Animal experiments were performed under approval and review by the Institutional Animal Ethics Committee (IAEC) at ICAR-Central Institute for Research on Buffaloes Hisar, Govt. of India. Heifers of 9 to 11 months age were subjected to control feeding in order to record individual dry matter intake (DMI) to determine residual feed intake (RFI) by estimating the difference between predicted and actual DMI for an individual animal. Dietary regimen of heifers included Jowar fodder and concentrate mixture. Nutrient value of feed was determined using standard methods. Chemical composition and nutrient digestibility of feed (g/kg) and nutritive value of diet were determined by reference methods.

Nutrient digestibility (Mean ± SE) of dry matter was 60.4 ± 2.43. Percentage of crude protein (CP), digestible crude protein (DCP) and total digestible nutrients (TDN) was determined as 14.2 ± 0.13, 12.5 ± 0.33 and 61.2 ± 1.77, respectively. Blood biochemical profiling revealed IGF1 and Glucose as significant biological attributes related to RFI efficiency in heifers [8] .

^aValues represent hex plicate assays of each material; ^bIngredient composition of concentrate mixture: [maize grain, 175 g/kg; barley grain, 175 g/kg; wheat bran, 270 g/kg; mustard cake, 200 g/kg; cotton seed cake, 150 g/kg; mineral mixture, 20 g/kg; common salt, 10 g/kg].

Daily feed intake and fortnightly body weight were recorded for heifers during the period of 78 days feeding trial. Average DMI (Kg/h/d) and daily weight gain (ADG) were determined. Average BW was attained as 200 Kg in heifers after 78 days of feed trial initiating from 155 Kg. Mean ADG and DMI per animal was recorded as 577 g/h/d and 4.8 kg/day, respectively (Table 1). Variation in daily weight gain and DMI was recorded ranging from 382 to 897 g/day and 3.3 to 6.0 kg/d over the 42 heifers during the trial. Mean DMI was regressed with RFI at (R² 0.970) in the range of 0.359381 to −0.43778 at high and low extremes (Figure 1). DMI was recorded as 2.41 Kg lesser in efficient feed conversant group of heifers (low RFI) as compared with less efficient (high RFI). BW gain in these heifers was recorded as 614.84 g/h/d. Low RFI heifers gained 16.4% higher BW through lesser DMI by 13.53% (Figure 1).

A large variation in DMI per unit BW gain over was determined in heifers (n = 42) allocated to low and high RFI feed efficiency (Table 1). DGEs in respect of FCE were identified in heifers comparing transcriptomic data obtained from high and low extremes of RFI efficient heifers through high throughput (HTP) sequencing.

Total RNA from blood cells collected through veni-puncture was extracted from

four heifers (two showing high RFI and two showing low RFI). 1 μg of intact RNA having integrity number (RIN) of 8.0 (Agilent 2100 Bio analyzer) was used to purify mRNA using oligo-dT beads (TruSeq RNA Sample Preparation Kit, Illumina) for each sample. cDNA library preparation and paired-end sequencing were done using denatured (90˚C) mRNA. The RNA fragments were reverse transcribed using random hexamers and Superscript II Reverse Transcriptase (Life Technologies). Second strand cDNA was synthesized on this first strand template, using RNaseH and DNA polymerase I. Clean cDNAs were obtained using Beckman Coulter Agencourt Ampure XP SPRI beads after end-repair and the addition of an “A” base. The clean cDNA molecules were ligated with Illumina adapters. Further, the cDNA library was amplified using PCR for the enrichment of the adapter-ligated fragments. The individual libraries were quantified using NanoDrop spectrophotometer (Thermo Scientific) and validated for quality by Bio analyzer (Agilent Technologies) for subsequent sequencing, using the Illumina HiSeq 2500 platform. Paired-end FASTQ files were subjected to standard quality control. High quality (HQ) filtered reads (Phred scores > 20) were selected using the NGSQC Tool Kit [9] .

RNA extracted from blood of the four heifers (Bubalus bubalis), covering extremely high and low RFI efficiency as estimated based on DMI/h/g BW gain as FCE1:FCE2: vs FCE3:FCE4 was subjected to standard paired end RNA-Seq library(s) preparation as per Illumina recommended protocol using 100 bp paired-end module. Approximately 40 - 60 million paired end sequence reads (Phred score > Q30) were obtained for each library and the same were subjected to stringent quality test using NGSQC tool kit [9] .

The improved transcripts resulting from the trans-assembly were subjected to the CD-HIT EST clustering pipeline. Redundant transcripts were removed and reference unigene transcriptome was generated from each library, all converging finally into a single non-redundant resultant transcriptome assembly by merging and clustering the transcripts from individual libraries. Resultant transcriptome assembly was aligned with the reference genome sequence, using Bowtie tool for quantification of expressed transcripts. Read count for each transcript was further normalized to determine value of RPKM. Transcripts with an RPKM of ≥1 were categorized as expressed transcript. Treatment specific transcripts were annotated with Gene Ontology and Pathways identified from expressed transcripts by DAVID Functional Annotation Tool [DAVID Bioinformatics Resources 6.8, NIAID/NIH [10] .

