Effect of fasting and subsequent refeeding on the transcriptional profiles of brain in juvenile Spinibarbus hollandi

Starvation is a common stress in fish. The underlying molecular mechanisms associated with growth depression caused by feeding restriction and compensatory growth are not well understood. We investigated the effect of fasting and refeeding on the transcriptome profiles of brain in juvenile S. hollandi using RNA-seq. A total of 4.73 × 108 raw reads were obtained from nine brain samples. De novo transcriptome assembly identified 387,085 unigenes with 2.1×109 nucleotides. A total of 936 annotated unigenes showed significantly differential expression among the control, fasting, and fasting-refeeding groups. The down-regulated differentially expressed genes (DEGs) during fasting were mainly associated with cell cycle, DNA replication, and mitosis. The up-regulated DEGs were mainly related to glucose and lipid metabolism, material transportation, and transcription factors. Most decreased DEGs during fasting were restored to normal levels after refeeding. Comparing with the control group, genes associated with protein synthesis, stimulus response, and carbohydrate metabolism were significantly over-expressed and pro-opio melanocortin (POMC) was down-regulated during the refeeding period. In conclusion, fish mobilized stored energetic materials and reduced energy consumption to prolong survival during fasting. After refeeding, the down-regulation of DEGs, e.g., POMC may be associated with compensatory growth. Up-regulation of DEGs related to ribosomal protein, stimulus response, and carbohydrate metabolism may contribute to eliminate negative effect of starvation on brain. This study provided the first transcriptome data related with impact of short-time starvation and refeeding in S. hollandi brains.


Introduction
Starvation is a common stress in fish. Extreme aquatic conditions, such as unsuitable temperature, hypoxia, and high population density, can affect the feeding behavior of aquatic animals. The starvation condition is strongly associated with the mobilization of the body's energy PLOS  Fish in the control groups were fed to apparent satiation everyday as mentioned above. The fish were sequentially sampled on day 0 (C0), day 7 (C7), day 14 (C14), and day 40 (C40). Fish in the fasting group were deprived of feed for 7 days, and then sampled (F7). The fish in the refeeding groups were fed twice per day to satiation after 7 days of fasting. The fish were sampled after 7 (R7) and 33 days refeeding (R33). All seven groups were then anesthetized by an overdose (100 ppm) of eugenol (Sangon Biotech (Shanghai) Co., Ltd., Shanghai, China). The body weights of each group were measured prior to sampling (n = 8). Brain tissues were rapidly isolated, immediately frozen in liquid nitrogen, and stored at -80˚C.

cDNA library preparation and sequencing
RNA samples of the C7, F7, and R7 groups were used for RNA-seq analysis. To minimize the variation among individuals, equal quantities of RNA from 5 samples belonging to the same group were pooled, and 3 replicates of mixed RNA pools were prepared for each group. A total of 9 RNA pools were used for cDNA library preparation. mRNA was purified from total RNA using oligo (dT) magnetic beads and fragmented using fragmentation buffer. First-strand cDNA was synthesized using SuperScript II Reverse Transcriptase (Applied Biosystems Ltd., USA) followed by second strand generation using the components dNTPs, buffer, DNA polymerase I, and RNase H. After purification, adapters were ligated to the cDNA, and then enriched using PCR. The nine cDNA libraries were sequenced on an Illumina-Hiseq 2000 platform using the 150 bp paired-end approach.

