Proteomic Analyses of Host and Pathogen Responses during Bovine Mastitis

Protein profiling by 2D-GE is characterized by a first dimension separation of proteins by isoelectric point, followed by a second dimension sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS- PAGE) separation by molecular size. Advances in 2D-GE technology including gel strips with immobilized pH gradients (IPG) for isoelectric focusing have dramatically increased the resolving power and reproducibility of 2D-GE. Additionally, the development of radioactive and fluorescent labeling has improved the ability to visualize proteins in a 2D gel, as well as the detection of low-abundance and post-translationally modified proteins [40].

In a MALDI-TOF/MS experiment, the protein or peptide(s) of interest is mixed with a suitable energy-absorbing matrix and allowed to co-crystallize by air-drying on a stainless steel plate. In addition to the ability to co-crystalize with the analyte and absorb the wavelength of the laser employed, compounds used as MALDI matrices must also be vacuum stable, soluble in solvents that are compatible with the sample being analyzed, and must enable and promote the ionization of the analyte of interest [41]. Matrices used in MALDI-TOF/MS are typically highly substituted aromatic molecules with carboxylic acid moieties capable of absorbing high levels of UV light at a specific wavelength, such as sinapinic acid, which is commonly used for the analysis of intact proteins, or alpha-cyano-4-hydroxycinnamic acid, which is often used for the MALDI analysis of peptides [41].

The generation of ions in MALDI-TOF/MS is initiated by short pulse irradiation with a laser, typically nitrogen or neodymium-doped yttrium aluminum garnet, and occurs when the matrix becomes electronically excited following absorption of photons from the UV laser. Proteolytically digested peptides in the sample accept a proton from the matrix and become singly charged positive ions as they are converted into the gas phase and ejected from the matrix. The ions are then directed into the TOF analyzer where they are separated by size and generate a mass spectrum. Separation of ions in a MALDI-TOF experiment occurs due to the velocity differences that exist between ions once the pulse of ions exit the ionization source and are dispersed in time down the flight tube. TOF mass analysis is based on the principle that after acceleration to a constant kinetic energy, ions travel at velocities that are inversely related to the square root of their mass-to-charge (m/z) values [42]. Thus, lighter ions “fly” down the TOF tube and reach the detector at faster speeds than heavier ions [42]. The subsequent identification of proteins from a MALDI mass spectrum is accomplished by the comparison of the set of peptide masses generated from a specific protein, or the peptide mass fingerprint, to a protein database containing theoretically calculated mass fingerprints of all known proteins.

Fragmentation reactions, or post-source decay (PSD), can often occur during travel down the flight tube of a TOF MS instrument, possibly due to collisions between ions and neutral matrix molecules, or between ions and residual gas molecules [43]. PSD fragment ions can be monitored using TOF instruments equipped with a reflectron, which is an electrostatic mirror consisting of a series of electrical lenses that possess progressively higher repelling potentials [42]. A reflectron not only improves the mass resolution of the MALDI MS spectrum, but likewise allows for the separation of PSD fragment ions and the generation of a PSD spectrum, which can be searched against a protein database to obtain peptide sequence information. Using a MALDI-TOF instrument, however, requires the generation of several PSD spectra from different mass regions in order to generate fragmentation information adequate for peptide identification. Subsequently, the sequencing of peptides using MALDI-TOF MS is most often performed using instruments equipped with dual TOF analyzers (MALDI-TOF/TOF). In contrast to single TOF MS instruments, MALDI-TOF/TOF instruments allow for the generation of complete fragment ion spectra in a single acquisition as well as improved precursor ion selection, and are thus preferable for accurate peptide sequencing using MALDI [44].

