Abstract

Milk fat is one of the most important economic traits in dairy animals. Yet, the biological machinery involved in milk fat synthesis remains poorly understood. In the present study, expression profiling of 45 genes involved in lipid biosynthesis and secretion was performed using a computational approach to identify those genes that are differentially expressed in mammary tissue. Transcript abundance was observed for genes associated with nine bioprocesses, namely, fatty acid import into cells, xenobiotic and cholesterol transport, acetate and fatty acid activation and intracellular transport, fatty acid synthesis and desaturation, triacylglycerol synthesis, sphingolipid synthesis, lipid droplet formation, ketone body utilization, and regulation of transcription in mammary, skin, and muscle tissue. Relative expression coefficient of the genes was derived based on the transcript abundance across the three tissue types to determine the genes that were preferentially expressed during lactation. 13 genes (ACSS1, ACSS2, ADFP, CD36, FABP3, FASN, GPAM, INSIG1, LPL, SCD5, SPTLC1, SREBF1, and XDH) showed higher expression in the mammary tissue of which 6 (ADFP, FASN, GPAM, LPL, SREBF1, and XDH) showed higher expression during adulthood. Further, interaction networks were mapped for these genes to determine the nature of interactions and to identify the major genes in the milk fat biosynthesis and secretion pathways.

1. Introduction

Milk fat content is regarded as one of the most important economic traits of milch animals; identification of gene networks that regulate lipid biosynthesis and secretion in the mammary gland is essential to our understanding of lactation physiology. Finding candidate genes for improved fat content represents a constant research goal [1] that may further provide opportunities for genetic manipulations to derive more or better milk fat. Comparing biomolecular composition of mammary tissue with other tissues can allow insights into the molecular responses that govern milk fat production. Transcriptional regulation is a major long-term mechanism for the control of metabolism, and switching on and off gene expression essentially drives a cell’s biological function and activity [2]. In the present study, an attempt has been made to identify the genes, which are differentially expressed during milk fat production in bovines, and determine their interaction networks using a computational approach.

2. Materials and Methods

2.1. Identification of Differentially Expressed Genes

The reference bovine gene sequences for 45 genes, previously known to be involved in lipid synthesis (Table 1) [3], were obtained from Ensembl [4]. Electronic Northern (e-Northern) was performed using dbEST and UniGene; briefly, the dbEST [5] was queried for these sequences by BLASTN v2.2.27 [6] using default parameters and the significant hits were looked up in UniGene ESTProfile [7] for transcript abundance based on normalized “transcripts per million” (TPM) values in mammary tissue , skin , and muscles . Where information was available, transcript abundance (value not shown) was also compared between adult and young stages.

Percent mammary transcript abundance for a gene was calculated using the formula:To confirm preferential mammary expression, relative expression coefficient was calculated as the ratio of mammary transcript abundance to the geometric mean of cutaneous and muscle transcript abundance; that is,A twofold change in was, arbitrarily, assumed to be significant; that is, upregulation of expression was inferred when and . Similarly, downregulation was inferred when and .

2.2. Gene Network Analysis

Interaction networks and coexpression profiles for the genes were derived using STRING v9.1 with default settings [8]. STRING is a web-based application for network generation and visualization that uses a database of physical and functional protein interactions derived from four separate sources, namely, genomic context, high-throughput experimental data, coexpression, and existing literature. It quantitatively combines the information from these four sources to generate a weighted interaction network.

3. Results and Discussion

3.1. Transcript Abundance

Transcript abundance was inferred from UniGene ESTProfile on the basis of normalized TPM values (Table 2) for the 45 genes involved in nine bioprocesses including fatty acid import into cells (CD36, LPL, and VLDLR); xenobiotic and cholesterol transport (ABCA1, ABCG2); acetate and fatty acid activation and intracellular transport (ACSS1, ACSS2, ACSL1, ACBP, and FABP3); fatty acid synthesis and desaturation (ACACA, FADS1, FADS2, FASN, and SCD5); triacylglycerol synthesis (AGPAT6, DGAT1, DGAT2, GPAM, and LPIN1); sphingolipid synthesis (ASAHL, LASS2, OSBP, OSBPL10, OSBPL2, SGPL1, SPHK2, SPTLC1, SPTLC2, and UGCG); lipid droplet formation (ADFP, BTN1A1, PLIN, and XDH); ketone body utilization (BDH1, OXCT1); and transcriptional regulation (INSIG1, INSIG2, PPARG, PPARGC1A, PPARGC1B, SCAP, SREBF1, SREBF2, and THRSP). Of the 45 genes included in the study, 23 genes did not have complete ESTProfiles and hence could not be included for further analysis. Notably, the absence of ESTProfiles of these 23 genes does not depress the robustness of the methodology that has been employed in the present study. Clearly, as more and more ESTProfiles get submitted to UniGene, it would become possible to use the same approach for analyzing the expression patterns of different genes including those of these 23 genes. Further, though ESTProfile TPM values lack exactitude as a measure of gene expression, the differences in TPM values tend to correlate with overall expression patterns.

