Deciphering the proteome of the in vivo diagnostic reagent “purified protein derivative” from Mycobacterium tuberculosis

Abstract

Purified protein derivative (PPD) has served as a safe and effective diagnostic reagent for 60 years and is the only broadly available material to diagnose latent tuberculosis infections. This reagent is also used as a standard control for a number of in vitro immunological assays. Nevertheless, the molecular composition and specific products that contribute to the extraordinary immunological reactivity of PPD are poorly defined. Here, a proteomic approach was applied to elucidate the gene products in the U.S. FDA standard PPD-S2. Many known M. tuberculosis T cell antigens were detected. Of significance, four heat shock proteins (GroES, GroEL2, HspX, and DnaK) dominated the composition of PPD. The chaperone activities and capacity of these proteins to influence immunological responses may explain the exquisite solubility and immunological potency of PPD. Spectral counting analysis of three separate PPD reagents revealed significant quantitative variances. Gross delayed type hypersensitivity (DTH) responses in M. tuberculosis infected guinea pigs were comparable among these PPD preparations; however, detailed histopathology of the DTH lesions exposed unique differences, which may be explained by the variability observed in the presence and abundance of early secretory system (esx) proteins. Variability in PPD reagents may explain differences in DTH responses reported among populations.

Introduction

Tuberculosis is a disease in humans and other vertebrates that progresses through complex interactions between the host’s immune system and the pathogenic bacterium Mycobacterium tuberculosis. In particular, the induction of a prominent inflammatory response as well as a strong adaptive immune response by M. tuberculosis, are hallmarks of this disease. Early after the discovery of M. tuberculosis as the causative agent of tuberculosis, Robert Koch introduced tuberculin as a potential vaccine against tuberculosis [1]. This material was a crude mixture of macromolecules produced by M. tuberculosis and released into the culture medium upon heat inactivation of the bacterium. Tuberculin failed as a vaccine because it elicited a damaging immune response known as the “Koch phenomenon” in individuals already infected with M. tuberculosis. However, this same immune mediated event was later recognized as having diagnostic value [2] based on observations that tuberculin produced localized edema within 1 to 2 days of its injection into the skin of animals infected with M. tuberculosis. This classical immunological reaction now known as delayed-type hypersensitivity (DTH) or Type IV hypersensitivity and has served for nearly a century as the primary basis for diagnosing (via the cellular immune response) exposure to M. tuberculosis.

The seminal work of Florence Seibert revealed that the tuberculosis specific DTH response induced by tuberculin was attributable to the “tuberculoprotein” [3]. Her work also led to refined production techniques that minimized carbohydrate and nucleic acid contamination and inconsistencies in the reactivity of tuberculin preparations [4]. The material produced in 1941 by Seibert established modern day purified protein derivative (PPD) and the first international standard, PPD-S1. From this, additional standards were produced to include RT23 by the Statens Serum Institute [5, 6]) and the replenishment of the PPD-S1 with PPD-S2 [7]. While each lot of PPD meets specific standards for biological activity [7], early studies noted significant differences in potency among standard PPD preparations [5]. Chemical analyses of standard PPD preparations also demonstrate significant compositional differences. Specifically, PPD-S2 is comprised of 93% protein, 1% nucleic acid, and 6% carbohydrate. In comparison PPD RT-23 from the Statens Serum Institut was 80% protein, 20% nucleic acid, and <1% carbohydrate (unpublished data M.J. Brennan).

Identification of a select number of individual proteins or protein fragments present in PPD has been reported [8-11]. However, a comprehensive identification of all proteins represented in human PPD does not exist. This is in sharp contrast to the evaluation of M. tuberculosis where extensive proteomic studies have been performed to characterize the protein profile of this bacterium with respect to subcellular location [12-15], differences among strains [16-21], alterations associated with growth conditions [22-29], and elucidation of antigens for the cellular and humoral immune responses [30-34]. A complete understanding of the molecular composition of PPD would not only allow for the construction of a more refined skin test antigen, but it may also provide insight as to why PPD elicits such a vigorous immune response. A major impediment to the molecular definition of PPD is the fact this material is a mixture of denatured, but soluble proteins and peptides that are generated through autoclaving in vitro grown M. tuberculosis at 100°C for two hours. Thus, individual proteins of PPD cannot be resolved by SDS-PAGE or 2D-PAGE. Kuwabara reported the purification of a dominant 9.7 kDa protein of PPD that exhibited strong DTH activity and further provided the sequence of this protein by N-terminal sequencing of selected peptides [10, 11] However, these reported sequences are now known not to represent any mycobacterial protein [35] nor be encoded by the M. tuberculosis genome [36]. This example underscores the complexity of PPD and the technical difficulties that have been associated with defining its composition.

Modern multi-dimensional chromatography techniques [37] coupled to soft ionization mass spectrometry technologies [38] and powerful computer algorithms to search tandem mass spectrometry (MS/MS) data against protein databases [39, 40] has allowed for the facile identification of proteins within complex mixtures such as tissue lysates or subcellular fractions. Quantitative approaches such as spectral counting [41] and isotope tagging [42] have also allowed for the relative quantities of individual proteins to be determined within and between complex protein preparations. We have now applied these approaches to decipher the protein composition of the standard PPD preparation, PPD-S2, and found that it possessed at least 240 protein representatives, with four proteins (GroES, GroEL2, HspX, and DnaK) dominating this composition. Our studies also reveal that spectral counting analysis is capable of distinguishing significant variability in the composition of standard PPD preparations, and that these variations may impact the biological activity.

