Analysis of pressure ulcer wound fluid using two‐dimensional electrophoresis

Abstract

The incidence rate of pressure ulcers in the USA ranges from 0·4% to 38% in acute care settings and from 2·2% to 23·9% in long‐term care settings, and their treatment costs are in the billions of dollars yearly. The proteome of wound fluid may contain early indicators or biomarkers associated with healing in pressure ulcers that would enable treatment regimes to be optimised for each individual. Wound fluid was collected from the interior and periphery of 19 chronic pressure ulcers at 15 time points during 42 days for an analysis of protein expression. Proteins were fractionated using two‐dimensional polyacrylamide gel electrophoresis. A comparison of the spot distributions indicates a biochemical difference between the interior and the periphery of wounds. Pressure ulcers that healed show a greater number of spots for interior and peripheral locations combined over time when compared with wounds that did not heal. Using this technique, protein S100A9 was identified as a potential biomarker of wound healing. The identification of differences within the proteome of healing versus non healing pressure ulcers could have great significance in the use of current treatments, as well as the development of new therapeutic interventions.

INTRODUCTION

The incidence rate of pressure ulcers in the USA ranges from 0.4% to 38% in acute care settings and from 2.2% to 23.9% in long‐term care settings, and their treatment costs are in the billions of dollars yearly (1). Proteomics, the study of protein structures and function, is a rapidly developing field with a wide range of applications in wound healing. The use of proteomics to assess wound healing has many potential benefits, which may include earlier evidence of healing and better understanding of how various treatments affect the wound at the protein level. The most common predictors of wound healing currently used are based on clinical appearance of the wound. Wound depth and size over time are most often used to assess wound progress with no change or an increase in size indicating a lack of healing. The identification of potential biomarkers associated with the development of pressure ulcers, healing or delayed healing in chronic wounds could have great significance in the use of current treatments, as well as in the development of new therapeutic interventions. As proteomics becomes more common in wound analysis, it can eventually be used as a bedside tool and change how wounds are assessed and treatments are evaluated.

Two‐dimensional polyacrylamide gel electrophoresis (2D‐PAGE) is one of the most widely used techniques for protein separation and quantification in the search for biomarkers (2). To date, there has been no report of the use of 2D‐PAGE to analyse wound fluid from pressure ulcers. In the current study, we analysed wound fluid samples from 19 pressure ulcers over 42 days using 2D‐PAGE.

METHODS

Subject population

Thirty‐four subjects with a total of 46 pressure ulcers were enrolled upon receiving Institutional Review Board approval for the study and consent from Daemen College Catholic Health, University of Buffalo and the US Army Medical Research and Materiel Command Human Research Protection Office. All pressure ulcers were present for a minimum of 4 weeks. Onset data were available for 30 ulcers and they were present, at most, 620 days prior to enrollment. Stage 2 (32%), stage 3 (23%) and stage 4 (45%) pressure ulcers were included. Ten men and 24 women were enrolled with the average age of study subjects being 72·3 years. Hypertension, diabetes and coronary artery disease were present in 52·2%, 54·3% and 30·4% of subjects, respectively. A total of 17 subjects with 19 pressure ulcers were included in a temporal study of protein expression. This subset includes 6 men and 11 women with an average age of 73·5 years. Ulcers stages 2 (42%), stage 3 (21%) and stage 4 (37%) are represented. Hypertension, diabetes and coronary artery disease were present in 52·2%, 54·3% and 30·4% of this subset, respectively.

Protocol

Subjects enrolled in the study were seen on days 0, 1, 2, 3, 4, 7, 8, 9, 10, 11, 14, 21, 28, 35 and 42. On each visit, the wounds were digitally photographed and sampled as previously reported on both the periphery and interior of the wound (3). Images were analysed using VeV MD (Version 1.1.14, Vista Medical Ltd., Manitoba, Winnipeg, Canada) to measure the wound size. The change in wound area expressed as a percentage of area on day 0 was used to place patients in one of the three categories for wound outcome—not healed, healing or healed (Table 1). Wound proteins were collected using sterile polyester tipped applicators gently rolled over the wound surface until saturated. The tip of the swab was broken off and placed in a 2 ml vial prefilled with 150 µl phosphate buffered saline (PBS) (10 mM, pH 7·4) to prevent sample dehydration. Wound fluid samples were placed in chilled coolers and transported immediately to the lab for analysis. Proteins were resuspended from the polyester tip by the addition of 350 µl dH₂O and vortexed for 30 seconds. The swabs were inverted and all liquid removed from the polyester tip via centrifugation for 10 minutes at 6000g. The swabs were removed from the vial and cellular debris pelleted by repeating the centrifugation. The supernatant was filtered using spin columns with 3‐kDa cut‐off membranes centrifuged for 99 minutes at 14 000g (Millipore, Amicon Microcon^® Ultracel YM‐3, Billerica, MA). These columns were used to concentrate the samples and remove ionic components in preparation for isoelectric focusing (IEF). After centrifugation, 50 µl dH₂O was added to the column membranes, and the columns were vortexed briefly and incubated at room temperature for 5 minutes to resuspend the wound proteins. The columns were then inverted into 1·5 ml Protein LoBind Eppendorf tubes (Eppendorf, Catalogue number 02243108, Hamburg, Germany) and spun at 1000g for 3 minutes. Replicates were combined prior to protein quantification using the method of Bradford in a microplate format (BioRad, Catalogue number 500‐0006, Hercules, CA). Bovine IgG (BioRad, Catalogue number 500‐0005) was used as a standard and the data collected by a BioRad Model 550 microplate reader at 595 nm controlled by BioRad Microplate Manager Software (version 5.1, build 75). After protein quantisation, samples were stored at −80°C.

