Reducing the complexity of protein mixtures by depletion and fractionation of intact proteins greatly simplifies the task for MS/MS analysis. There are essentially three patterns of depletion in Tab. 2 and Fig. 3: no depletion of the most abundant proteins, depletion only of albumin or Ig or both, and depletion of the top-6 proteins, which are albumin, IgG, IgA, haptoglobin, alpha-1 anti-trypsin, and transferrin (Agilent column). There is clear evidence from the main database and from a series of special project studies by PPP investigators that depletion makes it significantly more feasible to visualize, detect, and then identify lower abundance proteins (Echan et al. , Li et al. , Zolotarjova et al. , Huang et al. , Tang et al. , Misek et al. , Yang et al. , Barnea et al. , Moritz et al. , Cho et al. , Kim et al. ). However, when only 2-DE is employed, the many "new" spots detected after depletion are unmasked isoforms of medium-abundance proteins, rather than lower abundance proteins [5,26]. There is a counterbalancing problem, namely non-target or inadvertent removal of other proteins , which could be due to peptides and proteins bound to the target proteins, especially albumin; cross-reactivity with the bound antibodies; or non-specific binding to the column or resin or dye. Details of the protocols, proprietary buffers, column capacity, and previous use of the columns may be important variables. With older and much less expensive albumin-removal agents, such as Cibacron Blue dye, there is thought to be binding to the dye (as well as any binding to the albumin).
Moritz et al.  provide a preliminary report using free-flow electrophoresis (FFE-IEF) and rapid (6 min) RP-HPLC to fractionate citrate-plasma (Lab 33). They analyzed both bound and flow-through fractions from immunoaffinity depletion of the top-6 proteins. From 15 of 96 FFE fractions, with 72 780 MS/MS spectra analyzed with MASCOT and Digger and subjected to manual validation, they obtained 55 proteins based on two or more peptides and 23 more based on one peptide, across a mass range of from 4 to 190 kDa; these included several with estimated concentrations of 0.5-1 ng/mL. They highlight the identification in the bound fraction of a 35-amino acid serine protease protection peptide (CRISPP) that is cleaved from the C-terminus of alpha-1 anti-trypsin, non-covalently complexed with alpha-1 anti-trypsin, and not included in the IPI 2.21 database. They detected protein complexes by using non-denaturing, non-reducing buffers. They enhanced their yield by building a data-dependent exclusion list to prevent re-identifying abundant peptides.
Tang et al.  investigated many experimental parameters of depletion, fractionation, and such MS variables as gas phase fractionation. They combined solution isoelectrofocusing and 1-D SDS gel electrophoresis to generate "pixels" of proteins with defined pi and Mr ranges, then fractionated tryptic digests with 2D LC, followed by LCQ-Deca-XP+ or LTQ-linear IT-MS/MS for Bl-heparin-plasma and Bl-serum reference specimens, respectively. These methods yielded 575 and 2890high-confidence protein identifications (see Section 3.1) using the stringent HUPO PPP SEQUEST parameters; they did not remove potential homologous database entries; 319 of the 575 plasma proteins were identified in the serum specimen. Of these 319, half are single-peptide proteins in plasma, but many more are multiple-peptide proteins in serum, with the LTQ instrument, and have rich MS/ MS fragmentation patterns. They estimated that proteins in the low ng/mL range were detected from 45 mg of plasma protein using the LCQ-Deca XP+, whereas proteins in the low pg/mL range were detected from 204 mg of serum using the LTQ. They uniquely utilized a SEQUEST Sf score, which combines Xcorr, deltaCn, Sp, Rsp, and ions scores using a neural network to reflect the strength of peptide assignment on a scale of 0 to 1; scores >0.7 were considered to have a high probability of being correct, regardless of other parameters; when Sf scores replaced Rsp > 4, they obtained 744 and 4377 non-redundant protein identifications from the plasma and serum specimens, respectively.
Misek et al.  identified many isoforms and compared relative abundance of proteins in serum, EDTA-plasma, and citrate-plasma labeled, respectively, with the fluorescent dyes Cy3, Cy5, and Cy2 after top-6 immunoaffinity depletion. The three labeled, depleted samples were subjected to three-dimensional protein fractionation by pi, hydrophobicity, and Mr About 3000bands on 1-D SDS gels with ±> two-fold differences in intensity of fluorescence in dye pairs were excised and analyzed by MS/MS, yielding a total of only 82 non-redundant proteins; 28 proteins were identified in ten or more different fractions. Complement C3 and clusterin are presented as examples of proteins whose biologically significant cleavage products can be identified with this method. Not surprisingly, the yield in MS/MS was greater for proteins with higher intensity (abundance). Multiple isoforms reduce the concentration of a protein in any particular spot or fraction and may react very differently with antibodies used to quantify the proteins or detect the proteins, as on microarrays.
