 and CRACR2B in the cerebral of BR and FR hens, in contrast with increased PNMT expression in BC hens, suggests potential alterations in norepinephrine signaling between housing systems. To test the hypothesis, we measured concentrations of monoamines and their metabolites in the cerebrum using HPLC (Supplemental Figure 2). Although E was not detected, the concentration of NE was significantly higher in BC hens compared to FR hens (Fig. 4b; P < 0.01). No significant differences were observed in the concentrations of DA, DOPAC, HVA, and MHPG (Supplemental Fig. 2). However, the norepinephrine turnover, assessed by the MHPG/NE ratio, was significantly higher in FR hens than BC hens (Fig. 4c; P < 0.01).

Insulin signaling and insulin sensitivity

Since the transcriptomic analysis of both diencephalon and liver suggested alteration in insulin-related signaling, we examined the expression of insulin-related genes in the RNA-seq data (Fig. 5a and b). In the diencephalon, MLXIPL (MLX Interacting Protein like), RPS6KA2 (Ribosomal protein S6 kinase alpha-1), and RPS6KA3 (Ribosomal protein S6 kinase alpha-3) expressions were significantly higher in BC hens compared to BR (|log2 FC| = 4.52 and P < 0.05; |log2 FC| = 4.78 and P < 0.01; |log2 FC| = 2.27 and P < 0.05; respectively). In contrast, CARTL (Cocaine And Amphetamine Regulated Transcript Like), which is involved in appetite regulation and energy homeostasis, was highly expressed in BR hens (|log2 FC| = 3.59 and P < 0.01). In the liver, PCK1 (Phosphoenolpyruvate carboxykinase 1) expression was significantly upregulated in BC hens relative to BR (|log2 FC| = 2.30 and P < 0.01) (Fig. 5a and b).

Fig. 5.

insulin-related genes expression and insulin sensitivity.

(a, b) Gene expression of CARTL, MLXIPL, RPS6KA2, and RPS6KA3 in the diencephalon (a) and PCK1 in the liver (b). *P < 0.05. (c) Change of plasma glucose concentrations after oral glucose administration. (d) Area under the curve (AUC) calculated based on change in blood glucose after glucose administration. (e) Possible mechanism for differences in insulin signaling in the diencephalon and liver between BC and BR. Blue arrows indicate characteristics of insulin or appetite regulation of BC hens.

The observed upregulation of MLXIPL, RPS6KA2, and RPS6KA3 in the diencephalon, together with the increased expression of PCK1 in the liver of BC hens, suggests a tendency toward insulin resistance. To test the hypothesis, a glucose tolerance test was performed (Fig. 5c and d). Although no significant differences were found, plasma glucose concentrations of BC hens tended to remain elevated from 30 min after oral glucose administration through 240 min, compared with those of BR hens. In FR hens, plasma glucose concentrations also tended to elevate from 30 min but then decreased sharply from 90 min (Fig. 5c and d).

Discussion

Phenotypic characteristics

The differences in behaviors and physical conditions between cage (BC) and cage-free (BR and FR) systems were generally consistent with the previous studies (Fig. 1). For example, comfort behaviors—well-known indicators of animal welfare (Dawkins and Hardie, 1989; Norgaardnielsen, 1990; Shipov et al., 2010)—were observed more frequently in BR and FR hens than in BC hens. These results are in agreement with previous reports that the frequency of comfort behaviors is influenced by the amount of usable space, particularly in environments that provide larger or multi-dimensional areas for movement (Dawkins and Hardie, 1989; Norgaardnielsen, 1990; Widowski et al., 2016; Hemsworth and Edwards, 2021). In physical conditions, BC hens had longer claws than BR and FR hens, consistent with previous studies showing that the presence of litter or perches can reduce claw length through natural wear (Taylor and Hurink, 1994; Blokhuis et al., 2007). The comb color of FR hens was darker than that of BC and BR hens, which also corresponds to previous studies (Whay et al., 2007; Monique and Jan-Paul, 2014). Principal component analysis (PCA) using these phenotypic variables showed that the individuals were clearly clustered by housing systems, indicating that each housing system exhibited distinct phenotypic profiles (Fig. 1a). In the PCA plot, PC1 separated the cage (BC) and cage-free (BR and FR), suggesting that the degree of behavioral diversity contributes to the phenotypic differentiation in welfare assessment. PC2 further distinguished FR from BC and BR, implying that exposure to sunlight had an additional influence on the phenotype.