Materials and Methods

Reduction, alkylation, and trypsin and chymotrypsin digestion of PPD

Standard PPD-S2 was provided from the United States, Food and Drug Administration (FDA), PPD-RT23, was provided by the Statens Serum Institut (SSI), and PPD-KIT, was provided by the Korea Institute of Tuberculosis (KIT). For comprehensive analysis of PPD, PPD-S2 was reduced with dithiothreitol (DTT) (Sigma), and alkylated with iodoacetamide (Sigma) by standard methods [43]. Briefly, aliquots (1 mg) of lyophilized PPD-S2 were dissolved in 1 ml of 6 M guanidium hydrochloride (Pierce Chemical Co), 0.6 M Tris-HCl, pH 8.6, to which 1 μl of 4 M DTT was added and incubated for 3 h at 23 °C. Reduced PPD was alkylated by addition of 100 μl iodoacetamide (160 mM) in HPLC grade water and incubated in the dark for 30 min at 37 °C. The reduced and alkylated (RA) PPD was desalted by microdialysis using 50 mM ammonium bicarbonate (pH 7.8) for 48 h. The samples were dried under a vacuum and suspended in 10% acetonitrile, 0.2 M ammonium bicarbonate. Trypsin or chymotrypsin were added to individual samples at 1:50 enzyme:substrate ratio, and the samples incubated at 37 °C and 23 °C for 18 h, respectively. This resulted in reduced, alkylated and trypsin digested (RAT); and reduced, alkylated and chymotrypsin digested (RAC) PPD. Similarly, three aliquots (100 μg each) of PPD-RT23, PPD-KIT, and PPD-S2 were reduced, denatured, and digested with trypsin as above for quantitative analysis by spectral counting.

SDS-PAGE and Western blot

Aliquots (10 μg) of PPD, RA PPD, and RAT- and RAC-PPD were resolved by electrophoresis on 4-12% Bis-Tris polyacrylamide gels (Invitrogen, Carlsbad CA) using denaturing conditions and 2-(N-morpholino) ethanesulfonic acid (MES) running buffer per manufacturer’s instructions. Proteins were visualized by staining with silver nitrate. For Western blot analyses, aliquots (10 μg) of PPD-S2 and a preparation of Mycobacterium tuberculosis cytosolic protein [44] were resolved by SDS-polyacrylamide gel electrophoresis (PAGE) as above, electroblotted to nitrocellulose, and probed with M. tuberculosis specific monoclonal antibodies SA-12 (anti-GroES), CS-49 (anti-HspX), IT-56 (anti-GroEL2), and IT-41 (anti-DnaK). Antigen detection was visualized with 5-bromo-4-chloro-3-indolyl phosphate after incubating blots with anti-mouse IgG conjugated to alkaline phosphatase.

In-gel digestion

An aliquot (100 μg) of RAC-PPD was resolved by SDS-PAGE, stained by Coomassie blue R-250, and destained by 50% methanol. Five protein bands, 10, 14, 15, 25, and 50 kDa, were excised and digested in-gel with trypsin. The resulting peptides were extracted per standard methods [45].

Two dimensional liquid chromatography (2D-LC)

RAT- and RAC-PPD were fractionated by strong cation exchange (SCX) chromatography using a polysulfoethyl A^TM column (The Nest Group) and a Waters Alliance HPLC system. Peptides were eluted by stepwise fractionation at 0, 0.025, 0.05, 0.1, 0.2, and 0.5 M KCl at a flow rate of 1 ml min⁻¹. SCX fractions (six per sample) were dried under vacuum to 100 μl and further separated by reversed phase (RP)-HPLC using a Vydac C18 column (The Nest Group) and a linear gradient of acetonitrile (0-80%) in 0.1% trifluoroacetic acid. Fractions were collected manually based on peak intensity monitored at A₂₁₄. The number of individual fractions collected varied per run, and ranged from 16 to 40 per injection.

Protein Identification in PPD S2 by Mass spectrometry (MS)

Comprehensive analysis of PPD S2 to identify its molecular content was performed by taking an aliquot of each 2D LC fraction and applying it to a C18 capillary RP-HPLC column in-line with a ThermoFinnigan LCQ electrospray ionization (ESI) mass spectrometer (Thermo Electron) as described previously [30]. Briefly, an aliquot (10 μl) of the peptide mixture was injected onto a capillary (0.2 × 50 mm) C₁₈ RP column Microchom BioResources, Auburn, CA) and eluted with an increasing gradient (5-70%) of acetonitrile in 0.1% acetic acid using an Eldex MicroPro capillary HPLC system (Napa, CA) with a flow rate of 5 μl/min. The RP eluent was introduced directly into ThermoFinnigan LCQ ESI mass spectrometer operated using Xcalibur software version 1.3, and the peptides were analyzed by MS/MS. The electrospray needle was set at 4 kV with a N₂ sheath gas flow of 40 and a capillary temperature of 200 °C. MS/MS was automatically performed on the most dominant ion of the previous scan, and the normalized collision energy was set at 40%.