Despite the recent advancements in reproducibility and protein quantification, 2D-GE as a means of protein separation still suffers from issues including diminished capacity for the isolation of low abundance proteins, hydrophobic proteins, or proteins with extreme isolelectric points. Consequently, LC has emerged as the preferred method for in-solution separation of proteins and peptides prior to mass analysis using MS. Using the LC-MS/MS approach, proteins are proteolytically digested into peptides using a protease such as trypsin, which cleaves at every arginine and lysine residue, and separated online by 1- dimensional (1-D) or 2-dimensional (2-D) LC prior to introduction into the mass spectrometer for mass analysis. Separation of peptides is accomplished by either strong cation-exchange chromatography, reverse-phase (RP) chromatography, or a combination of both separation strategies, followed by ionization and mass analysis of the peptides and peptide fragments by ESI-MS/MS [45]. In traditional 2-D LC, peptides are separated in the first dimension by charge using ion exchange chromatography, and are then further separated in the second dimension by hydrophobicity using RP-LC. Alternatively, peptides can be separated by 2D-LC using RP-LC in both dimensions. In 1-D LC-MS/MS experiments, peptide mixtures are typically separated only by hydrophobicity by passage over a column packed with non-polar stationary phase. Using either approach, the number of proteins identified using LC-MS/MS is directly dependent on the efficiency of peptide separation prior to introduction into the mass spectrometer [46]. Poor chromatographic resolution increases the frequency of the co-elution of peptides off the LC column prior to introduction into the mass spectrometer, which can result in the production of a tandem mass spectrum of a peptide mixture that often fails to yield a match when searched against a protein database. Additionally, the potential for selection of a peptide from a low abundance protein for further fragmentation by a process called collision-induced dissociation (CID) will decrease with poor LC peak resolution, and peptides from dominant proteins will be preferentially selected for analysis.

In LC-MS/MS experiments, ESI is the dominant method of ionization. Ionization occurs in ESI after the peptide solution is dispersed as a fine spray of charged droplets after passage through a heated metal capillary tube to which voltage is applied. The charged droplets get desolvated by a dry inert gas, and multiply charged ions are produced. Nano-spray ionization (NSI) functions in a manner similar to ESI, but flow rates from the LC instrument into the ionization source of the mass spectrometer are much lower with nano-spray than those used for ESI. Because flow rates are reduced from microliters per minute to nanoliters per minute when using nanospray, droplet formation can occur without the additional of heat or a sheath gas, smaller charged droplets are formed, and ionization efficiency is greatly improved [47].

Ions resulting from either ESI or NSI are directed into the vacuum chamber of the MS instrument, and are resolved according to their m/z ratio. In tandem mass spectrometry (MS/MS), the masses of precursor ions are determined in the first MS scan, and an MS spectrum is generated. From each MS scan, a pre-determined number of ions can be selected for further fragmentation by CID. Fragmentation by CID involves the introduction of an inert gas such as argon (Ar) or helium (He) into the collision cell of the mass spectrometer which, through impact with the selected precursor ions, results in further fragmentation of the ions. The second stage of tandem MS is used to analyze the masses of the fragment, or product, ions produced by CID, and results in the production of a tandem or MS/MS mass spectrum. Peak lists generated from the fragment ion masses in tandem mass spectra are then distilled and searched against a protein database to determine the amino acid sequence of the peptides in the complex mixture. The assignment of the sequenced peptides to a given protein is the means by which protein identification is ultimately accomplished [46].

Comparative proteomic analyses are designed to elucidate changes in the relative abundance of proteins among different biological states, most commonly healthy versus diseased. Detection of the same peptides from a given protein is not always possible in comparative studies, however, because PTMs of peptides as a result of disease is expected. Characterization of PTMs is crucial for biomarker discovery, because much of the regulation of the biological activity of proteins is mediated by the modification of peptide amino acid residues, including the phosphorylation of serine and threonine, and the glycosylation of asparagine, arginine, or tyrosine. Unfortunately characterization of PTMs has been hindered in past experiments due to the fact that modifications are labile and are often lost in a CID experiment. Electron-transfer dissociation (ETD), which is a superior fragmentation strategy for the analysis of PTMs, however, was recently introduced, and shows promise as a strategy for the characterization of protein modification during disease [48]. The ETD technique uses electrons to promote fragmentation along the peptide backbone, which produces a series of c- and z- ions, instead of CID fragmentation, which most often produces a series of b- and y- ions. Protonation during ionization occurs most often at the N-terminal amino group; however, the charge can be localized to any of the nitrogen atoms that comprise the amide bond [42]. As a result, all three of the peptide backbone bonds can be cleaved, and either the N- or C- terminus fragments may retain the charge (Fig. 1). When the charge is retained on the N-terminus, the ions are denoted as either a _n , b _n , or c _n, whereas C-terminus ions are designated as x _n , y _n , or z _n. While side chain information (R group) is lost during the formation of b- and y- ions, the fragmentation of the peptide backbone using ETD and the generation of c- and z- ions allows for amino acid side chains and modifications such as glycosylation and phosphorylation to remain intact, making it possible not only to deduce the amino acid sequence of a peptide, but also to detect any modified residues [48].