Table 1.

Subject and wound profiles with sampling site t‐tests of protein spots after 2D‐PAGE

Subject	Days enrolled	% change in wound area *	Outcome	Stain	Collection times		I versus P t‐tests
					Internal (I)	Peripheral (P)	DF	Pr>ltl
BM021	42	−88%	Healed	CBB250	15	12	1537	<.0001
BM024	35	−100%	Healed	Sypro 40	12	12	950	0.2096
BM027	42	−100%	Healed	CBB250	11	8	1439	<.0001
BM034	35	−100%	Healed	Sypro 40	10	8	701	0.7272
BM036	42	−89%	Healed	CBB250	15	15	1586	<.0001
BM010	42	−84%	Healed	CBB150	8	14	1746	0.0019
BM038	35	−56%	Healing	Sypro 40	10	10	734	0.0079
BM002	42	−43%	Healing	CBB150	15	15	1363	<.0001
BM008	42	−37%	Not healed	CBB150	14	13	1503	0.1622
BM048	42	−36%	Not healed	Sypro 40	15	15	923	<.0001
BM029	28	−30%	Not healed	Sypro 40	11	11	1217	0.0001
BM015	42	−22%	Not healed	CBB250	15	15	736	0.9067
BM023	42	−19%	Not healed	Sypro 40	12	14	1236	0.112
BM016	35	−18%	Not healed	CBB250	15	14	1673	0.0001
BM017	42	−8%	Not healed	CBB250	12	12	1289	<.0001
BM043	42	43%	Not healed	Sypro 40	15	15	703	0.0401
BM044	42	63%	Not healed	Sypro 40	14	12	909	0.0001
BM001	42	73%	Not healed	Sypro 40	13	10	1717	0.1348
BM045	42	362%	Not healed	Sypro 40	14	13	1046	0.8358

DF, degrees of freedom.

^*The % change in wound area is the final wound area divided by the wound area on day 0, expressed as a percentage. Outcome was assigned based on the final wound area. Three protein concentrations and two stains were used to detect protein spots on 2D‐PAGE gels. Both the interior and periphery of wounds were sampled and the protein spots were compared for all days at each location using Student's t‐test.

IEF of wound fluid proteins

Protein samples were thawed to 4°C and thoroughly vortexed before removing an aliquot for IEF. One hundred and seventy microlitres of rehydration buffer, 8 M urea, 2% CHAPS, 0·002% bromophenol blue, 1 M dithiothreitol (DTT), and 0·5% GE Healthcare 3‐10NL ampholyte (GE Healthcare Biosciences, Piscataway, NJ) were added to each sample and the resulting solution vortexed before being pipetted into the lane of a 11 cm IEF tray (BioRad, Catalogue number 165‐4020). Eleven centimetre, pH 5–8, linear gradient IPG ReadyStrips (BioRad, Catalogue number 163‐2018) were placed on top of the sample solution. The tray was placed into a Protean IEF cell (BioRad, Catalogue number 165‐4001). The strips were rehydrated for 12 hours. Mineral oil (BioRad, Catalogue number 163‐2129) was pipetted over the strips after 1 hour of rehydration to prevent evaporative sample loss. After rehydration, focusing was via rapid gradient to a maximum voltage of 8000 V with a total focusing time of 55 000 V h and a 50 mA resistance limit per strip. When focusing was complete, strips were either immediately run in the second dimension or were sealed in a tray and stored at −80°C.