The Reference genome of Bubalus bubalis (https://www.ncbi.nlm.nih.gov/assembly/GCF_003121395.1) was used for read alignment and identification of transcripts coding regions using Kallisto followed by quantification and annotation [11] . Differentially expressed (DE) transcripts (fold-change ≥ 2.0, P-value < 0.05) were identified by DESEQ2 [12] which revealed underlying genetic make-up of efficient RFI, compared with high RFI as control.

The molecular functions, biological mechanisms and gene networks that emerged from the transcripts which were over-represented were determined. Dataset containing gene identifiers, corresponding expression and P-values were uploaded into IPA. Identification of de-regulated genes was done based on significant P-adj value (P < 0.05). The “focus” genes were overlaid onto a global molecular network using Cytoscape software [13] to lineate bio-molecular interaction networks into a unified conceptual framework by evaluating a score ranking the network obtained according to their degree of relevance to the eligible molecules in the dataset. The score is the result of the number of knowledge base considerations of the network eligible molecules into account out of the total number of network eligible molecules that emerged and the total number of possible molecules to be included in the network.

Ethics Approval: The protocols followed in the present animal study were reviewed and approved by the Institute Animal Ethics Committee (IAEC) of CIRB Hisar. India (Reg. No. 406/GO/RBI/L/01/CPCSEA).

An average of 51.9 million (varying between 36.9 - 72 million) paired end raw RNA Seq reads were obtained per sample using Illumina HiSeq platform. Average size of reads qualifying for all the QC criteria was 210 bp (Table 1). FASTQ sequences were filtered using NGSQC tool kit thus, an average of 52 million high quality reads per sample (~92.7%) were accounted for alignment and variant detection analysis. Around 88.7% of the HQ reads of created assembly were mapped on the human reference genome using KALLISTO pipeline, which confirms optimum alignment (Figure 2) of the two genomes.

Per sample high quality reads obtained from all the heifers were comparable covering nearly 9% of the total reads and having average GC content of 47.22% (Table 1).

RNAseq dataset obtained from selected heifers were evaluated for DEGs including 49781 significant transcripts (Padj = 2.44995163703878E-54 to 0.0504821201656885). Percent alignment of reads with reference genome of Bubalus bubalis varied between 64.1% to 67.7% with respect to heifers (Figure 2) in the study. Differentially expressed (P < 0.05) transcripts were 850. Correlation matrix coefficients were determined as 0.821*** between replicates of high RFI, i.e. FCE1:FCE2 and 0.967*** between replicates in low RFI individuals i.e. FCE3: FCE4 (Table 2). It indicates high similarity in genetic make-up underlying the selected heifers as per the estimated RFI efficiency. Correlation coefficients while regressing the transcript data across high and low RFI category of heifers was significantly low (r² 0.851***) for FC1:FC3 and FC2:FC4 (r² = 0.816***) while comparing within high or low RFI.

RFI of an individual heifer is influenced by digestibility, tissue metabolism and protein turnover [14] under the changing patterns in environmental stress, thermoregulation, heat increment, and body composition [15] while evaluating RFI efficiency having uniform feeding regimen.

Distinctive pattern of transcripts expression is depicted through Heatmaps (Figure 3(a) and Figure 3(b)) generated from high and low RFI efficient heifers. De-regulated genes (P-adjusted < 0.05) are presented as fold change value (>2 or <−2) expressing variation in RFI (Table 2 and Table 3). A total of 850 DEGs (p < 0.05) were detected based on the Cuffdiff analysis (FDR = 0.00174998654944857 to 58.5661012869287 (FDR < 0.05). 427 genes were down-regulated and 420 genes were up-regulated in the study, where top thirty genes in each category of deregulation are shown in Table 2 and Table 3.

Significant GO terms enriched and the pathways identified (https://david.ncifcrf.gov/) in this study are listed in Table 4. Selected GO terms (Padj ≤ 0.05) were used as input to plot gene network (Figure 4) in respect of FCE trait. Clusters (Padj < 0.05) that emerged in Cytoscape showed Intracellular

Transduction signals (GO: 0035556) as pronounced function in Biological Process (4.3%) followed by Cell migration and Cell proliferation (Figure 5).

Cytoplasm functions (GO: 0005737) were found more pronounced than Nucleoplasm (GO: 0005654) due to a higher number of genes (103) annotated with former than the latter (65 genes), respectively (Table 4), sharring 9.8% of Cellular component (Figure 5) term, however, transcripts of genes governing transcriptional regulation in nucleus were emerged and annotated by Nucleoplasm (GO: 0005654) significantly (Padj < 0.05) (Table 4). GO: terms annotated by high rate of energy dependent translational changes as ATP binding (GO: 0005524), Metal Ion binding (GO: 0046872), Zinc Ion binding (GO; 0008270, bta04151) (Table 4), corroborate with the energy related genes reported earlier as Rap 1 gene, [24] which harvests MAP kinase activity [25] to regulate body weight and feed intake in cattle [26] [27] . The Rap1 (bta 04015) and Ras signalling (bta 04014) pathways emerged as enriched (P < 0.05, FDR in range of 0.00174998654944857 to 58.5661012869287) pathways (Table 4) as regulator of fatty acid content in response to retinoic acid and glycerol transport [28] in higher feed efficiency heifers.