Other limitations that obstruct comparative proteomic experiments include the types of proteins present in the sample to be analyzed, protein abundance or dynamic range, and the chosen proteomic strategy. A select number of highly abundant proteins can sometimes represent a large percentage of the total protein concentration in a given sample, which can lead to a higher probability of selection of peptides from the abundant proteins for MS/MS analysis, and hindered detection of low abundant proteins. For example, serum albumin can account for more than half the total protein concentration of plasma (30-50 g/L), whereas minor proteins such as interleukins are present in only picogram concentrations [49]. However, identification of low abundance proteins is critical to biomarker discovery, as low-copy-number gene products have a high probability of being potential drug targets and biological markers of disease [50].

Preparation strategies are typically employed to deplete high abundance proteins prior to LC-MS/MS in an effort to reduce sample complexity and enhance detection of low abundance proteins. Some traditional depletion strategies include the precipitation of abundant proteins, the use of size exclusion filters or 1D-SDS PAGE to fractionate a complex mixture of intact proteins by mass, and targeted removal of abundant proteins using immuno-depletion techniques. There are several commercially available depletion kits optimized for use with serum and plasma that selectively deplete a number of the most abundant blood proteins including serum albumin, immunoglobulins, complement, and acute phase proteins. Additionally, the use of a bead-bound library of peptides with random rearrangements of six amino acids and the capacity for binding a large number of different proteins is gaining popularity as a means to equalize the abundance of proteins in a complex biological sample [51].

Specific depletion techniques used in prior proteomic analysis of bovine milk have included acid precipitation of the casein proteins [23, 28, 29], and the use of bead-bound peptide libraries [27]. When considering comparative proteomic analyses of normal and mastitis milk, however, the divergence that exists in the number of hydrophobic and hydrophilic proteins in a given sample at each physiological state, the varying sizes and charge states of proteins present in each matrix, PTMs, and the cellular distribution of proteins in one sample versus another makes the application of a universal sample preparation strategy extremely unfeasible. Additionally, though depletion strategies are commercially available, they are often optimized for only one biological matrix, and are not amenable to all types of samples, most notably bovine milk, and can lead to the non-selective depletion of proteins of potential interest [27]. Likewise, profound changes in the dynamic range, modification, and protein composition of a given sample due to the induction of a disease such as mastitis can further confound the use of sample preparation strategies [27]. When no effective sample preparation strategy is applicable, several proteomic approaches are often combined to increase protein detection and subsequent proteome coverage. For example, LC techniques are more efficient for detection of hydrophobic, low molecular mass, and basic proteins than 2D gel based methods, whereas 2D-GE enables the detection and visualization of PTMs [50]. Despite limitations and inherent drawbacks, proteomic strategies offer a wider range of capabilities than classic approaches to protein characterization such as ELISA and, with further advances, could factor prominently in the establishment of biomarkers of mastitis.

Advances in MS have provided a means for the accurate and sensitive identification of differentially expressed proteins in complex biological samples, but another important criterion for the establishment of disease biomarkers is reliable quantification. Relative and absolute quantification of modulation in protein abundance using proteomic strategies is a topic that has garnered significant attention in recent years [52‐55]. Proteomic-based quantification methods can be assigned to one of two broad categories: the incorporation of stable isotope labels into proteins or peptides prior to MS analysis, or the alternative, which is to use a label-free quantitative method [55].

The basis of most labeled quantification methods is the theory that a labeled peptide will behave in the same fashion chemically as its unlabeled counterpart, and therefore the two peptides will have identical chromatographic and MS properties [55]. The addition of a label, however, does impart a mass difference between the two peptides, and thus relative abundance can be inferred by comparing the respective signal intensities of the labeled and unlabeled peptides in the same MS run [55]. Labels can be incorporated in several ways, with the most popular means being either metabolically, chemically, or enzymatically [53, 55]. Examples of labeling strategies include: metabolic labeling commonly known as stable isotope labeling by amino acids in cell culture or SILAC [56]; proteolytic labeling with ¹⁸O [57]; or isotope incorporation by several means including chemical derivatization also known as isotope coded affinity tags or ICAT [10], isobaric tags for relative and absolute quantitation (iTRAQ) [58], or global internal standard technology [59].