Second dimension SDS‐PAGE of wound fluid proteins

IPG strips were incubated in Equilibration Buffer I, 50 mM tris base, 6 M urea, 30% glycerol, 2% SDS, 0·002% bromophenol blue and 65 mM dithiothreitol, for 20 minutes followed by Equilibration Buffer II, 50 mM tris base, 6 M urea, 30% glycerol, 2% SDS, 0·002% bromophenol blue and 135 mM iodoacetamide, for 20 minutes. The IPG strip was placed into the well of a Criterion 10–20% Tris–HCl IPG+1 gel (BioRad, Catalogue number 345‐0107) and cemented into place with 0·5% molten agarose (GE Healthcare Biosciences, Catalogue number 17‐0554‐01) and 0·02% bromophenol blue in Tris‐Glycine‐SDS (sodium dodecyl sulphate) (TGS) running buffer (25 mM tris base, 192 mM glycine, 0·1% SDS). Twelve gels were run simultaneously (200 V) in a Criterion Dodeca cell (BioRad, Catalogue number 165‐4130) filled with chilled TGS running buffer. Blocks of frozen TGS were placed between gels and at either end of the tank to prevent overheating during the run. Precision Plus Dual Color Protein standards (BioRad, Catalogue number 161‐0374) were used to monitor the progress of the second dimension separation.

Gel staining, image acquisition, data storage

Gels were stained with either Bio‐Safe colloidal coomassie stain (BioRad, Catalogue number 161‐0787) or Sypro^® (BioRad) according to manufacturer's instructions. The staining method and protein load were tested at several levels to compromise sample consumption with protein detection. Gel images were acquired using a GelDoc XR Gel Imaging System (BioRad) and PD Quest 2D analysis software, version 7.4.0, build 036 (BioRad). Both PD Quest and Proteomeweaver 4.0 (BioRad) software packages were used to analyse gel images. Proteomeweaver software settings were default with the exception of the input image resolution (190 dpi) and remained consistent for all experiments.

After spots were recognised by Proteomeweaver, they were matched within location, internal or peripheral, for a wound before matching between locations. Spots that matched to at least one other 2D‐PAGE gel spot (global frequency = 2) were termed super‐spots and were used for a comparison of the internal and peripheral regions of individual wounds. The number of occurrences of a super‐spot within a wound area (I versus P) across days was tabulated. The super‐spot numbers within patients were tested for I versus P differences with paired t‐tests (Proc TTest, SAS Institute 2002).

For each 2D gel, the number of super‐spots identified by Proteomeweaver was compared for healed and non healed chronic wounds. To avoid pseudoreplication, data were averaged by sample. A generalised linear model (GLM) was used to test for differences in the total number of super‐spots (Proc GLM, SAS Institute 2002). The GLM included day, stain type, site within the wound (I versus P), wound outcome and all first‐level interactions among these effects. After the initial run of this model, outlier analysis was conducted, and any datum whose studentized residual was above X was removed. For the outlier analysis, the significance level was P > 0·01 and we imposed a maximum limit of 1% of the data to be deleted (4 points).

Protein depletion

Wound fluid samples were partitioned using either a ProteomeLab™ IgY‐12 High Capacity SC Spin Column kit (Beckman Coulter, Catologue number A24618, Fullerton, CA) or Multiple Affinity Removal Spin Cartridges (Agilent Technologies, Catalogue number 5188‐5230, Palo Alto, CA) according to manufacturer supplied protocols. Briefly, 8–20 µl of wound serum (33–52 µg/µl total protein) was diluted, 0·22 µm spin‐filtered and applied to the column. Unbound proteins were collected using centrifugation, and subsequently, bound proteins were stripped and eluted. Column fractions were concentrated (Amicon Microcon^® Ultracel YM‐3, Billerica, MA) and processed for 2D electrophoresis.

Protein identification

Protein spots were cut from 2D gels using sterile pipette tips and aseptic technique. Gel plugs were placed in micro‐centrifuge tubes and sent to either Tufts University, Boston, MA, or Midwest Bio Services LLC, Overland Park, KS, for protein identification. Results from additional service labs are not shown. For samples analysed at Tufts University, excised spots were subjected to in‐gel reduction, alkylation, and enzymatic digestion (Roche Applied Science, Indianapolis, IN) in a HEPA‐filtered hood to reduce keratin background. LC/MS/MS analysis was performed on the in‐gel digest extracts using Agilent (Santa Clara, CA) 1100 binary pump directly coupled to a mass spectrometer. A measured quantity of 2–8 µl sample was injected on column using a LC Packings (Sunnyvale, CA) FAMOS autosampler. Nanobore electro spray columns were constructed from 360 µm o.d., 75 µm i.d. fused silica capillary with the column tip tapered to a 10 µm opening (New Objective, Woburn, MA). The columns were packed with 200‐Å , 5‐µm C18 beads (Michrom BioResources, Auburn, CA), a reverse‐phase packing material, to a length of 10 cm. The flow through the column was split pre‐column to achieve a flow rate of 320 nl/min. The mobile phase used for gradient elution consisted of (A) 0·3% acetic acid 99·7% water and (B) 0·3% acetic acid 99·7% acetonitrile. Tandem mass spectra (LC/MS/MS) were acquired on a Thermo LTQ ion trap mass spectrometer (Thermo Corp., San Jose, CA). Needle voltage was set to 3 kV, isolation width was 3 Da, relative collision energy was 30% and dynamic exclusion was used to exclude recurring ions. Ion signals above a predetermined threshold automatically triggered the instrument to switch from MS to MS/MS mode for generating fragmentation spectra. The MS/MS spectra were searched against the human subset of the whole NCBI non redundant protein sequence database using the SEQUEST computer algorithm (4) to produce a list of proteins identified in each sample. Each protein confirmed required a minimum of four matching peptides. The total ion current (TIC), sum of individual ion currents for all confident peptide matches of a protein identification, was divided by the TIC for the entire LC/MS/MS run and expressed as %TIC. The %TIC was used as a coarse estimate of protein abundance.