The locomotor activity was higher in BR and FR hens than BC hens during the extended light period (Fig.1b and c), indicating that the restricted space in the cage system limited movement. Shimmura et al. (2024) automatically quantified the behaviors of hens using a 3-axis accelerometer-based wearable inertial sensor and reported that behavioral diversity was greater and resting was observed more frequently in cage-free compared with cages. On the other hand, small steps and stopping without sleeping were observed more frequently in cages due to disturbances from cage mates (Shimmura et al., 2024). However, because the use of a multi-axis inertial sensor requires substantial battery capacity, the observation period was limited to a total of only four hours. In the present study, we captured dynamics of behavioral changes in the three housing systems continuously for two days by simplifying the computation and reducing battery consumption (Fig. 1b). Our data showed that small steps occurred more frequently in BC hens during the shortened night period, resulting in higher activity at nighttime in BC hens than BR hens (Fig. 1d). This behavioral pattern suggests that resting in caged hens is sometimes disturbed by cage mates within restricted space. In contrast, cage-free hens rested on elevated perch (Shimmura et al., 2024), resulting in longer sleeping during the night period (Fig. 1d).

Transcriptomic profiling of peripheral tissues

Hierarchical clustering based on TMM-normalized data revealed that the transcriptomic landscapes of multiple tissues in laying hens varied markedly depending on the housing system (Fig. 2a). In the PCA using the phenotypic measurements, PC1 and PC2 corresponded to the effect of environmental enrichment (EE) and sunlight (SUN), respectively (Fig. 1a). Based on these components, we identified two categories of DEGs: EE DEGs, derived from comparisons between cage and cage-free systems (BC vs. BR or FR), and SUN DEGs, identified from comparisons between indoor and outdoor systems (BC or BR and FR) (Fig. 2b and c).

The DEGs identified in peripheral tissues related to egg production, including the ovary, magnum, and uterus, suggested alterations in metabolic and reproductive processes such as follicle development and ovulation (ovary), albumen deposition (magnum), and eggshell calcification (uterus). In the ovary, several genes known to affect ovarian function showed significant differential expressions (Zhang et al., 2019; Mishra et al., 2020; Mu et al., 2021) (Supplemental Table 9). Specifically, IRX3 (Iroquois Homeobox 3), CFAP77 (Cilia And Flagella Associated Protein 77), ENSGALG00000046141, ENSGALG00000051811, and MYLK (Myosin Light Chain Kinase) were upregulated in high-productivity strains (Mishra et al., 2020) as well as under BR and FR. In contrast, OVA (Ovalbumin) were downregulated, and CNTN3 (Contactin 3) and LRRC18 (Leucine Rich Repeat Containing 18) were upregulated in high-productivity strains (Mishra et al., 2020; Mu et al., 2021) but showed the opposite expression pattern in BR. This may be due to the differences in breed, age, and timing of sample collection. On the other hand, the pathway analysis of the ovary detected enrichment in immune-related pathways, including intestinal immune network for IgA production, cytokine-cytokine receptor interaction, cell adhesion molecules, and tight junctions (Supplemental Table 7). Yu et al. (2023) showed that environmental enrichment by litter provision can enhance immune and tight junction gene expression even in caged hens. In our profiling, immune genes including PIGR (polymeric immunoglobulin receptor), CCL20 (C—C Motif Chemokine Ligand 20), CD86 (CD86 Molecules), CCR10 (CC Motif Chemokine Receptor 10), and CSF3 (Colony Stimulating Factor 3) were upregulated in cage-free hens compared with caged hens (Supplemental Table 8). Tight junction genes such as CDH1 (Cadherin 1), CLDN3, and CLDN18 (Claudin 3 and 18) were also upregulated in BR (Supplemental Table 8). These results suggest that immune activity was enhanced in cage-free hens by environmental enrichment.

In the magnum and uterus, no significant changes were observed in genes related to albumen formation, thick egg albumen, eggshell formation, or shell strength (Wan et al., 2017; Sah et al., 2018, 2021; Fu et al., 2024) (Supplemental Tables 10 and 11). On the other hand, pathway analysis detected upregulation of serotonin receptor genes in both tissues: HTR7L in the magnum and HTR2B (5-hydoroxytryptamine receptor 7 and 2 B) in the uterus were upregulated in FR hens (Supplemental Table 8). Serotonin has been implicated in oviductal reproductive function (Paczoska-Eliasiewicz and Rzasa, 1987). Since sunlight increases plasma serotonin levels in humans particularly under UVA exposure (Gambichler et al., 2002; Lambert et al., 2002), sunlight may similarly elevate serotonin in FR hens, leading to higher expressions of its receptors.