Data processing and analysis
Clean reads were generated by removing low-quality reads reads containing adapter and ploy-N from raw data. Clean reads were de novo assembled using Trinity software version 2.5.1 [21]. The assembled transcripts were identified using Blast+ (version: ncbi-blast-2.2.28+) against NCBI non-redundant (Nr, e-value <1e -5 ), Clusters of Orthologous Groups of proteins (COG, e-value <1e -3 ), Kyoto Encyclopedia of Genes and Genomes (KEGG, e-value <1e -5 ), and Swiss-Prot databases (e-value <1e -5 ). Gene ontology (GO) annotation was performed using Blast2GO with an e-value threshold of 1e -6 .
Total number of mapped reads for each transcript was determined and normalized, to calculate the expected number of Fragment Per Kilobase of transcript sequence per million base pairs sequenced (FPKM) by using RSEM V1.2.15 [22]. Differential expression analyses of three groups (C7, F7, and R7) were performed using DESeq R [23]. The adjusted p-value (padj) with a cut-off of 0.05 was used to identify significant differentially expressed genes (DEGs). Pathway enrichments of DEGs were analyzed using the KEGG database. expression changes in seven groups (C0, C7, C14, C40, F7, R7, and R33) through quantitative real-time PCR (qRT-PCR). A total of 10 DEGs including the above six genes, and TFRC1, TUBB5, MCH1, and ACSBG2, were used to verify RNA-seq results. It was performed by detecting expression changes of these DEGs during fasting (C7 vs. F7) and refeeding (C7 vs. R7) using qRT-PCR. Three biological replicates were used for each RNA sample and the betaactin gene was used as an internal control. Primers used for qRT-PCR analysis are shown in S1 Table. The qRT-PCR analysis was performed using ABI 7000 platform with 20 μL reactions containing the following components: 100 ng of cDNA, 10 μL Power SYBR Green PCR Master Mix (Vazyme Biotech, Nanjing, China), 0.3 μL of each primer (10 μmol/L), and 7.4 μL doubledistilled water. The reaction procedure followed by 95˚C for 30 s, 40 cycles at 95˚C for 10s and 60˚C for 30s. After amplification, a melting curve was obtained by default processes in ABI7000. All samples were analyzed in triplicate, and fold changes were calculated using the comparative Ct method (2 −ΔΔCt method) [25]. All data are provided in terms of relative mRNA expression as mean ± S.E. (n = 3).

Changes of body weights during fasting and refeeding periods
The body weights of fish in the control groups (C7, C14, and C40) and the test groups (F7, R7, and R33) were measured. The growth curves for these groups were shown in Fig 1. At the end of 7-day fasting period, the body weight of C7 and F7 were 1.57 g and 1.16 g, respectively.
There was a significantly difference in body weights between groups (p < 0.05). After 7-day refeeding, the body weight of C14 and R7 were 1.94 g and 1.63 g. After 40 days, the body weights for C40 and R33 was 3.34g and 3.28 g, respectively. However, no significant difference between groups was detected at the same time points.

RNA-seq data analysis
Through high-throughput paired-end sequencing, 4.93 × 10 8 raw reads (150 bp) were yielded from nine brain samples. After removing low-quality sequences, 4.73 × 10 8 clean reads with 71.01 Gb were generated. The Q20 and Q30 of all samples were above 96.27% and 90.97%, respectively. The sequence data were deposited in the NCBI SRA database, under the accession number, SRP156354.
After de novo assembly, 864,488 transcripts (ranging from 201 to 23,693 bp) were generated. The longest transcript of a gene was defined as a unigene for further analysis. A total of 387,085 unigenes with 2.1×10 9 nucleotides were obtained, with an average length of 568 bp. The N50 and N90 were 752 and 254 bp, respectively.
Functional prediction and classification of all unigenes were performed using the GO database. A total of 71,164 unigenes were classified into 56 GO terms, including 10 molecular function terms, 20 cellular component terms, and 26 biological process terms. Cellular process (39,913), binding (37,484), single-organism process (32,048), metabolic process (31,836), and cell (24,674), were the five most abundant functional terms at the second GO level (S1 Fig). Furthermore, all unigenes were aligned to COG database for phylogenetic classification. In total, 22,886 unigenes were divided into 26 functional categories (S2 Fig). The largest category was signal transduction mechanisms with 4,653 unigenes followed by general function prediction (3,511), posttranslational modification, protein turnover, chaperones (2,283), intracellular trafficking, secretion, and vesicular transport (1,725).
To interpret the biological pathways, we annotated all assembled unigenes in the KEGG database, and 39,265 unigenes were matched in level 2 of the KEGG pathway. Signal transduction (6744), endocrine system (3302), cellular community (2976), immune system (2643), and nervous system (2432) were the five most significant matched catalogs in the KEGG pathway (S2 Table).

Identification of DEGs between control and fasting group
A total of 936 annotated unigenes showed significantly differential expression among the C7, F7, and R7 groups. Of the DEGs, 464 were significantly expressed between C7 and F7 (fasting period), 491 were significantly expressed between F7 and R7 (refeeding period), and 230 were significantly expressed between C7 and R7.
Except the above catalogs, KEGG pathway analysis identified that the down-regulated genes were also enriched in other pathways, such as steroid biosynthesis, pathogenic Escherichia coli infection, phagosome, gap junction, and sesquiterpenoid and triterpenoid backbone biosynthesis (Fig 2).