The derivatization of the primary amine groups of proteins or peptides with isobaric tags, or iTRAQ, has become a very popular LC-MS/MS quantification strategy in recent years. Quantification using iTRAQ is performed by the fragmentation of the attached isobaric tag, which generates a low molecular mass reporter ion [58]. The popularity of iTRAQ in proteomic screens, including the analysis of mastitis pathogens [36], the bovine milk fat globular membrane [60], and bovine milk following in vivo LPS challenge [28], is due primarily to the capacity to simultaneously analyze four or more of differential protein pools.

However, though very accurate, labeling strategies can be cost-limiting, unsuitable for some types of biological matrices, and cannot be performed retrospectively. In terms of comparative proteomic analysis of normal versus mastitic milk aimed at biomarker discovery, labeling strategies are not always feasible for protein quantification. Due to dramatic changes in protein composition during clinical mastitis, accurate comparisons of the abundance of a peptide that is present at one physiological state but not the other is problematic [27]. As a result, label-free strategies, which are based on the correlation between the abundance of a protein or peptide in a sample and the MS signal [55], have increased in popularity in recent years. Ion intensity, determined by extracted ion chromatograms (XIC), is one of the most accurate and widely used methods of label-free quantification. Using XIC, the number and intensity of selected precursor ions at a particular m/z are summed, and the peaks areas used as a measure of relative abundance [61]. Another popular label-free approach is spectral counting, which is defined as the number of MS/MS spectra that contribute to the identification of a given protein [62, 63]. Spectral counting is based on the assumption that the abundant proteins, when proteolytically digested, will yield numerous copies of the same peptide, and that peptides from abundant proteins have higher probabilities of triggering multiple MS/MS events than peptides from lower abundance proteins. Similar to spectral counts, the number of unique peptides assigned to each protein in a sample has likewise been used as a measure of relative protein abundance [26, 62]. An inherent drawback of label-free methods, though, is the assumption that the linearity of response will be the same for each protein, which often does not hold true because the chromatographic behavior of peptides tends to vary [55]. Additionally, because the amino acid composition of every peptide differs, the ionization potential of each peptide is unique, and may not always be correlated to abundance. Thus, many spectra must be acquired and data normalized when using label-free methods [61]. Nevertheless, label-free quantification does not require any extra sample processing and can be performed retrospectively; two attributes which make non-labeled quantification methods the continued focus of development in biomarker discovery research [55].

The first proteomic endeavor aimed at the identification of novel markers for bovine mastitis was a comparison of differentially expressed proteins in normal and mastitic bovine milk from seven cows with naturally-occurring clinical infections using 2D-GE followed by MALDI-TOF-MS PSD [29]. Milk sample preparation prior to proteomic analysis was relatively simple, and consisted of removal of the milk fat by centrifugation at 4°C, and acid precipitation of the caseins. The results of the proteomic analyses indicated the isolated presence of lipocalin-type prostaglandin D synthase in 2D gels generated from milk samples collected from cows with clinical mastitis [29]. Other proteins detected in the 2D-GE profiles of the milk samples analyzed included serum albumin and κ- casein [29]. Though the results were extremely limited and only a single differentially expressed protein in mastitic bovine milk was reported, the study nonetheless marked the first instance of published data on a marker of inflammation identified in bovine milk during mastitis using proteomic strategies.

A later attempt to identify differentially expressed proteins in whey from healthy cows versus cows with clinical mastitis also employed 2D-GE followed by enzymatic digestion of isolated proteins and MALDI-TOF-MS [22]. Mastitis was defined strictly by identification of clinical signs and milk sample preparation prior to analysis included fat removal and precipitation of the casein proteins by the addition of salt, followed by dialysis to remove traces of the ammonium sulphate [23]. Despite attempts to remove the casein fraction from the milk prior to proteomic analysis, however, α_S1-casein, β-casein, and κ-casein were still found in rather high abundance in normal milk. Also identified were the proteins serum albumin, transferrin, microsomal triglyceride protein, β-lactoglobulin, and α-lactalbumin [23]. Marked increases in serum albumin and transferrin, concurrent with apparent decreases in the caseins, β-lactoglobulin, and α-lactalbumin were apparent in the 2D-GE profiles generated from whey samples from cows with mastitis, and supported the well established theory that serum protein concentration increases in milk during clinical mastitis infections as a result of the breakdown of the blood-milk barrier [23]. While no additional biomarker information was gained and protein identification was restricted to major milk proteins, the results were the first to demonstrate temporal changes in bovine milk protein expression during mastitis without a reliance on antibody-based strategies.