Database Searching

MS/MS data were extracted, deconvoluted, and deisotoped by BioWorks version 3.3. The resulting .dta files were searched against the M. tuberculosis protein database (R6, 3993 entries; Genbank accession #AL123456) using Sequest (ThermoFinnigan, San Jose, CA; version 27, rev. 12), and Mascot (Matrix Science, London, UK; version: 2.1 Mascot) programs. Search parameters included fragment ion mass tolerance of 1.00 Da, parent ion tolerance of 2.5 Da, and variable modifications of oxidation of methionine, and iodoacetamide or iodoacetic derivatives of cysteine for all samples, and acrylamide modification of cysteine for in-gel digested protein bands.

Scaffold (version Scaffold-3_00_01, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they achieved the Mascot program thresholds of 20, 30, and 40 or greater ion score for +1, +2, and +3 charged peptides, respectively, or achieved the Sequest program thresholds of a dCN of 0.1 or greater and 1.5, 2.5, and 3.0 or greater for +1, +2, and +3 charged peptides, respectively. Proteins represented by one peptide were validated by manual interrogation of the MS/MS data. Criteria for validation included: 1) there must be a minimum of 80% coverage of theoretical y or b ions (at least 5 in consecutive order); 2) there must be an absence of prominent unassigned peaks greater than 5% of the maximum intensity; and 3) indicative residue specific fragmentation, such as intense ions N-terminal to proline and immediately C-terminal to aspartate and glutamate, were used as additional parameters of confirmation. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.

The false discovery rate (FDR) was determined for the comprehensive PPD-S2 data set by searching the 357 individual .raw files against a decoy database that contained a reversed version of the proteins included in the customized database composed from the Mycobacterium tuberculosis H37Rv FASTA file (7982 entries). The resulting files were pooled in Scaffold and the analysis was identical to that described above. A total of 9 proteins and 12 spectra matched to the reverse database, resulting in a protein FDR of 3.9% and peptide FDR of 0.3%. The final data set was defined by 240 unique proteins represented by 3,873 spectra in PPD-S2. All Mascot generic files (MGF) containing MS/MS spectra are available on the PRoteomics IDEntifications database [46] (PRIDE; http://www.ebi.ac.uk/pride/, Accession numbers 19499 and 19500) for review. The scaffold file, containing all compiled, compressed PPD-S2 data, is available at http://mrl.colostate.edu/mtb_db/pages/ppd.aspx.

For determination of the 10 most dominant protein representatives in PPD-RAC and PPD-RAT, normalized spectral abundance factors (NSAFs) were determined [47]. Quantitative values were then assigned for each RAC-PPD and RAT-PPD identified protein so that the two preparations could be directly compared [48].

Quantitative comparison of separate PPDs by mass spectrometry

Individual digests of PPD-S2, PPD-RT23, and PPD-KIT were analysed by LC-MS using a nanospray LC-LTQ as described previously [49]. Each sample was injected in triplicate, resulting in each PPD being represented by 9 individual data sets.

Spectrum counting was used to compare PPD-S2, PPD-RT23, and PPD-KIT relative protein abundances. Each sample was represented by a total of nine injections (3 biological replicates of each injected 3 times each). Each injection was interrogated individually using the Mascot and SEQUEST algorithms as described above. The data were pooled such that each biological replicate included the pooled spectral data from the 3 injections. The resulting spectra were qualified and validated as described above. All Mascot generic files (MGF) containing MS/MS spectra are available on the PRoteomics IDEntifications database [46] (PRIDE; http://www.ebi.ac.uk/pride/, Accession numbers 19499 and 19500) for review. The scaffold file, containing all compiled, compressed PPD spectral counting data, is available at http://mrl.colostate.edu/mtb_db/pages/ppd.aspx.

Total spectral counts from the validated dataset were exported for statistical analysis using Dante (version 1.2, Battelle Memorial Institute). Briefly, individual data sets for each injection were quantile normalized and data compressed for each biological replicate. Replicates were analyzed against each PPD to generate hierarchical clustering analysis (HCA) heat maps. Pair-wise anova analysis for each protein was performed to determine statistically significant changes between samples (p <0.05).

DTH analysis

Hartley guinea pigs (Charles River Laboratories, North Wilmington, MA) weighing ∼450 g were infected with approximately 10-30 colony forming units (CFU) of M. tuberculosis H37Rv (TMCC# 102) via aerosol delivery using a Madison chamber (University of Wisconsin, Madison, WI). At 10 weeks post-infection, guinea pigs were anesthetized, shaved, and injected intradermally with 0.1 ml of PPD (PPD-S2, PPD-SSI, or PPD-KIT) containing 1, 5, 10, or 20 tuberculin units (TU; 1 TU is equal to 0.2 μg/ml protein). Reactions were read at 24, 48, and 72 h and expressed in millimetres of diameter of erythema or induration.