Identification of DEGs between fasting and refeeding group
After 7-day refeeding, of the 491 DEGs, 374 and 117 genes were up-regulated and down-regulated, respectively. The up-regulated DEGs were remarkably enriched in antigen processing and presentation, cell cycle, steroid biosynthesis, protein processing in endoplasmic reticulum, DNA replication, mitosis, pathogenic E. coli infection, and terpenoid backbone biosynthesis (Fig 3).

qRT-PCR analysis of key genes
To investigate the expression profiles of key DEGs under the continuous feeding condition and fasting-refeeding condition, genes associated with cell cycle (CDC6 and CDC20), DNA replication (MCM2), steroid biosynthesis (DHCR7), fat metabolism (FABP7), and stimulus response (HSP70) were analyzed using qRT-PCR. All of these genes (CDC6, CDC20, MCM2, DHCR7, FABP7, and HSP70) were significantly down-regulated after 7-day fasting. CDC20 and HSP70 were sharply increased and higher than the control group after 7-day refeeding. On day 33 of refeeding, HSP70 remained significantly higher in the refeeding group than the control group. CDC20, however, was restored to normal level, and other genes were also restored and stable after refeeding. All genes had stable expressions under the continuous feeding condition (Fig 4).

Confirmation of DEGs by qRT-PCR
To confirm the RNA-Seq results, 10 DEGs in different pathways were selected for qRT-PCR verification. Expression changes of these genes at fasting and refeeding were analyzed. The changes in gene expression obtained from qRT-PCR were consistent with the RNA-Seq under fasting and refeeding conditions, with R 2 values of 0.7315 and 0.8251 (Fig 5), respectively. The results of the qRT-PCR analyses confirmed the reliability and accuracy of the data obtained by RNA-seq.

Changes in transcriptome profile of fish brain after 7-day fasting
The present study used RNA-Seq to investigate the impact of fasting and refeeding on transcriptome profiles of brain in S. hollandi. The global gene expression pattern of fish brain under continuous feeding condition, fasting condition, and refeeding condition were displayed. After 7-day fasting, 295 down-regulated and 169 up-regulated genes were identified in the brain.  Effect of fasting and subsequent refeeding on the transcriptional profiles of brain in S. hollandi The down-regulated DEGs were mainly enriched in the pathway of cell cycle, DNA replication, mitosis, steroid biosynthesis, immunity, gap junction, and sesquiterpenoid and triterpenoid backbone biosynthesis. Cyclins are important factors in the control of cell cycle, and can be activated by cyclin-dependent kinase enzymes [26]. Cyclin A2 (CCNA2), Cyclin-dependent kinase 1 (CDK1), and Cyclin-dependent kinase 2 (CDK2) were down-regulated during fasting. CHEK2, E2F3, and ESPL1 play crucial roles in the control of cell cycle [27][28][29]. CDC20 is essential for the regulation of cell division; however, its most important function is to promote chromatid separation [30]. CDC6 is crucial to the maintenance of checkpoint mechanisms in the cell cycle and the loading of minichromosome maintenance (MCM) proteins onto DNA [31]. MCM, a DNA helicase composed of six subunits, is essential for genomic DNA replication. During fasting, MCM2, MCM3, MCM4, MCM5-B, and MCM7-A were all down-regulated. Mitosis is a part of the cell cycle wherein replicated sister chromatids are separated into two chromosomes. Several genes related to mitosis (NCAPD2, NCAPG, PTTG1, SMC2, and SMC4) were down-regulated. NCAPD2, NCAPG, SMC2, and SMC4 were involved in chromosome condensation [32,33]. Securin, encoded by PTTG1, is a protein involved in the control of the metaphase-anaphase transition [34]. It has been widely reported that cell proliferation can be    Effect of fasting and subsequent refeeding on the transcriptional profiles of brain in S. hollandi inhibited by starvation in mammal and fish [2,4]. The down-regulation of DEGs associated with steroid biosynthesis, immunity, and other functions may reduce energy expenditure and prolong the survival of S. hollandi. Several genes associated with regulating glucose and lipid metabolism (PDK4, ADIPOR1, ARRDC3, PMM1, and BBOX1) were markedly up-regulated. PDK4 decreases glucose utilization and increases fat metabolism through inhibiting pyruvate dehydrogenase. This plays an important role in regulating the shift in fuel economy to prolonged fasting and starvation [35]. ADIPOR1 is essential in the regulation of normal glucose and fat homeostasis, and for maintaining normal body weight during the fasting period [36]. ARRDC3 can decrease energy expenditure through decrease in the thermogenesis of adipose tissues [37] and PMM1 has a phosphoglucomutase activity, which converts glucose-1-P into glucose-6-P [38]. BBOX1 is an enzyme involved in the biosynthesis of L-carnitine, a key molecule in fatty acid metabolism [39]. ZBTB16, a transcriptional repressor, plays a key role in various biological processes such as adipogenesis, regulation of lipid levels, and insulin sensitivity [40]. Maintaining metabolic homeostasis is essential for the survival of fish during fasting, and similar results have been reported in the rainbow trout, yellow croaker, and grass crap [3,14,41].
Three nutrition transport genes (SLC43A2, SLC16A3, and SLC19A3) were up-regulated. These trans-membrane proteins are substrate-specific and their increased expression may aid in the mobilization of nutrients. Several genes associated with ketone body metabolism (SLC16A3 and PDK4) were up-regulated, and brain energy metabolism usually changes toward oxidation of ketone bodies during starvation [42]. Atrophy and autophagy can be induced by starvation, and protein synthesis can be inhibited. EEF2K can inhibit the eukaryotic elongation factor 2 (EEF2), an essential factor for protein synthesis. FBXO32 is associated with muscle atrophy during the fasting period [43] and plays a similar function in brain. GABARAPL1 and TP53INP2 are essential genes for autophagosome maturation [44,45]. These DGEs can assist in alleviating starvation stress in fish. The change of transcriptome profiles associated metabolism and mobilization of stored energy were similar to the liver tissues [3]. Effect of fasting and subsequent refeeding on the transcriptional profiles of brain in S. hollandi