The first extensive report of protein modulation in separate sub-cellular fractions of bovine milk collected during different physiological states included direct LC-MS/MS, as well as 2D-GE, excision of protein spots, and identification by either MALDI-TOF-MS or LC-MS/MS [24]. The milk fractions analyzed included milk from a cow in peak lactation, colostrum from a fresh cow, and milk from a cow with naturally-occurring clinical mastitis. Though the sample size was limited, the results of the analyses were the most comprehensive to date in terms of the number of low abundance proteins indentified, and the number of host response proteins that were isolated from mastitic milk [24]. An additional novel aspect to the proteomic analyses conducted by Smolenski and colleagues was the fact that no sample clean-up or attempts to deplete high abundance proteins were carried out on the samples prior to analysis. Despite the presence of caseins in the fractions analyzed, the study marked the first reported identification of proteins such as apolipoprotein A-1, cathelicidin-1, heat shock protein 70kD protein, peptidoglycan recognition receptor protein (PGRP), calgranulin B and C, and serum amyloid A (SAA) in milk fractions collected from a cow with naturally-occurring mastitis [24].

The first reported comparison of protein expression patterns in milk from cows before and after experimental induction of E. coli mastitis utilized 2D-GE followed by peptide sequencing using MALDI-TOF PSD to characterize differentially expressed proteins [25]. Unlike prior 2D-GE-MALDI analyses of mastitic bovine milk, no removal of high-abundance proteins was performed prior to analysis. The analyses involved milk samples collected from 8 cows before, and at 18 h following intra-mammary inoculation with E. coli, and only proteins present in whey fractions of all 8 cows were sequenced to avoid reporting a protein that represented a potentially unique response to infection. Despite the lack of sample clean-up, the low abundance proteins transthyretin, lactadherin, β-2-microglobulin precursor, α-1-acid-glycoprotein (A1AG), and complement C3 precursor were identified in whey samples from healthy cows. Whey samples at 18 h post infection were characterized by an abundance of serum albumin, in spots of varying mass and isoelectric point, as well as increased transthyretin and complement C3 precursor levels. Also detected at 18 h post inoculation were the antimicrobial peptides (AMPs) cathelicidin −1, −2, −3, and −4, and the proteins β-fibrinogen, α-2-HS-glycoprotein, S100-A12, and α-1-antiproteinase [25]. The most notable results of the analyses, however, were the detection of the cathelicidin cationic AMPs, and the identification of the acute phase protein (APP) A1AG in both normal and mastitic whey samples, as discoveries of the AMPs and A1AG were not reported in previous comparative 2D-GE proteomic analyses of bovine milk [23, 24, 29]. As well, prior reports of APP expression in milk during bovine mastitis were all antibody-based detections, and only documented the identification of the APPs SAA, haptoglobin (HPT), and lipopolysaccharide binding protein [68‐70].