RESULTS

Protein depletion

Wound fluid samples subjected to protein depletion columns for the enrichment of low abundance proteins did not appear to have a consistently better or different distribution of protein spots than the native sample. For some samples, the spot pattern was similar for both the high‐abundant eluent and the low‐abundant flow‐through fractions (Figure 1B–D). In other trials, the number of spots for the low‐ and high‐abundant fractions combined was significantly reduced from the native sample indicating protein loss (Figure 2).

Figure 1.

Wound fluid separated by two‐dimensional electrophoresis before (A and B) and after (C and D) prefractionation to remove high‐abundant proteins. Coomassie blue was used to detect proteins if a minimum of 150 mg sample was separated (A). Sypro^® was used to detect protein spots when 40–60 mg of sample protein was separated (B–D). The low‐abundant enriched sample (C) and high‐abundant flow‐through proteins (D) for a fractionated sample showing negligible differences.

Figure 2.

Partitioning of wound fluid samples before electrophoresis to remove high‐abundant proteins with concomitant protein loss. The original sample (A), column flow‐through (B), and eluent (C) are all stained with Sypro^®.

Gel analysis

Proteomeweaver settings were tested by comparing the number of spots recognised in a gel after digital imaging by the software and comparing it with the number that could actually be cut by hand. Figure 3 includes a gel before and after spot cutting. Proteomeweaver recognised 490 spots using experimental settings while 485 were actually cut. Because both the total numbers of spots and the intensity of an individual spot are dependent on protein load and detection agent, only equivalent gels were used for comparisons (Figure 1A, B). Of the 46 ulcers enrolled in the study, complete temporal comparisons were limited to 19 wounds (Table 1). Proteomeweaver identified 271–772 spots per gel (n = 517) for coomassie stained gels and 226–733 spots per gel (n = 407) for Sypro^® stained gels. For each wound, all swabs were grouped by location, internal and peripheral. Gel spots were matched within group before between groups by Proteomeweaver.

Figure 3.

Side‐by‐side comparison of a 2D‐PAGE gel before and after spot cutting. Proteomeweaver and PD Quest software (BioRad) were used to identify protein spots from digital images of gels. Proteomeweaver identified 490 spots on this gel, and when cut by hand, 485 could be isolated and removed from the polyacrylamide matrix.

Table 2 shows the results of the GLM analysis of total spot numbers. The effect of day was statistically significant and showed an interaction with wound healing (Figure 4). Healed wounds showed an increasing number of spots coincident with wound closure, while unhealed wounds showed no temporal trend. There was also a statistically significant interaction between stain and location. Table 1 shows the results of paired t‐tests between wound sites for super‐spot abundance. The internal and peripheral sites were significantly different for 12 of the 19 wounds.

Table 2.

2D‐PAGE protein spot number analysis for chronic wound fluid ^†

Source	DF	Sum of squares	Mean square	F value	Pr > F
Model	14	218950.726	15639.338	4.67	<.0001
Error	415	1388902.347	3346.753
Corrected
Total	429	1607853.073
	R‐square	Coefficient of variation	Root MSE	Mean
	0.136176	13.62153	57.85112	424.7035
Source	DF	Type III SS	Mean square	F value	Pr > F
Day	1	21877.3175	21877.3175	6.54	0.0109
IP	1	87.55122	87.55122	0.03	0.8716
Heal	1	461.70374	461.70374	0.14	0.7105
Stain	2	7298.59039	3649.2952	1.09	0.337
Day*IP	1	4781.07036	4781.07036	1.43	0.2327
Day*heal	1	28777.63228	28777.63228	8.6	0.0036
Day*stain	2	9268.80862	4634.40431	1.38	0.2515
Heal*IP	1	1187.29279	1187.29279	0.35	0.5518
Stain*IP	2	48612.87503	24306.43751	7.26	0.0008
Heal*stain	2	5716.86459	2858.43229	0.85	0.4264

^†Outlier data removed (4 points) before a generalised linear model was used to test for differences in the total number of protein spots between wounds that healed and those that remained unhealed during the 42‐day observation period.