Histopathology

Skin samples were obtained at 72 h post-intradermal injection of PPD. These samples were placed on a piece of paper in cassettes, fixed in 10% neutral-buffered formaldehyde for 72 h, and embedded in paraffin and followed by sectioning. Tissue sections (4-5 μm) were stained with hematoxylene and eosin (HE) and examined on an Olympus BX41 microscope equipped with a Q-Color3 camera (Olympus) and Stereo Investigator software (MicroBrightField Inc., Williston, VT). The epidermal, dermal and subcutaneous changes were assigned a severity scored based on the following features 1) epidermis: basal cell hyperplasia and acanthosis, necrosis or apoptosis, intraepidermal microabscesses, subepidermal clefts, transepithelial leukocyte migration, ortho- and para-hyperkeratosis; 2) dermis: edema, collagen necrosis, vasculitis, transendothelial leukocyte migration, neutrophil infiltration, monocyte infiltration, mast cell infiltration and degranulation; and 3) subcutis: necrosis, edema, neutrophil infiltration, lymphocyte & plasma cell infiltration, monocyte-macrophage infiltration and vasculitis. Each feature was scored on a scale of 0 to 4, and the total score for each sample summed up. Photomicrographs are shown without further manipulations.

Results

Protein composition of PPD-S2

Initial assessment of a standard PPD preparation (PPD-S2) used by the U.S. FDA was performed by SDS-PAGE, but as expected this method was unable to resolve discreet protein bands (Fig. 1A),confirming that protein denaturation and degradation occurs during the production process [50]. Similarly, PPD was refractory to 2D-PAGE analyses (data not shown). Thus, we hypothesized that the denatured proteins and representative peptides of PPD could be identified by a non-gel based proteomic approach. PPD-S2 was reduced, alkylated with iodoacetamide and digested with trypsin or chymotrypsin to yield RAT-PPD or RAC-PPD, respectively. Analysis of the RAT and RAC-PPD by SDS-PAGE confirmed that both trypsin and chymotrypsin digestion reduced the overall molecular mass of PPD-S2 (Fig. 1A). However, the RAC-PPD displayed three relatively well resolved bands of ∼37, 25, and 15 kDa. In gel digestion of these products with trypsin and analysis of the resulting peptides by MS/MS revealed that all three bands possessed peptides of GroES (Rv3418c) and the 15 kDa product also contained peptides of GroEL2 (Rv0440) and MPT63 (Rv1926c).

Figure 1.

SDS-PAGE and Western blot analyses of PPD-S2 preparations. A) A silver stained SDS-polyacrylamide gel of PPD-S2 untreated (UN), reduced and alkylated with iodoacetamide (RA), and digests of the reduced and alkylated material with Trypsin (RAT) and chymotrypsin (RAC). The double asterisk mark products that were resistant to chymotrypsin digestion. B) Western blot analysis of untreated PPD-S2 and the cytosol (CYT) preparation of M. tuberculosis for the presence of GroES, GroEL2, Acr (Rv2031c) and DnaK (Rv0350). The left panel is a silver stained gel of PPD-S2 and cytosol. The panels marked GroES, GroEL2, Acr, and DnaK are Western blots probed with the monoclonal antibodies SA12, CBA1 (IT56), CS49, and HAT3 (IT41), respectively.

The non-gel based proteomic analysis of PPD-S2 was initiated with 2D-LC separation of RAT- and RAC-PPD via SCX- and RP-HPLC. This process yielded 208 and 149 fractions respectively, for RAT- and RAC-PPD. The 357 2D-LC fractions were resolved by liquid chromatography coupled to LC-MS/MS and data analyzed by Mascot and SEQUEST [51]. Statistical validation of peptide assignments using the program thresholds, reduction of spectra by manual interrogation and FDR analysis resulted in identification of 3,873 bona fide peptide spectra. Further validation of these using the ProteinProphet algorithm [52] led to identification of 240 protein representatives in PPD-S2 at a confidence level greater than 90%. Thus, we determined that at least 6.0% of coding sequences in the M. tuberculosis genome were represented in PPD-S2 (Supplementary Tables 1 and 2). Of the 240 protein representatives in PPD-S2, 205 and 77were identified in RAT- and RAC-PPD preparations, respectively, with 75 protein representatives common to both. One hundred seven of the 240 protein representatives identified in PPD were based on a single peptide identified by MS/MS. Stringent criteria (see experimental design) were required for these single peptide hits to be designated as true components of PPD-S2.

Identification of the most abundant protein representatives of PPD

Our analyses of PPD-S2 demonstrated that a large number (50) of previously identified T cell antigens were present (Table 1; [53-57]). Given the complexity of PPD it is also expected that T cell reactivity is largely driven by the most abundant T cell antigens. Thus, to investigate the relative quantities of individual protein representatives the method of spectral counting [41] was employed. The ten most abundant protein representatives in RAT- and RAC-PPD are shown in Table 2. Of these, nine and eight of the most abundant proteins representatives in RAT- and RAC-PPD, respectively, were known T cell antigens. Spectral counting also revealed 57.9% and 29.8% of the accepted spectra for RAT- and RAC-PPD, respectively, belonged to four heat shock proteins (HSPs), DnaK (Rv0350), GroEL2 (Rv0440), Acr/HspX (Rv2031c, RAT-PPD only), and GroES (Rv3418c). Analysis of undigested PPD by Western blot confirmed the presence of these four proteins and notably, that they were aggregates and/or partially degraded in PPD (Fig. 1B).