The first proteomic analyses of a more extensive longitudinal series of bovine milk samples collected from 8 mid-lactation cows before and over the course of experimental infection with E. coli, utilized an ultra pressure LC instrument coupled to a quadrupole TOF-MS [26]. Similar to earlier analyses [24], only a select number of proteins related to the host response were identified in whey from milk following E. coli challenge. LC-MS/MS conducted on whey from milk samples collected just prior to infusion with E. coli and at various time points following infection resulted in the identification of the high to medium abundance proteins αS₁-, αS₂- β-, and κ-casein, and the whey proteins serum albumin, β-lactoglobulin and α-lactalbumin. Additionally, a select number of lower abundance markers of inflammation including lactoferrin, transferrin, apolipoprotein A-I, fibrinogen, Glycam-1, PGRP, and cathelicidin-1 were also identified. Despite limited protein identifications, normalized peptide counts for each protein identified were used to evaluate temporal changes in milk proteins following infection [26]. Additionally, to assess the accuracy of LC-MS/MS-based label-free quantification strategies, modulation in the abundance of serum albumin, lactoferrin, and transferrin in milk during disease evaluated using proteomic-based methods were compared with protein abundance measured using ELISAs. The outcome of the comparison of label-free, proteomic-based quantification with abundance profiles generated by ELISAs revealed that label-free LC-MS/MS methods yielded results that were both comparable to antibody-based detection, and a viable means of tracking changes in relative protein abundance in milk during disease [26]. Despite the identification of primarily abundant milk proteins, the results likewise indicated that, with further methodology refinement, LC-MS/MS could be used to evaluate temporal changes in proteins related to host response for which no antibody existed [26].

More recent comparative proteomic analyses of whey from bovine milk before, and at time points after experimental challenge with E. coli [27] or LPS [28], have resulted in the more robust detection of bovine milk proteins, including the identification of a greater number of proteins related to the host response to infection. The separate comparative analyses both involved evaluation of samples collected before and following experimental induction of clinical mastitis, both reported quantification of changes in the relative abundance of milk proteins related to host response, and both detected equivalent numbers and types of proteins (Table 1). The distinguishing factor between the two comparative proteomic analyses was the application of the iTRAQ labeling strategy for quantification of protein modulation [28], versus the use of spectral counts, a label-free proteomic approach, to assess changes in the relative abundance of minor milk proteins [27]. Other divergent aspects of the in-vivo challenge studies included the number of animals enrolled in each study, as well as sample preparation strategies. Methodologies employed in the proteomic analysis of host response to LPS-mediated intra-mammary inflammation differed slightly from previous E. coli challenge studies [25‐27] in that only 3 cows were enrolled in the study, and acid precipitation of caseins was performed prior to LC-MS/MS analyses.

Similar to the prior reports on the use of label-free quantification strategies [26], the accuracy of LC-MS/MS label-free quantification performed using samples collected in the E. coli challenge study was compared to data generated using antibody-based strategies [27]. Unlike prior reports, however, total spectral counts were used as a measure of relative abundance and, due to increased sensitivity of the instrument system, the temporal expression of the low abundance acute phase proteins HPT and SAA were evaluated as opposed to major milk proteins [27]. Comparisons of ELISA data and total spectral counts for the two APPs revealed trends in temporal expression with very similar overall patterns, which reaffirmed the utility of using LC-MS/MS data to model expression patterns of proteins related to the host response in bovine milk during clinical mastitis. Conversely, quantification of host responses to LPS-mediated inflammation were calculated using fold-change in equivalent iTRAQ reporter ion intensities for peptides identified in pre-challenge milk samples and time points following challenge [28].

Though the time frame of the reported modulation in protein expression differed due to inoculation with E. coli versus purified LPS, which is known to stimulate a more rapid inflammatory response than E. coli in the bovine mammary gland [18], and the methods of determining changes in protein abundance were not the same, the identified proteins with altered expression patterns were similar for the two separate analyses. Proteins related to the host response to infection detected in whey from bovine milk at time points following infection were predominantly vascular derived, acute phase, antimicrobial, complement, or related to immune response, and fell into categories that could be broadly classified as secondary effects of cytokine induction [27, 28]. Cytokine expression during the bovine innate immune response to invading pathogens has been well characterized, and the induction of vascular leak and APP synthesis, as well the stimulation of neutrophils as a result of cytokine production have been previously established [14]. Following intra-mammary inoculation with E. coli, several vascular-derived proteins including all three chains of the blood coagulation protein fibrinogen, apolipoprotein A-1, and serotransferrin were detected in bovine milk as early as 12 h after challenge, with peak abundance detected at 24 h following induction of mastitis [27]. The same proteins were detected following challenge with LPS, and were 2-fold or higher in expression level 7 h after infusion [28]. Similarly, peak expression of complement C3 determined using spectral counts, was detected in bovine milk 18 h after induction of coliform mastitis [27], while 2-fold changes in reporter ion intensity was apparent for complement C3 peptides in bovine milk 7 h after LPS challenge [28]. The reported increases of serum proteins in milk following E. coli or LPS challenge are most likely results of vascular leakage induced by tumor necrosis factor-alpha (TNF-α) or interleukin-1 beta (IL-1β) expression, and corresponded well with prior reports of peak cytokine expression during E. coli and LPS induced mastitis [14, 18]. In all, 13 of the same proteins were found to have biologically relevant changes in relative abundance during clinical mastitis in both comparative proteomic analyses, including complement factors, the AMPs cathelicidin-1 and PGRP, apolipoprotein A-1, the APPs serotransferrin, HPT, and SAA, and the somewhat poorly characterized proteins kininogen, inter-alpha-trypsin inhibitor heavy chain-4 (ITIH4), and clusterin [27, 28].