Figure 4.

Wound fluid from chronic pressure ulcers was separated by 2D‐PAGE. The number of protein super‐spots resulting from samples isolated from the interior and periphery of the wounds was combined for each sampling time point and compared between wounds that healed during the 42‐day observation period and those that remained unhealed. The effect of day was statistically significant and showed an interaction with wound outcome.

Mass spectrometry

Excised spots were sent for protein identification on nine occasions. Five different service labs in both the public and private sectors were used. Technical issues prevented results for three trials and two more trials were performed to alleviate identification issues originating in sample preparation. Protein identifications from spots chosen based on analysis of gels using PD Quest are shown in Table 3 with their corresponding location shown in Figure 5. Contaminating keratin proteins including epidermal cytokeratins were removed pre‐search and post‐search and are not included in Table 3. For spot 6, only keratins were identified. Spot 14 contained S100A9 protein, a novel but potential biomarker of wound healing. Proteomeweaver matched this spot in 21 of 28 gels from the 14 days that this wound was included in the study. Day 10 samples did not adapt to the 2D‐PAGE technique well and all three replicate gels were unmatched for the S100A9 super‐spot, accounting for nearly half of the total unmatched gels for this wound. The intensity of this spot during sampling is shown in Figure 6. The S100A9 spot was matched in four different subjects including both peripheral and internal swabs by Proteomeweaver. Spots were cut from five gels and sent for protein identification. Only one spot returned a S100A9 positive identification (results not shown).

Table 3.

Mass spectrometry of proteins in pressure ulcer wound fluid after 2D‐PAGE*

Spot no.	MW (kDa)	Name	% TIC	Peptide matches	GenInfo identifier (gi)
1	61.3	Albumin	22.45	21	78101701
		Haemopexin	2.4	7	11321561
		Ig heavy chain c region gamma 1	0.93	4	12054072
2	56.9	Albumin	13.47	23	78101701
		Ig alpha‐1 chain c region	1.05	5	113584
		Serpin peptidase inhibitor, clade A	0.69	4	15080499
		Haptoglobin	0.44	4	1212947
		Ig heavy chain c region gamma 1	0.52	4	12054072
		Mutant beta globulin	0.45	4	18418633
3	60.8	Albumin	10.33	21	21706456
		Ig alpha‐1 chain c region	0.64	4	113584
4	56.6	Albumin	8.49	27	78101701
		Ig alpha‐1 chain c region	1.28	4	113584
5	60.5	Albumin	15.33	22	21706456
		Haemopexin	1.38	5	11321561
		Ig alpha‐1 chain c region	0.8	4	113584
6	55.9	Keratin	na	na	na
7	60.4	Albumin	10.65	18	21706456
		Haemopexin	1.74	5	11321561
		Serpin peptidase inhibitor, clade A	0.65	4	15080499
		Ig alpha‐1 chain c region	0.89	4	113584
8	55.4	Albumin	9.8	19	78101701
		Ig alpha‐1 chain c region	3.38	5	113584
		Haemopexin	0.78	4	11321561
		Serpin peptidase inhibitor, clade A	0.87	4	15080499
		Haptoglobin	0.52	4	1212947
9	17.2	hp2‐alpha	17.51	13	296653
		Albumin	2.12	17	21706456
		Ig kappa light chain vlj region	1.02	6	21669463
		Chain B, Apo‐human serum Transferrin (glycosylated)	0.8	13	110590600
		cn/Zn‐seperoxide dismutase	0.32	4	1237406
10	16.7	Albumin	8.94	25	23307793
		Ig lambda chain V region	0.73	5	478602
		Haptoglobin alpha (2FS)‐beta precursor	0.63	4	459813
		anti‐TNF antibody light chain Fab	1.17	5	11275302
		Unnamed protein product sickle beta haemoglobin	0.18	5	29446
		Purine nucleoside phosphorylase (Pnp)	0.15	6	4557801
11	15.8	Albumin	4.99	28	4389275
		Anti‐HBsAg Ig Fab kappa chain	4.33	10	3721651
		Chain L, crystal structure of human factor Ix Glas domain	0.65	5	42543068
		Haemoglobin deoxy form	0.17	4	109157483
		Chain A, fourth Ch domain from human L‐plastin	0.14	4	109157450
12	14.1	Ig lambda heavy chain	1.63	7	2765425
		Transferrin variant	5.33	13	62897069
		Albumin	1.62	14	4389275
		Haemoglobin deoxy form	0.37	7	109157483
		Anti TNF‐alpha antibody light chain Fab	0.5	6	11275302
13	38	Haptoglobin	6.52	10	1212947
		Albumin	1.39	18	31615331
		Zn‐alpha2‐glycoprotein	0.36	4	38026
14	13.3	S100 A9	2.22	4	4506773
		Albumin	3.9	18	78101701
		hp2‐alpha	2.77	6	296653
		Anti TNF‐alpha antibody light chain Fab	2.26	8	11275302
		HSP 70 kDa protein 8	1.94	4	16741727
		Haemoglobin deoxy form	0.26	4	109157483
15	11	Albumin	4.04	17	78101701
		Haemoglobin chain D deoxy	1.15	4	1942684