Table 1.

T cell antigens of PPD-S2

Rv Number	Protein Name	Mol. Mass (Da)	Ref
Rv0009	PpiA	19221.6	50
Rv0054	Ssb	17303.1	50
Rv0125	PepA	34907	52
Rv0129c	FbpC2	36753.8	51
Rv0350	DnaK	66812.9	51
Rv0440	GroEL2	56709.3	51
Rv0475	HbhA	21517.1	51
Rv0577	Tb27.3	27324.5	50
Rv0652	RplL	13422.6	50
Rv0667	RpoB	129200.9	54
Rv0685	Tuf	43542.6	52
Rv0733	Adk	20075.6	50
Rv0798c	Cfp29	28812.9	51
Rv0831c	Hypothetical protein	30171.2	53
Rv0934	PhoS1	38193.8	51
Rv0952	SucD	31211.6	50
Rv1211	Hypothetical protein	7792	50
Rv1352	Hypothetical protein	12831.6	50
Rv1411c	LprG	24529.7	51
Rv1626	Probable transcriptional regulator	22651.8	50
Rv1827	Cfp17	17200.5	51
Rv1860	ModD	32702.8	51
Rv1886c	FbpB	34563.6	51
Rv1908c	KatG	80555.9	51
Rv1926c	Mpt63	16495.8	51
Rv1932	Tpx	16878.2	51
Rv1980c	Mpt64	24805.3	51
Rv1984c	Cfp21	21763.9	51
Rv2031c	HspX	16209.6	51
Rv2220	GlnA1	53522	51
Rv2346c	EsxO	9935.6	53
Rv2376	Cfp2	16617.0	53
Rv2428	AhpC	21548.7	50
Rv2461c	ClpP	21658.5	50
Rv2476c	Gdh	176884.9	52
Rv2626c	Hypothetical protein	15499.6	51
Rv2744c	35kd_ag	29239.9	52
Rv2780	Ald	38695.2	52
Rv2878c	Mpt53	18364.9	51
Rv2945	LppX	24121.9	52
Rv3028c	FixB	31672	50
Rv3029c	FixA	28062.9	50
Rv3418c	GroES	10754.1	51
Rv3763	LpqH	15096.2	51
Rv3804c	FbpA	35668.3	51
Rv3841	BfrB	20423.6	50
Rv3846	SodA	22984.3	51
Rv3874	Cfp10	10775.6	51
Rv3875	Esat6	9885.8	51
Rv3914	TrxC	12526.5	51

Table 2.

Dominant Protein Representatives Identified in PPD-S2

PPD Digestion	Protein Name	# of Spectra (Quantitative value)	% Total Weighted Spectra
Trypsin	Co-chaperonin GroES	512	26.1
	Chaperonin GroEL	378	19.2
	HEAT SHOCK PROTEIN HSPX (ACR)	159	8.1
	Molecular chaperone DnaK	88	4.5
	Elongation factor Tu	51	2.6
	ALANINE AND PROLINE RICH SECRETED PROTEIN APA (MPT-32)	48	2.4
	SECRETED L-ALANINE DEHYDROGENASE ALD	46	2.3
	Acyl carrier protein	40	2.0
	10 KDA CULTURE FILTRATE ANTIGEN ESXB (CFP10)	34	1.7
	THIOREDOXIN TRXC (MPT46)	33	1.7
Chymotrypsin	Co-chaperonin GroES	383	19.8
	Acyl carrier protein	126	6.5
	Hypothetical protein Rv0020c	116	6.0
	Chaperonin GroEL	105	5.4
	Molecular chaperone DnaK	89	4.6
	CATALASE-PEROXIDASE-PEROXYNITRITASE (KATG)	89	4.6
	IMMUNOGENIC PROTEIN MPT63	58	3.0
	CONSERVED HYPOTHETICAL PROTEIN CFP17	58	3.0
	10 KDA CULTURE FILTRATE ANTIGEN ESXB (CFP10)	53	2.7
	Thiol peroxidase	47	2.4

Biological variability among PPD preparations

Production methods for PPD can vary among manufacturers, and this variability correlates with variance in potency [58]. To examine this phenomenon three PPD preparations (PPD-S2, PPD-RT23, and PPD-KIT produced at the Korean Institute of Tuberculosis) were assessed for biological activity. Hartley guinea pigs infected with a low dose aerosol of virulent M. tuberculosis [59] were injected intradermally with 0.2 μg of each preparation and erythema measured at 24, 48, and 72 h. For all preparations DTH responses were observed at 24 h, increased slightly at 48 h, and began to wane at 72 h (Fig. 2A). A comparison of the mean erythematous responses at each time point did not show significant differences between PPD formulations. However, analysis of variance among animals using the Tukey-Kramer post-test analysis demonstrated greater variability in the DTH responses to PPD-KIT and PPD-RT-23 at 48 h compared to PPD-S2, with p values of 0.034 and 0.026, respectively. Histopathology of skin biopsies from infected animals sampled at 72 h post PPD injection revealed severe inflammatory reactions characterized by edema, fibrin transudation, vasculitis, infiltration of neutrophils, monocyte-macrophages and lymphocytes, and tissue necrosis regardless of the PPD preparation (Fig. 2 B-D). Yet, inflammation was notably more severe in animals receiving PPD-KIT (Fig. 2C) and PPD-RT23 (Fig 2D) as compared to the PPD-S2 (Fig. 2B). The increased inflammation was due to enhanced edema, fibrin transudation, haemorrhage and very pronounced leukocyte infiltration. Additionally, edema of the subcutis was observed in animals receiving PPD-KIT.