Norepinephrine signaling in the cerebral

In the cerebral, several genes were significantly enriched in neuroactive ligand-receptor interaction and calcium signaling pathways (Supplemental Figure 5). Notably, PNMT was significantly down-regulated in FR than BC, whereas ADRB1 and CRACR2B were up-regulated in BR and FR than BC (Fig. 4d). Additionally, ADRA1D expression was higher in BR than in BC and FR (Fig. 4d). PNMT encodes an enzyme that catalyzes the final step in monoamine synthesis by converting norepinephrine to epinephrine. Both ADRA1D (α1) and ADRB1 (β1) are G-protein-coupled adrenergic receptors that activate intracellular calcium signaling: α1 receptors activate intracellular Ca²⁺ release via the Gq/11 and PLC-IP3 signaling cascade, whereas β1 receptors stimulate the Gs and cAMP-PKA signaling cascade, which also leads to Ca²⁺ mobilization. CRACR2B, a calcium sensor involved in store-operated Ca²⁺ entry; it promotes channel clustering and Ca²⁺ influx upon store depletion (Srikanth et al., 2010). CRACR2B is also expressed in hypothalamic neurons, where it contributes to neuronal excitability (Cahalan, 2009; Peterson et al., 2018). Moreover, intracellular Ca²⁺ is critical for vesicular neurotransmitter release (Freedman, 2025). Taken together, these transcriptomic differences suggest that adrenergic calcium and norepinephrine signaling in neural circuits was enhanced in cage-free hens (BR and FR) compared to caged hens (BC) (Fig. 4d).

The transcriptomic changes were well supported by differences in concentrations of monoamines and their metabolites. The ECD-HPLC analysis showed elevated NE levels and a reduced MHPG/NE ratio in the cerebral of BC hens compared to FR hens. NE released into the synaptic cleft binds to their receptors on pre-synaptic membrane and is subject to feedback inhibition primarily via α2-adrenergic autoreceptor such as ADRA2A (Fig. 4d). After signal transmission, norepinephrine is degraded to MHPG by monoamine oxidase (MAO) and catechol-O-methyltransferase (COMT), and reabsorbed into the nerve terminal by NE reuptake transporters, including SLC18A1 and VAT1. Transcriptome analysis showed no significant differences in gene expression related to these degradation or reuptake pathways among housing systems (Supplemental Figure 3). However, ADRA2A, which attenuates neurotransmitter release during high-frequency stimulation, showed a tendency for increased expression in BC hens compared to BR and FR hens, suggesting possible synaptic NE accumulation in BC. Excess synaptic norepinephrine, if not efficiently degraded, can impair neurotransmission by reducing signal specificity (Freedman, 2025). These findings support the hypothesis that norepinephrine signaling is enhanced in cage-free systems, whereas elevated norepinephrine levels and ADRA2A expression in BC hens may reflect compensatory responses to synaptic saturation.

The transcriptomic and metabolic differences also align with the change by environmental enrichment, a key distinction between cage and cage-free systems. Environmental enrichment, characterized by a stimulating and dynamic environment, is known to enhance behavioral diversity and cognitive function, and to induce neuroplastic changes (Rampon et al., 2000; Van Praag et al., 2000; Appleby et al., 2011). Previous studies have shown that environmental enrichment modifies monoaminergic dynamics in the brain and reduced anxiety- and depression-like behaviors (Oshea et al., 1983; Naka et al., 2002; Brenes et al., 2008). Whereas NE levels in BC hens were elevated compared to BR and FR, the transcriptomic data consistently point to enhanced NE signaling in BR and FR. These findings suggest that long-term exposure to environmental stimulation leads to sustained upregulation of norepinephrine-related gene expression, indicative of increased norepinephrine turnover (MEPG/NE ratio).

Insulin resistance in the Diencephalon and Liver

KEGG pathway analysis detected enrichment of genes involved in insulin resistance and feeding regulation in the diencephalon (Figure 5, Supplemental Figure 4 and 5). In the liver, enriched pathways also included the adipocytokine signaling pathway, which is associated with glucose and lipid metabolism and insulin signaling, as well as FoxO signaling pathway, a downstream component of insulin signaling (Supplemental table 7). Specifically, MLXIPL, RPS6KA2, and RPS6KA were significantly upregulated in the diencephalon of BC hens, whereas CARTL was significantly downregulated in BC hens compared to BR hens (Fig. 5a, e). MLXIPL (MLX-interacting protein-like), also known as ChREBP (Carbohydrate-Responsive Element Binding Protein), binds to and activates triglyceride synthesis genes in a glucose-dependent manner. RPS6KA2 and RPS6KA3 (Ribosomal Protein S6 Kinase A1 or 2), members of the ribosomal S6 kinase (RSK) family, encode serine/threonine kinase, known to mediate insulin resistance through hypothalamic signaling that influences hepatic function (Ono et al., 2008). Alterations in insulin sensitivity within the diencephalon are also transmitted to the liver via hepatic vagus nerves (Pocai et al., 2005). Therefore, the upregulation of MLXIPL, RPS6KA2, and RPS6KA3 in BC hens is suggested to mediate insulin resistance in the liver via hepatic vagus nerve (Fig. 5e). In arcuate nucleus (Arc) of diencephalon, CARTL (Cocaine- and Amphetamine- Regulated Transcript-Like) is implicated in appetite and energy balance. Previous research has shown that intracerebroventricular injection of CART suppresses fasting-induced feeding in chicks (Tachibana et al., 2003), and CART expression in the diencephalon increases following intracranial injection of insulin (Honda et al., 2007). In contrast, AGRP (Agouti related neuropeptide) involved in the control of feeding behavior tended to be upregulated in BC hens than BR hens (Supplemental Figure 5). AGRP expression suppressed insulin and leptin via STAT3 or FoxO signaling cascades (Varela and Horvath, 2012). Therefore, the change in expression of CARTL and AGRP also suggests the enhancement of insulin resistance in BC hens (Fig. 5e).