^*Mass spectrometry results are the result of a search of the human subset of the entire NCBI non redundant database by SEQUEST software. Keratin and cytokeratin contaminants are not reported except in the case of spot no. 6 where only keratin proteins were identified. The %TIC is the percentage for a particular ion signal relative to the total ion signal for the full LC/MS/MS analysis and is given as a coarse estimate of protein abundance. Spot numbers correspond to Figure 5.

Figure 5.

Wound fluid was collected from the interior portion of a chronic pressure ulcer that existed for more than 6 months prior to study inclusion and eventual closure. The proteins present were separated by 2D‐PAGE using a pH 5–8 linear gradient and a 10–20% polyacrylamide gel with a separation range of 10–200 kDa. Fifteen spots identified by PD Quest software (BioRad) were excised. Protein identifications from mass spectrometry analysis are given in Table 3 for the locations indicated.

Figure 6.

Protein S100A9 was identified from the interior wound fluid of a chronic pressure ulcer from day 1 of sampling. Proteomeweaver matched this spot, shown in red outline, in 21 of 28 gels from the 14 days that this wound was included in the study. The intensity changes between gel replicates are shown below the average intensity and coefficient of variation for each day.

DISCUSSION

Proteomic assays are reliant on the detection of target molecules in the presence of a molar excess of other proteins. In blood plasma, albumin, IgG, alpha‐1 antitrypsin, IgA, transferrin and haptoglobin are responsible for approximately 85% of the total protein concentration. Depletion columns are designed to remove highly abundant proteins from a sample and therefore enrich low‐abundant proteins. While simple in principle, techniques for the optimisation and prefractionation of samples have been published ^{5, 6}. The Agilent Multiple Affinity Removal System depletes samples of 6 proteins (albumin, IgG, antitrypsin, IgA, transferrin and haptoglobin) while ProteomeLab IgY‐12 Proteome Partitioning Kit removes 12 proteins (albumin, IgG, fibrinogen, transferrin, IgA, IgM, haptoglobin, alpha2‐macroglubulin, alpha1‐acid glycoprotein, alpha1‐antitrypsin, apolipoprotein A‐I and A‐II).

Because the partitioned samples are reduced in protein concentration compared with the native sample, they must be concentrated and/or pooled (low‐abundant fraction) before 2D electrophoresis. Sypro^® was used to detect spot patterns after fractionation. Sypro^® is advantageous for its extended dynamic range, but spots are more tedious to excise, and because the protein concentration is less than that found in coomassie gel spots the identification of proteins by LC MS/MS is generally more difficult. Additionally, depletion removes small proteins that are bound to albumin and immunoglobulins and as a result, loss of biomarkers may result unless an evaluation of coeluted proteins is performed in parallel (7).

In samples collected from the wounds in the current study, the use of protein depletion columns yielded inconsistent results. Thus, because of the high number of samples and their probable protein loss, dilution and degradation during extra processing steps, prefractionation of samples was not performed. With optimisation, commercially available kits may prove useful to enrich low‐abundant proteins from exudate of different wound types, such as venous ulcers, where sample volume is not limited.

After removing the 20 most abundant proteins in the fluid from diabetic foot ulcers, Chojnacki et al. used multidimensional protein identification technology (MudPIT) to identify a variety of growth factors and inflammatory proteins including epidermal growth factor (EGF), platelet‐derived growth factor subunit B homodimer (PDGF‐BB), keratinocyte growth factor‐2 (KGF‐2) and transforming growth factor beta (TGF‐β) as well as calgranulin A, B and C (8).

Fernandez et al. used 2D‐PAGE and mass spectrometry to evaluate wound fluid from chronic venous ulcers. Samples were pooled for analysis and high‐abundant proteins were removed (9). Fibrinolytic and coagulation proteins were identified. In wound fluid collected for the current study, we did not note significant amounts of native high‐abundant serum proteins in 2D gels as is seen in 2D gels prepared from unfractionated serum samples. However, the difference between the aetiology of venous leg ulcers and pressure ulcers must be noted. Additionally, our LCMS/MS results indicate that albumen fragments were a common component in many spots (Table 3), but partially degraded.