Figure 2.

Analysis of the DTH response to three preparations of PPD. A) Strength of the DTH response as measured by erythematous reactions at PPD injection sites of guinea pigs infected with M. tuberculosis for 10 weeks. Measurements were made at 24, and 48 h post intradermal injection of PPD-S2 (diagonally slashed bar), PPD-KIT (speckled bar), and PPD-RT23 (black bar). The saline control is represented with the open bar. All data are average values and standard deviation of five guinea pigs per PPD preparation. Panels B-D are photomicrographs of histological preparations of dermal tissue from PPD-S2 (panel B), PPD-KIT (panel C), and PPD-RT23 (panel D) injection sites. Microscopic analysis of tissue demonstrated extensive inflammation typical of a DTH response for all PPD preparations. For reference, layers of the integumentary system are identified; dermis (D), subcutis (SC), and muscle (M). All photomicrographs are at 40 Xdsew34 magnification.

Compositional variability among PPD preparations

The histological differences in the DTH response may be associated with significant biochemical differences present in PPD preparations. While total carbohydrate, protein, and nucleic acid content are known to vary among PPD preparations (unpublished data M.J. Brennan), differences in protein profiles have not been assessed. The complexity and characteristics of the protein representatives in PPD do not allow for classical quantitative proteomic studies, such as 2-D PAGE or Western blot. Nevertheless the PPD preparations should be amenable to relative quantification via spectral counting [47, 48]. Thus, the three PPD preparations were processed to generate RAT-PPDs and assayed for quantitative differences in their peptide profiles as described in the methods. Three pair-wise PPD combinations (PPD-S2 vs PPD-S2, PPD-RT23 vs PPD-S2 and PPD-KIT vs PPD-S2) were analyzed to identify statistical differences in individual protein abundance between samples. Proteins were removed from analysis if they did not meet the validation criteria (described in the methods) or, if they met the validation criteria but were not present in at least 2 biological samples. After removing proteins not meeting this criteria, the number of spectra for identical peptide ions were normalized as described in the methods. After normalization, 173 proteins remained in PPD-S2. Of these 173 proteins, 108 were common to PPD-RT23, and 120 were common to PPD-KIT for comparison. Analysis of global clustering between the expression level for identical genes demonstrated that biological replicates clustered together and that PPD-RT23 was more similar to PPD-KIT then either were to PPD-S2 (Fig 3).Analysis of the paired data sets demonstrated 28 and 17 proteins were significantly different (p < 0.01) between PPD-S2 and PPD-RT23 or PPD-S2 and PPD-KIT, and an additional 31 proteins for each pair that were significantly different with p < 0.05 (Figure 3). Specific analysis of relative quantities of the 10 dominant protein representatives found in PPD-S2 (Table 2) versus PPD-RT23 or PPD-KIT, demonstrated that four protein representatives were present in different quantities in both PPD-RT23 and PPD-KIT versus PPD-S2 (Table 3), including two proteins, DnaK and ALD, present in lower quantities in both PPD-RT23 and PPD-KIT versus PPD-S2, and one protein, EsxB (CFP-10), present in higher quantities in both PPD-RT23 and PPD-KIT versus PPD-S2 (Table 3). Given the known potential of EsxB to induce potent, even cytotoxic DTH responses [60], the possible role of other esx proteins to contribute to the immunopathogenesis of M. tb (reviewed in Brodin, et al [61]) and our findings of severe inflammatory responses for PPD-RT23 and PPD-KIT (Figure 2), the relative quantity of other members of the Esx family were examined. Five additional Esx proteins were common to all three PPD preparations (EsxA, L, O, G, and K), and for all five, the relative quantity of each protein was significantly greater (p < 0.05) in both PPD-RT23 and PPD-KIT versus PPD-S2 (Table 3).

Figure 3.

Hierarchically clustered (HCA) heat map of the protein representatives present in the PPD samples. Each row represents individual proteins and each column represents a PPD sample, each of which (PPD-S2, PPD-RT23, and PPD-KIT) were run in triplicate. Cells are colored by z-score which represents the deviation for each protein in each sample from the mean spectral count for that protein. The complete list of proteins compared is available (Supplementary Table 3).

Table 3.