3.2. TPMma/TPMs and Percent Transcript Abundance: Functional Inferences

Based on and : values, 13 genes (ACSS1, ACSS2, ADFP, CD36, FABP3, FASN, GPAM, INSIG1, LPL, SCD5, SPTLC1, SREBF1, and XDH; Table 2; Figure 1) were found to exhibit higher mammary expression over skin or muscle; 6 of these 13 genes (ADFP, FASN, GPAM, LPL, SREBF1, and XDH) further showed preferential expression during adulthood.

The skin has been included for comparison because the mammary tissue is known to be modified cutaneous tissue [9] and differences in the expression pattern of genes between mammary and cutaneous tissue are likely to signify functional differences; muscle tissue has been included as a control. ACSS1, ACSS2, ADFP, CD36, FABP3, FASN, GPAM, INSIG1, LPL, SCD5, SPTLC1, SREBF1, and XDH had higher mammary expression over skin or muscle; ADFP, FASN, GPAM, LPL, SREBF1, and XDH showed preferential expression during adulthood and, hence, was considered most likely to be differentially expressed during milk fat synthesis.

Among genes responsible for fatty acid import into cells, both LPL and CD36 appeared to have greater expression in mammary tissue. LPL primarily functions in the hydrolysis of triglycerides of circulating chylomicrons and very low-density lipoproteins (VLDL). CD36 binds long-chain fatty acids and functions in their transport and also as a regulator of fatty acid transport. LPL showed more than 5-fold increase in TPM values in mammary tissue over cutaneous tissue whereas CD36 showed a more than 13-fold increase. Further, the expression of LPL was greater in adult-derived tissues than in tissues derived from young ones. Our findings support the predication that LPL has higher mammary activity by virtue of high transcript abundance [10]. LPL was the fifth most abundant transcript. Also, more than 8-fold increase in transcript abundance of CD36 has been previously reported during in vivo studies [3].

Among the five genes for acetate/fatty acid activation and intracellular transport, three showed relatively higher expression in mammary tissue. ACSS1 showed a >3-fold increase, and ACSS2 showed more than 7-fold increase in transcript abundance. These findings are comparable to previous findings; Bionaz and Loor have reported a higher (~13-fold) increase in ACSS2 over ACSS1 (~4-fold) [3]. ACSS1 and ACSS2 are responsible for activation of short-chain fatty acids; while ACSS1, primarily mitochondrial enzyme, activates acetate for energy production, ACSS2, the cytosolic enzyme, activates acetate for fatty acid synthesis [11]. With acetate being the chief substrate for energy production and fatty acid synthesis in the mammary tissue [9], overexpression of ACSS1 and ACSS2 during lactation is teleologically expected. In the same study [3], FABP3 was the second most abundant transcript with a nearly 80-fold change in transcript abundance at 60 days of lactation. However, the relative change in transcript abundance at onset and 15, 30, 120, and 240 days of lactation ranged about 20–40. In our study, FABP3 showed a >26-fold increase and was also the fourth most abundant transcript among all ones considered in the study. FABP3 is involved in the intracellular trafficking long-chain fatty acids and their acyl coesters.

Fatty acid synthesis and desaturation per se are the most important step in milk fat synthesis. However, of the five genes studied, only two appear to be involved during the milk fat synthesis response in the mammary tissue. FASN that catalyzes the formation of long-chain fatty acids from acetyl-CoA, malonyl-CoA, and NADPH was the second most abundant transcript and showed ~3-fold increase in expression. SCD5, responsible for introducing a double bond in fatty acyl-coenzyme A at the delta 9 position, was the most abundant transcript (~15.6%) with more than 5-fold increase in mRNA expression. Bionaz and Loor have also reported SCD5 to be the most abundant (~23%) among transcripts of genes involved in milk fat synthesis. However, in their study, the relative increase in expression has been reported to be much higher (~10–40-fold increase) [3].

GPAM, with more than 2% of all transcripts studied, was the only one of five genes involved in triacylglycerol synthesis found to be overexpressed (>10-fold increase). Bionaz and Loor have reported identical values of transcript abundance and relative expression of this gene [3]. Among the genes involved in sphingolipid synthesis, SPTLC1 appeared to be overexpressed (>3-fold) whereas the expression of SGPL1 appeared to be downregulated at about 1/20th of cutaneous expression.

Among the genes involved in lipid droplet formation, ADFP and XDH were overexpressed with 1.6- and a 10-fold increase in relative expression, respectively, over the cutaneous tissue. Both of these genes also showed preferential expression in adult-derived tissues. XDH includes xanthine dehydrogenase and xanthine oxidase; the enzyme can be converted from the dehydrogenase form (D) into the oxidase form (O) irreversibly by proteolysis or reversibly through the oxidation of sulfhydryl groups. XDH was the third most abundant of all transcripts (>11%). Bionaz and Loor have similarly reported >7% abundance of XDH transcripts and about 8-fold increase in its relative expression in the lactating mammary tissue [3].