Up-regulation of regulatory function genes for ATP binding (bta 04152: AMPK signalling pathway (Table 4) in B cell activation in present study along with MHC class I protein complex (presenting intra-cellular degraded proteins to cytotoxic T cells); GTPase activity, cell proliferation and differentiation, proteins translocation, signal transduction through transmembrane receptors, followed by subsequent protein synthesis at ribosome, as reported earlier [29] and Platelet activation (bta04611, Table 4) [30] are expected physiological processes in efficient feed utilizing (low-RFI) heifers for achieving higher weight gain by reduced feed intake thus, lowering the input price as compared to rearing the less efficient counterparts.

Biological analysis of transcripts (Padj < 0.05) obtained from system biology analysis was summarized in GO summary (GO: terms and pathways) (Table 4). The DEGs identified were annotated by 93 GO terms/pathways enriched, allocated to the category of post transcriptional regulation and chromatin organization in Nucleus, Cell energy balancing, Immunity, hormonal regulation functions and Cell signaling including the biological pathways as Rap1, Notch, Prolactin, Osteoclast differentiation, m TOR, Ras, PI3K-AKT, FoxO, TNF, ErbB and Sphingolipid. High RFI heifers were associated with high IGF1 (p < 0.05) and other blood parameters such as Cholesterol and Triglycerides (p < 0.001) as reported in buffalo heifers [7] and cattle [31] .

Feed utilization is an energy dependent function requiring availability of higher reducing power to maintain optimum productivity [7] . Metabolic profile differs in high and low RFI heifers. High level of IGF1 has a positive correlation with LDL, HDL fat and SGPT enzyme in this study [7] and might favor the deposition of fat by channeling free triglycerides (P < 0.05) toward muscle tissue to gain weight in less efficient animals. Lower blood level of non-esterified fatty acids and higher muscle growth has been reported in low RFI cattle [32] and pig [20] . Thus, IGF-1 is recognized as a potential biological marker for FE/RFI, in dairy heifers [7] and beef cattle [33] determining a negative correlation between RFI efficiency and circulatory level of IGF-1 [34] in heifers. However, the systemic IGF-I concentration is known to be influenced by environment also [7] [14] . In the contrary, a positive genetic relationship between plasma IGF-I level and RFI in sheep [35] has been reported. RFI DEGs may not be comparable with respect to age, breed and physiological status. However, feeding management and climate change may be the additive factors influencing the DEG patterns.

Differentially expressed genes in respect of RFI efficiency trait are listed the first time in Murrah dairy growing heifers. DG patterns were obtained comparing blood transcriptome [21] data obtained from low and high RFI heifers tried on controlled feeding trials. Feed efficiency is a complex functional expression resulting from the synergism between energy built-up (ATP synthesis) and growth reflected through basal metabolic rate, homeostatic control of body, immune response through lipid metabolism and hepatic inflammation respectively [18] .

A large variation (p < 0.05) in DMI and average daily weight gain (ADG) over young heifers maintained under common management was attributed to the difference in their respective feed utilization efficiency. Feed efficiency was estimated as the residual feed intake (RFI Kg/h/d) in this study. DMI (kg/h/d) was 13.14% lower in the low RFI heifers as compared to high RFI heifers. It hypothesized feed efficiency as a selection trait due to underlying variation expected in genetic make-up of these animals. Feed input cost can be significantly reduced by selective breeding of feed efficient heifers. Variation in RFI is pertinent to metabolism [36] of individual animal which is translated into low consistency in phenotypic growth and RFI.

Hierarchical clustering of differentially expressed transcripts obtained from high and low RFI heifers showed no common transcripts in categories of up and down regulated genes. It supports the selection proposal of the animals in high and low RFI subgroups, especially having common transcripts between the two groups being non-significant in enrichment analysis. The identified set of deregulated genes that were obtained from the analysis are considered as a fingerprint of differences in metabolic homeostasis. This leads to variation in growth rates that are governed by the underlying genetic potential of the heifers. Hence, these transcripts can be used to identify extreme FE phenotypes in buffaloes.

Immune response emerges as a potential body function underlying the biological variation in RFI in this study. Immunity is an energetically costly physiological process [37] as higher incidences of chronic inflammation are expected in high-RFI cattle, consequently making less energy available for growth. Down regulation of genes governing the immune functions indicates low immunity in the high-RFI heifers as reported earlier [19] , thus affecting the efficiency of feed utilization. It is consistent with the regular/increased level of immunity governing gene expression in energy efficient manner as reported in low-RFI cattle [38] [39] .

Efficient (low RFI) heifers showing down-regulated ROS function genes suggest the lower energy expenditure for metabolic function to achieve growth in feed efficient heifers. Genes related to lipid transport and energy production were up-regulated probably, in response to high ROS in the high-DMI group of less efficient animals. High-ADG is consistently increased in response to carbohydrate and lipid transport in favor of cell differentiation.