The site complexity of samples collected from chronic wounds makes the analysis of 2D gels more challenging than samples collected from more privileged or isolated fluid cavities such as blood or spinal fluid. The proteolytic nature of the exudates promotes variable degrees of protein fragmentation independent of the technique, further complicating the spot patterns. Wound samples had shifting protein concentrations, sometimes very low, necessitating SYPRO stain and lower protein loading levels to avoid sample exhaustion. The unpredictability of changes in protein concentration over time for the same wound and the significant effect of stain and protein load on spot detection eliminated many wounds from a full temporal analysis of spot intensity changes.

Analysis of gels is based on the presence of the spot, as well as the density of the spot. Gel to gel variability dictates multiple runs of each sample for replicate averaging. Gels can be difficult to compare when run at different times, even when run under the same conditions (pH, time, voltage, etc.) because of slight but inevitable positional shifts and intensity changes of individual spots. Wound fluid is limited in amount, and multiple gel replicates eventually consume the entire sample leaving no options for additional quantitative or confirmatory proteomic techniques. Depending on the tissue present in the wound bed (i.e. eschar versus granulation), the amount of exudate from the wound can vary widely. Additionally, the type of wound can impact the amount of exudate, but even highly exudative wounds may have limited protein present. In the current study, as wounds healed and the area of the wound decreased the amount of exudate decreased, but no correlation between the amount of protein and the amount of exudate was noted. A compromise between replicate averaging and sample preservation is paramount for follow‐up quantisation of proteins in the native sample (10).

Sophisticated software is required to detect spots, match spots between gels, and finally determine density differences for individual spots on multiple gels. Identifying spots of interest based on expression differences is necessarily focused on a minimum twofold change as a result of high gel to gel variation (11). In ten replicate gels of human lumbar cerebrospinal fluid, the coefficient of variation (CV) for the volume of individual spots ranged from 4% to 23% while the percentage of matched spots between replicate gels ranged from 88% to 100% (12). The percentage of matched spots decreased while the CV increased for samples that were processed at different times. This source of error is unavoidable in our study because of the sheer number of samples (1101) collected and analysed without pooling. Because protein stains bind proteins in a dye‐specific manner, spot density can be compared only between gels that had equal protein loads and consistent post‐electrophoretic staining protocols. Spot matching and quantisation including subsequent protein identification is a common bottleneck for proteomic analysis using 2D‐PAGE (10). Some of the disadvantages of 2D‐PAGE have been overcome by 2D fluorescence difference gel electrophoresis (2D DIGE).

2D DIGE is a technique used to simultaneously compare different samples on the same gel by labelling the samples with a fluorescent dye prior to 2D‐PAGE. Up to three different samples can be separated with the same gel, two samples and a standard. The three samples on the gel are clearly discernable by visualisation of the different labelling dyes, allowing differences in spot intensity to be more easily evaluated. Pollins et al. evaluated protein extracts from normal and burned skin and identified temporal patterns of expression of cytoskeletal and heatshock proteins using 2D‐DIGE (13). While this technique is valuable for a small number of individual or pooled samples, the advantages are lost when a large number of gel comparisons are required, as in our study.

The identification of proteins by LC MS/MS from excised gel plugs has been used successfully to characterise proteins and identify potential biomarkers. It is precise but costly if many samples are analysed. Even after protein fractionation by both molecular weight and isoelectric point, each spot usually contains multiple proteins. Proteins of interest identified from 2D gels need to be quantified to determine if the intensity changes of a matched spot are due to the protein of interest or any other protein component of the spot. Further complicating this is the consideration that the composition of the spot can change over the time course of the experiment. While this technique can be simplified by pooling samples and analysing a single treatment versus control, changes in spot intensity and protein identification must be confirmed with another technique such as western blots or enzyme‐linked immunosorbent assay ELISA.

Tarran et al. used 2D‐PAGE and matrix assisted laser desorption/ionisation time‐of‐flight (MALDI‐TOF) mass spectrometry to study wounds in a rat model at different time points. Tissue from the wounds was collected, frozen and ground into a powder prior to analysis. Twenty‐six distinct proteins were identified, including the major spots in the samples. The level of hemoglobin changed significantly over time and was elevated in wound samples at 5 minutes to 3 hours after the wound was created, but was near control levels by 12 hours (14). Oh et al. used 2D gels and MALDI‐TOF to evaluate the collagen and collagen‐related proteins present in a line of human skin fibroblasts. Eighty proteins were identified on the 2D gels, including collagen alpha 1 (I) chain precursor, collagen alpha 1 (III) chain precursor and other collagen‐related structures, as well as metabolic enzymes (15). 2D‐PAGE was used to study the differential expression of epidermal proteins during epithelialisation in human skin equivalents and one protein RS/DJ‐1 was identified (16).