The most interesting aspect of both quantitative proteomic analyses of bovine milk following experimental induction of mastitis was the modulation of apolipoproteins and the somewhat novel candidate biomarkers ITIH4, kininogen, and clusterin. Quantification of changes in abundance of the potentially novel candidates was accomplished using both iTRAQ [28] and spectral counts [27]. The temporal expression patterns of the candidate proteins, modeled using LC-MS/MS spectral count data (Fig. 2), likewise appeared to be in accord with prior reports of inflammatory mediator expression during coliform mastitis [14]. Though apolipoproteins are known to be involved in lipid transport and are major components of high density lipoproteins, other roles for the apolipoproteins during disease and inflammation have been proposed [71]. The specific role of the apolipoproteins during inflammation related to coliform mastitis has not yet been determined, but implications are that the apolipoproteins could inhibit neutrophil activation, as well as the release of inflammatory cytokines [72, 73]. By similarity, bovine ITIH4 is assumed to be a serine protease inhibitor involved in the acute phase response. Reports of ITIH4 expression in cattle, however, have been limited to isolation of the APP from the serum of heifers with experimentally induced summer mastitis [74]. Prior to recent comparative proteomic analyses of bovine milk following in vivo challenge, ITIH4 was never identified in bovine milk, and its exact role during coliform mastitis remains unknown. Kininogen, on the other hand, belongs to the family of plasma kallikreins, and is known to play key roles in complement activation [75], and the release of bradykinin [76]. Bradykinin levels in milk from cows with experimentally induced coliform mastitis have yet to be investigated, but the kinin peptides are presumed to be potent mediators of vasodilation, pain, and udder edema during clinical mastitis [77]. Similar to ITIH4, the function of clusterin during coliform mastitis also remains unclear. Prior reports have indicated that clusterin could possess anti-inflammatory properties [78, 79], but the only inference in regards to the role of clusterin in the bovine mammary gland is that clusterin could be associated with mammary gland involution, the clearance of cellular debris, or apoptotic cell death [80]. Though more focused follow-up analyses are required, the results of the comparative proteomic analysis of bovine milk during experimentally-induced clinical mastitis have provided information that could prove useful in the design and execution of future studies, and have shed light on potential candidates for the establishment of inflammatory biomarkers in bovine milk.

Associations between proteins related to the host defense present in human and bovine milk have recently been reported [31]. The total numbers of proteins identified following proteomic analyses of both human and bovine milk differed by only one protein, and more than half of the reported proteins were detected in samples from both species [31]. Using LC-MS/MS, analyses of whey from milk and the milk fat globular membrane from both human and bovine milk resulted in the identification of 44 proteins related to the host response in human milk fractions, and 51 host defense proteins in the equivalent bovine samples. Thirty-three (33) defense-related proteins were identified in milk fractions from both species, and included several APPs, complements factors, members of the cathelicidin family of cationic AMPs, clusterin, immunoglobins, osteopontin, several mucins, lactoferrin, calgranulins, as well as protease inhibitors [31]. Though similar proteins were detected in the human and bovine milk fractions, the abundance and distribution of the proteins differed between the two species. Specifically, the authors noted the higher prevalence of antibacterial proteins in the bovine milk fractions, compared to a marked increase in the detection and abundance of immunoglobins in human milk fractions [31]. Possible explanations for the differential prominence of AMPs, including lactoperoxidase in bovine milk, and immunoglobins in human milk, were higher thiocyanate levels in the ruminant diet and differences in immune system development in human babies and bovine calves, respectively [31].