Subfractionation of the complex mixtures that are plasma and serum can be performed chemically or with capture agents. A very good example is the glycoprotein subproteome. Labs 2 and 11 (Tables 1 and 2) utilized hydrazide chemistry and binding with three lectins, respectively, to enrich for glycoproteins. The chemical method, which captures N-linked glycoproteins subsequently treated with PNGase F, was published by Zhang et al. in 2003 . Yang et al.  used wheat germ agglutinin, Jacalin lectin, and Con A together on agarose to isolate and characterize approximately 150 glycoproteins in PPP serum and plasma reference specimens after analysis by LCQ-MS/MS, with confirmation in some cases using a linear IT LTQ instrument. There was close similarity for the composition of the glycoproteome across the plasma and serum specimen sets, except for fibrinogen, which was absent from serum (after clotting). Samples from the individuals from three different ethnic groups showed only a few individual differences. Together the two laboratories identified 254 glycoproteins, of which 164 were identified by other laboratories in this collaboration. That means that 90 were found only in the glycoprotein-enriched studies. Glycoprotein has an important incidental benefit in that the non-glycosylated albumin protein should be excluded; in fact, some albumin remains, given its very high abundance and its tendency to bind glycoproteins.
Cho et al.  combined immunoaffinity depletion of the top-6 proteins with freeflow electrophoresis or 2-DE of fractions, and MALD-TOF-MS PMF; they found only minor differences across the donor and specimen preparation variables. With 2-DE they found few non-target proteins in the immunoaffinity bound fraction.
Kim et al.  sought to identify and eliminate false-positive peptide identifications and subsequent protein matches by analyzing molecular weight on 1-D SDS gels after immunoaffinity depletion. Of 494proteins identified with 2-D-LC/ESI-MS/MS of 28 1-D fractions, using SEQUESTwith stringent PPP filters, 202 were excluded as single-peptide hits as well as estimated Mr too deviating from theoretical M„ but 166 one-peptide matches were retained based on good Mr match. This approach requires careful curation for biologically cleaved proteins. Their method actually increased the number of accepted proteins, since only 128 (26% of 494) were based on two or more peptides among the total of 292 protein identifications claimed for the B1-serum specimen.
Echan et al.  compared the immunoafffinity top-6 depletion column and corresponding spin cartridge from Agilent with a prototype ProteoPrep dual anti-albumin/anti-IgG antibody column from Sigma Aldrich, with five commercially available kits using Cibacron Blue for albumin and/or Protein A or G for immu-noglobulin depletion, and with no depletion. These variables correspond to the categories depicted in Fig. 3. The polyclonal antibody column gave nearly complete depletion, showed low non-specific binding, based on 2-DE profiles, and permitted many new spots to be visualized. However, the number of new proteins was quite small, due to the emergence of newly visualized spots representing numerous iso-forms of the now-most abundant remaining proteins. They estimated that silver staining on 2-D gels should have been able to detect proteins originally present in the serum or plasma at 40 ng/mL or higher, while the protein identified with lowest known concentration is at about 30 mg/mL, before accounting for heterogeneity of isoforms. The two-protein column had more capacity for albumin and IgG removal, but also removed many non-target proteins, which may be improved with optimized buffers. Apparently, buffer variables are very influential with all of the antibody columns. Given published reports of up to 63 proteins bound to albumin , secondary binding conditions can introduce major variability in results. Clearly, more potent technology combinations are required to adequately evaluate the non-target binding of proteins during immunoaffinity depletion, as well as to reach down to the ng/mL to pg/mL concentration range. Echan et al.  point out that the inexpensive and convenient dye and protein A/G methods can be used for fractionation rather than depletion. They also note the potential to specifically deplete many more proteins with expanded immunoaffinity columns.
Additional papers by Zolotarjova et al.  and by Huang et al. , scientists at Agilent and at GenWay Biotech, respectively, present laboratory results with their immunoaffinity products. The polyclonal rabbit antibody column from Agilent removes albumin, IgG, IgA, haptoglobin, transferrin, and alpha-1 anti-trypsin. The polyclonal chicken IgY antibodies on microbeads from GenWay remove six (albumin, IgG, IgA, IgM, transferrin, and fibrinongen) or 12 (also alpha-1 anti-trypsin, alpha-2 macroglobulin, haptoglobin, apolipoproteins A-I and A-II, and orosomu-coid/alpha-1 acid glycoprotein). Both groups report highly effective removal and little to no non-target binding. These products were introduced during the conduct of the PPP pilot phase and were made available to investigators.
One way to maximize identifications is to analyze bound fractions as well as pass-through fractions, as done by He et al.  and by Labs 29 and 46 (Tables 1 and 2). He et al.  report large numbers of proteins in the top-6 immunoaffinity bound fraction when extensive LTQ-MS/MS is applied, utilizing the stringent PPPSEQUEST filters. They may not have used the full system optimized by the column manufacturer.
Was this article helpful?