In the liver, PCK1 was upregulated in BC hens compared with BR hens (Fig. 5b and e). PCK1 (Phosphoenolpyruvate Carboxykinase 1), a rate-limiting enzyme in gluconeogenesis, is typically inhibited by insulin (Hanson and Reshef, 1997). In contrast, AKT3 (AKT Serine/Threonine Kinase 3) was downregulated in BC hens (Supplemental Fig. 4). AKT proteins are central mediators of insulin signaling and regulate transcription of gluconeogenic enzymes (Hatting et al., 2018). This resistance likely contributed to downregulation of CARTL, whose expression is normally stimulated by insulin, in the diencephalon, and to the upregulation of PCK1, a key gluconeogenic enzyme, in the liver (Fig. 5e). Taken together, the changes of gene expression of both brain and liver commonly suggested that insulin resistance is enhanced in BC hens. Therefore, it is reasonable to speculate that BC hens exhibit elevated blood glucose levels, where insulin is less effective in lowering glucose concentrations, thereby impairing glycemic control.

This hypothesis is partially supported by the glucose tolerance test results, which showed that plasma glucose concentrations in BC hens tended to remain elevated from 30 min after oral glucose administration through 240 min, compared with those in BR hens (Fig. 5c). The insulin resistance, potentially exacerbated by restricted movement in cage systems, reflects metabolic alterations associated with fatty liver hemorrhagic syndrome (Shini et al., 2019; Zhuang et al., 2019). In this study, BC hens performed fewer comfort behaviors and exhibited significantly lower locomotor activity than BR and FR hens (Fig. 1b-c and Supplemental Table 2), and some of BC hens showed fatty liver hemorrhagic syndrome (no data).

Despite exhibiting similar activity levels to BR hens (Fig. 1b), FR hens surprisingly displayed similar transcriptomic and glucose clearance profiles to BC hens. The primary environmental distinction between BR and FR housing was the presence of outdoor sunlight. These results underscore the need for further research to elucidate the molecular mechanisms underlying differences in insulin signaling between BR and FR hens.

In conclusion, interactive changes in gene expression between the diencephalon and liver suggest that insulin resistance is enhanced by behavioral restriction in BC hens. In addition, comfort behaviors and norepinephrine signaling in the cerebrum are promoted by enriched environments in BR and FR systems. This study provides a comprehensive examination of whole-body characteristics at the molecular level and elucidates the physiological landscape of laying hens in different housing systems by multi-tissue transcriptomic profiling. Although the sample size and the scale of housing system were limited, and the findings should be validated in future studies, this work contributes to the understanding of molecular mechanisms underlying welfare-related outcomes and offers a basis for developing novel molecular markers for welfare of laying hens.

Data availability

The RNA-seq data are available in the DDBJ Bioproject database PRJDB38014.

CRediT authorship contribution statement

Nonoko N. Shimura: Writing – review & editing, Writing – original draft, Visualization, Validation, Formal analysis, Data curation. Eiki Asagi: Formal analysis, Data curation. Tadahiro Matsubara: Formal analysis, Data curation. Itsufumi Sato: Formal analysis, Data curation. Yuki Higashiura: Formal analysis. Saki Nakamura: Formal analysis, Data curation. Chihiro Kase: Visualization, Validation, Project administration. Atushi J. Nagano: Formal analysis. Shozo Tomonaga: Visualization, Validation. Jun-ichi Shiraishi: Visualization, Validation. Kaito Kurogi: Formal analysis. Ryohei Matsuo: Formal analysis. Shinobu Yasuo: Visualization, Validation. Tatsuhiko Goto: Visualization, Validation. Kan Sato: Visualization, Validation. Tsuyoshi Shimmura: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Disclosures

We declare no competing financial interests.