Comparison of Dominant Protein Representatives and Esx Proteins in PPDs

PPD Comparison	Protein Name	P value	Fold change relative to PPD-S2*
PPD-RT23 vs PPD-S2
RAT-PPD Dominant
proteins	Co-chaperonin GroES	NS	ND
	Chaperonin GroEL	p < 0.01	2.9
	HEAT SHOCK PROTEIN HSPX (ACR)	NS	ND
	Molecular chaperone DnaK	p < 0.01	4.9
	Elongation factor Tu	p < 0.01	7.4
	ALANINE AND PROLINE RICH SECRETED PROTEIN APA (MPT-32)	p < 0.01	-3.6
	SECRETED L-ALANINE DEHYDROGENASE ALD	p < 0.05	Detected in PPD-S2 Only
	Acyl carrier protein	NS	ND
	10 KDA CULTURE FILTRATE ANTIGEN esxB (CFP10)	p < 0.01	-2.9
	THIOREDOXIN TRXC (MPT46)	p < 0.05	2.1
esx proteins	esxA	p < 0.05	-2.1
	esxL	p < 0.01	-3.9
	esxO	p < 0.01	-10.1
	esxG	p < 0.01	Not detected in PPD-S2
	esxK	p < 0.01	Not detected in PPD-S2
PPD-KIT vs PPD-S2
RAT-PPD Dominant
proteins	Co-chaperonin GroES	NS	ND
	Chaperonin GroEL	NS	ND
	HEAT SHOCK PROTEIN HSPX (ACR)	p < 0.01	2.3
	Molecular chaperone DnaK	p < 0.05	4.9
	Elongation factor Tu	p < 0.05	-1.7
	ALANINE AND PROLINE RICH SECRETED PROTEIN APA (MPT-32)	NS	ND
	SECRETED L-ALANINE DEHYDROGENASE ALD	p < 0.05	Detected in PPD-S2 Only
	Acyl carrier protein	p < 0.05	1.0
	10 KDA CULTURE FILTRATE ANTIGEN esxB (CFP10)	p < 0.01	-8.2
	THIOREDOXIN TRXC (MPT46)	NS	ND
esx proteins	esxA	p < 0.05	-5.1
	esxL	p < 0.05	-7.3
	esxO	p < 0.01	-9.4
	esxG	p < 0.01	Not detected in PPD-S2
	esxK	p < 0.05	Not detected in PPD-S2

NS = Not Significant, ND = Not Determined.

^*Fold change relative to PPD-S2, where a positive fold change indicates significantly more counted spectra in PPD-S2 and a negative fold change indicates significantly more counted spectra in either PPD-RT23 or PPD-KIT.

Discussion

Despite significant advances in tuberculosis control since the first use of tuberculin as a diagnostic reagent, this disease remains a major global health issue. One of the most pressing issues is improved diagnostics. An alternative to PPD for diagnosis of tuberculosis are the use of interferon gamma release assays (IGRAs). These tests measure the cellular immune response through the release of interferon gamma after stimulation of peripheral blood mononuclear cells collected from suspect TB patients with Mtb specific proteins/peptides. As a result, IGRAs are exceptionally specific and sensitive, however they require significant lab infrastructure and are cost prohibitive in resource limited countries. The PPD skin test is also highly sensitive and in contrast to the IGRAs are inexpensive, stable in a range of environmental conditions [62, 63] and can be administered in the field. However, PPD is not widely applied for TB diagnosis in those regions with the greatest burden of disease because of its inability to differentiate between active and latent disease and continuing concerns with specificity. A major impediment to developing a better skin test antigen has been a lack of knowledge as to the molecular composition of PPD and the contribution of individual gene products. Early studies on the definition and standardization of PPD led Affronti to hypothesize that specific qualities of PPD are essential for its immunological activity: 1) PPD must be prepared from denatured material to alleviate sensitization from repeated administration, 2) PPD represents protein aggregates and not individually resolvable products, and 3) the biological activity of PPD is not attributable to a single protein [50].Our studies of human PPD derived from M. tuberculosis via the process described by Seibert [4] has in-part tested these hypotheses and clearly demonstrated that PPD is a complex mixture of denatured proteins and protein fragments representing at least 6.0 % of the proteins encoded on the M. tuberculosis genome. Moreover, even with modern 2D-LC approaches the complexity of PPD did not allow for clean resolution of individual peptides or for a specific peptide to be found in only one 2D-LC fraction. The denatured state and complexity of PPD also explains why previous investigators were unable to define more than a few major constituents. The complexity and form of the products in PPD likely resulted in the misidentification of PPD proteins, such as the 9.5 kDa product of PPD erroneously identified by Kuwabara [11], that is now known not to be a true gene product of M. tuberculosis [35, 36]. Another early effort to characterize the composition of PPD was based on antibodies to M. tuberculosis proteins. Specifically, Rv0350 (DnaK) and Rv0440 (GroEL2)were described as components of PPD using polyclonal sera with immuno-electrophoresis [64]. Previous 2D-PAGE analysis of PPD prepared by SSI revealed an unresolved mixture of denatured proteins, similar to what we observed [9]. However, this earlier study was able to differentiate two low molecular mass proteins that were identified as Rv3418c (GroES) and Rv0652 (L7/L2 50s ribosomal protein). Based on our analyses, all of these proteins were present in the PPD-S2 and three (DnaK, GroEL, and GroES) were identified as dominant products. Coler et al. purified from PPD a low molecular weight protein termed DPPD that induced a significant DTH response in tuberculosis patients [8]. It should be noted, however, that a standard PPD was not used to isolate DPPD, and unlike PPD-S2 or other PPD standards the PPD used by Coler and colleagues produced a distinct banding pattern by SDS-PAGE. Most recently proteomic methodologies similar to those applied here were used to analyze the composition of PPD derived from Mycobacterium bovis, a close relative of M. tuberculosis [65]. Two separate M. bovis PPD preparations were evaluated and a total of 116 proteins representatives were identified. The difference in the total number of protein representatives identified in PPD-S2 (240) and that for the M. bovis PPD preparations (116) is likely a reflection of the methodologies used. We utilized 2D-LC separation of trypsin and chymotrypsin digested PPD-S2, while the evaluation of M. bovis PPD was performed with peptides derived from trypsin digests of gel slices after SDS-PAGE separation of PPD. Thus, the large amount of starting materials used in the evaluation of PPD-S2, the inclusion of both trypsin and chymotrypsin digests, and the extensive fractionation we employed likely resulted in the identification of more protein representatives. Of the 116 protein representatives in the M. bovis PPD preparations, 80 (69%) were also identified in our analyses of the M. tuberculosis PPD. Likewise, seven of the ten most abundant protein representatives identified in RAT- or RAC-PPD preparations based on spectral counting [41] were also identified as highly abundant in the M. bovis PPD using the emPAI function of MASCOT [66]. More interestingly, the seven highly abundant protein representatives that correlated to those in M. bovis PPD were the same seven found to be highly abundant in both the RAT- and RAC-PPD preparations (Table 2). Thus, although there are considerable differences in the experimental approaches and likely the methods used to generate the various PPD preparations it would appear that the most dominant products of PPD remain consistent.