The blister fluid of 11 human subjects was examined by Volden et al. Two subjects were subjected to ultraviolet light to create erythema, and then suction blisters were created on those sites. Blister fluid from individuals without erythema was also obtained. Blister fluid was collected using a syringe and compared with serum samples collected. Several hundred spots were present in the 2D gels, and protein patterns that were obtained using 2D‐PAGE were similar between blister fluid and serum from the same individual. Blister fluid had haemoglobin present, and increased alpha‐antitrypsin dimmer and actin, as well as two unknown spots compared with the serum. The blister fluid haemoglobin was thought to be a product of the syringe collection technique. One hundred low molecular‐weight compounds were identified in this study using gas chromatography and mass spectrometry (17).

Macdonald et al. evaluated blister fluid from normal skin and nonlesional skin from an individual with plaque psoriasis before and after IL‐1β intradermal application using 2D‐PAGE and MALDI‐TOF. Six hundred and seventy protein spots were present in the samples and compared, and nine proteins were identified. Vitamin D binding protein was present in the greatest amounts in the psoriasis subject both with and without IL‐1β treatment. Haptoglobin was absent in the IL‐1β‐treated psoriasis subject sample, and present in low values in the fluid from the psoriasis subject without treatment. Haptoglobin was present with the greatest density in both normal samples (18). Although some of the same proteins were identified in our study (albumin, haemopexin, transferrin, immunoglobulins, glycoproteins and haptoglobin) as in these two studies using blister fluid, spot locations were not directly comparable because of the differences in pH gradients and the size of the gels used in each study.

In the absence of protein identifications from 2D‐PAGE, significant differences in the spot distribution between the interior and periphery of a wound were found in 12 of 19 chronic wounds. However, it is likely that wound fluid sample collection limitations affected these results. Eleven of the 12 subjects showing significant differences had equal numbers of collection dates between internal and peripheral sites. Seven of eight subjects with significant differences between the internal and peripheral locations of a wound had high numbers of collection days. In the seven wounds that did not show significance, the number of collection dates was either few or not well matched between the interior and periphery, with the exception of a single wound (Table 1). These data support our hypothesis that local biochemistry differs between the interior and periphery of a wound. The biochemistry reflects cellular processes and is an indication that different events occur at the periphery in comparison to the interior of a chronic wound. The temporal dependence of spot number and wound outcome was significantly different for healed and unhealed chronic wounds (Figure 4). Again, this supports a different biochemistry that parallels patient outcome.

2D electrophoresis is intrinsically hampered by poor gel to gel reproducibility, limited dynamic range of detection by protein binding dyes, loss of hydrophobic and low‐abundant proteins, multiplicity of proteins in spots, high sample demand and finally, it is expensive and labour intensive (19). The S100A9 spot matched between wounds by Proteomeweaver resulted in a positive identification in only one out of five spots. It cannot be determined if the software matched incorrectly or if the spot's composition varied between subjects and/or days or even both without cutting the spot and analysing it for each day and subject in the study.

Our sample source is a chronic wound with variable degrees of inflammation caused by the presence of both environmental pathogens and cellular debris. Wound healing is multi‐hierarchal and involves a transition from tissue degradation and clearing to synthesis (20). Even for a single protein, in order to understand the balance between build and destroy, the measurement of degrading proteins such as matrix metallic proteinases must be coupled with their inhibitors. Therefore, it seems unlikely that a single protein will be useful as a biomarker, but rather a set of proteins that represent multiple stages of healing.

Traditionally, the standard work flow for proteomic analysis uses 2D gel electrophoresis to separate proteins and mass spectrometry to identify proteins in excised spots. Although this 2D proteomic approach has proven successful for biomarker discovery in other systems, because of the high numbers of spots present in gels, as well as the complexity of spots, high cost of protein identification from gel plugs and sample consumption, further biomarker identification via 2D technology was abandoned in favour of other techniques including iTRAQ and label‐based microarrays that simultaneously measure a panel of proteins for a more complete assessment of the proteome.

ACKNOWLEDGEMENTS

This work was supported by the Telemedicine and Advanced Technology Research Center (TATRC) at the US Army Medical Research and Materiel Command (USAMRMC) through award W81XWH‐05‐1‐0401.

We thank Catholic Health System Partners in Rehab, ElderWood Health Care, and Kaleida Health for allowing us to enroll subjects from their facilities. We thank Jon DeGnore, Tufts University, for mass spectrometry discussions and manuscript review. Charles Anderson and Tera Kane are credited with the development and completion of a manual 2D‐PAGE spot analysis method used for preliminary results and comparison with the digital analysis. We thank Gary Kohl, Instrumentation Specialist, Bio‐Rad Laboratories, Life Science Group Hercules, CA, for use and evaluation of Proteomeweaver 4.0 software.