Changes in transcriptome profiles of fish brain after refeeding
After refeeding, most DEGs down-regulated after fasting were restored to normal levels. Many of the genes associated with immunity and protein synthesis were up-regulated, and are known to play important roles in antigen processing and presentation, protein processing in endoplasmic reticulum and pathogenic E. coli infection. An improvement in disease resistance capability and immune status was observed in the refeeding fish. These fish were also seen to outperform the fish that were continuously fed.
DEGs between control group and refeeding group were also analyzed. Many ribosomal proteins were over-expressed in the refeeding group. These proteins consisted of the ribonucleoprotein complexes and involved in the cellular process of translation. Stress-specific alterations in ribosomal proteins were reported in many species, the over-expressions of ribosomal proteins contributed to the starvation response. Similar result has been observed in the muscle of rainbow trout [46].
Several members of the heat shock proteins were over-expressed after refeeding. HSP is an important protein in response to the exposure to stressful conditions. Many members of HSP act as chaperones to assist new proteins, ensuring them correct folding and preventing protein damage by cell stress [47]. The over-expression of HSP can enhance translational efficiency and increase the ability of stimulus response.
Glycolysis and oxidative phosphorylation are the main metabolic pathways for releasing energy in organisms and provide energy for various physiological activities. The up-regulation of LDHB, TPI1, GAPDH, PGAM1, LDHA, PGK1, ALDOA, GPI, MT-ND5, and MT-CO1 was beneficial for the production and restoration of energy.

Analyses of expression profiles of key DEGs using qRT-PCR
Expression profiles of six key DEGs were investigated using qRT-PCR. CDC6, CDC20, MCM2, DHCR7, and FABP7 were down-regulated during fasting and restored after refeeding. DHCR7 plays a vital role in the cholesterol biosynthesis pathway, which is part of the steroid biosynthesis pathway [50]. Starvation inhibited steroid biosynthesis; thus, DHCR7 was down-regulated during fasting and restored after refeeding. FABP7 was down-regulated during fasting, and is most likely because the ketone body was the main energetic materials of brain during starvation. FABP3 was also down-regulated during fasting. CDC20, CDC6, and MCM2 were all down-regulated because of the blockage of cell division by starvation. After refeeding, the expression of CDC20 was higher than in the control group, despite there was no significant difference. HSP70 is important for protein folding, and protects cells from stress. Thus, the overexpression of HSP70, after refeeding, can protect organisms from undergoing stress.

Conclusion
A transcriptome analysis was carried out to reveal the effects of fasting and refeeding on brains of juvenile S. hollandi. During fasting, the fish significantly reduced gene expression in the cell cycle, DNA replication, mitosis, steroid synthesis, and other physiological activities. After 7-day refeeding, the up-regulated DEGs were remarkably enriched in antigen processing and presentation, cell cycle, steroid biosynthesis, protein processing in endoplasmic reticulum, DNA replication, mitosis, pathogenic infection, and terpenoid backbone biosynthesis.
The study indicated that growth was inhibited during starvation since the fish mobilized the stored energy to prolong survival. After refeeding, the energy was stored again, and compensatory growth occurred, the body weight was recovered. In response to the starvation and feeding, a great deal of related genes and pathways were changed. This study provided the first transcriptome data on impacts of short-time starvation and refeeding to fish brains.