Among transcriptional regulators that drive or sustain milk fat synthesis, INSIG1 and SREBF1 appeared to be overexpressed. Percent transcript abundance and relative increase in expression for the genes were about 1.8%, ~3-fold, and 4.2%, ~5-fold, respectively; 2.4- and 2.5-fold increases in the expression of these two genes have been reported previously [12]. Increase in INSIG1 [3] and SREBF1 [13] activities during lactation to much greater extents than being reported in the present study have also been reported earlier. A greater function of SREBF2 than SREBF1 in milk fat synthesis has been hypothesized [3]. Our study could not include SREBF2 due to insufficient information on this gene in the UniGene ESTProfile. However, based on our results, SREBF1 is expected to play a role at least equivalent to, if not greater than, SREBF2 in regulating the transcriptional response during milk fat production in the mammary tissue. None of the genes involved in xenobiotic and cholesterol transport and ketone body utilization appeared to be differentially expressed as part of the lactational milk fat synthesis response.

3.3. Gene Interaction Networks

Interaction network for all the 45 genes, obtained using STRING, has been shown in Figure 2. The interactions were further purged to map only those 13 genes that showed preferential expression in mammary tissue in UniGene ESTProfile (Figure 3); in Figure 3(a), the weight of the edges shows the strength of the interactions. The nature of these interactions has been depicted in Figure 3(b).

Network analysis shows that FASN, SREBF1, SREBF2, PPARG, and ACSS2 are the major components of the milk fat synthesis pathway. Two subnetworks are evident: one under the predominant control of PPARG and the other one majorly under the joint control of SREBF1 and SREBF2; both these subnetworks appear to converge at FASN. SCD5, the most abundant transcript, was the only gene under the direct control of PPARG, SREBF1, and SREBF2. Also, three of the four genes showing the maximum relative change in expression, namely, FABP3, CD36, and XDH, were chiefly under the control of PPARG. Thus, PPARG, though not found to be overexpressed based on TPM values, appears to play a major role in the transcriptional regulation of milk fat synthesis. Bionaz and Loor [3] have also advocated a role of PPARG in regulating the entire bovine milk fat synthesis machinery notwithstanding its downregulation and low mRNA abundance in mammary tissue. The genes involved in sphingolipid synthesis and ketone body utilization appeared to form two nearly independent clusters with sparse interaction with the rest of the network; ASAHL (NAAA), LASS2, and UGCG did not interact with any other gene at all. THRSP did not form part of the cluster of genes involved in transcriptional regulation. While all other gene products were involved directly or indirectly in interactions with each other, SPTLC1 and XDH did not interact with any of these gene products.

STRING was also used to determine coexpression patterns between these genes; a functional association of the gene products can be assumed if a group of genes exhibits strong coexpression. Only a low level of association could be inferred between some of the genes based on the coexpression pattern (Figure 4). Again, FASN appeared to be the central component of the milk fat synthesis pathway.

To conclude, in this study we have put forward a simplistic approach for determining the relative expression of genes based on their transcript abundance values in UniGene ESTProfile. Further, we used this approach for the expression profiling of genes involved in milk fat biosynthesis and secretion in bovines. Based on our findings, an updated model of the transcriptional profile of the genes involved in milk fat production by the mammary gland has been presented. For the genes studied, the results were in good agreement with the previously reported results from wetlab studies, indicating the satisfactory performance of our computational approach. Our study included cutaneous tissue as a control assuming its ontogenetic equivalence to the quiescent, nonlactating mammary gland; the congruity of our findings with those from previous studies projects this equivalence beyond the histological landscape to a biomolecular level. Previously, SREBF2 has been upheld as the major regulator of transcription during milk fat biosynthesis, refuting the role of SREBF1. Our results reinstate SREBF1 as a major transcriptional regulator, along with INSIG1, during the process. Using interaction network analysis of the genes, we could also show two separate transcriptional controls under PPARG and SREBFs. FASN, SREBF1, SREBF2, PPARG, and ACSS2 were the major components of the milk fat synthesis pathway. However, expression profiles could not be studied for nearly half of the genes due to incomplete UniGene ESTProfile. Also, the inferences would have been more conclusive if UniGene ESTProfile also included information on the stage of lactation during which the mammary glands had been sampled. Thus, further studies are warranted to verify the proposed model and to fill in the research gaps in the present study.

Conflicts of Interest

The authors declare no conflicts of interest.

Acknowledgments

Bhaskar Ganguly acknowledges the financial assistance received from Council of Scientific and Industrial Research (CSIR), India, in the form of a Senior Research Fellowship during this study.