It is widely accepted that variability exists in the potency of different PPD preparations [58]. However, no studies have been published to define compositional differences among PPD standards or commercially manufactured PPDs used in human diagnostics. Given the complexity of PPD and the state of the PPD proteins, a protein-by-protein comparison of the composition of individual PPD preparations would be difficult based on current proteomic methods. The proteomic evaluation of M. bovis PPD included the comparison of PPDs manufactured in Brazil and the United Kingdom, and showed that only 33 of the 116 total protein representatives were present in both preparations [65]. Additionally, only four of the ten most abundant protein representatives in the two preparations overlapped. The protocols used to generate the Brazilian and United Kingdom PPDs differed significantly and this likely contributed to the modest overlap (28%) in identifiable products. However, it was not possible to determine how experimental variability in the proteomic analysis of each sample contributed to the lack of concordance between the two M. bovis PPD preparations. In contrast to the protein-by-protein comparison our evaluation of PPD-S2, PPD-RT23, and PPD-KIT utilized nanospray-LTQ tandem MS with spectral counting and comparative evaluation of proteins after quantile normalization of individual peptides. By this method, 72% (173) of the proteins identified from 2D-LC analysis of PPD-S2 could be compared between replicate samples of PPD-S2, and 45% (108) and 50% (120) of these proteins could be compared between PPD-RT23 and PPD-KIT versus PPD-S2, respectively. This analysis demonstrated that the other two PPD preparations differed considerably from PPD-S2. These relative compositional differences also correlated to differences in the histopathology of the DTH response to the PPD preparations, but not in the measurement of erythema. Discreet analysis of the differences in protein composition identified relative increases in the esx proteins present in PPD-RT23 and PPD-KIT. Although speculative, the increased abundance of this protein family in PPD-RT23 and PPD-KIT may contribute to the increased inflammatory responses observed when these two reagents were administered in this guinea pig model. Thus, these series of experiments have demonstrated methods by which individual preparations of PPD can be compared both at a molecular and biological level, and could be applied to the further standardization of PPD. With this base knowledge and approach, it is also possible to extend the analyses of PPD preparations such that individual protein representatives with quantifiable differences between PPD preparations are identified and the impact that specific compositional differences have on the biological activity of PPD can be measured at a cellular level.

Elucidating the relative quantities of individual protein representative in PPD-S2 led to the recognition of a disproportionate abundance of the heat shock proteins (Acr, DnaK, GroEL, and GroES), and given this preponderance it is likely the heat shock proteins significantly influence the biological activity of PPD. A lack of specificity, has been cited as one of the major limitations to the use of PPD as a diagnostic for tuberculosis in endemic settings [67] and the potential of cross reactive epitopes on the highly conserved heat shock proteins [68] would be expected to account for the reported cross reactivity to PPD. However, a recent retrospective review of studies conducted since 1966 argues that PPD does possess high specificity, even in individuals vaccinated with M. bovis BCG as infants and skin tested at greater than ten years of age [69]. This review also revealed that exposure to non-tuberculosis mycobacteria (NTM) does not significantly contribute to false positive PPD responses unless the individual is in a region with a high prevalence of NTM and low prevalence of M. tuberculosis infections. Thus, the data on the composition of PPD when coupled with this retrospective evaluation of clinical studies would suggest that the abundance heat shock proteins in PPD is not a shortcoming of this diagnostic reagent. On the other hand, recent evaluation of bacterial heat shock proteins and their interaction with the innate and adaptive immune response [70-72] would argue that these products may actually enhance sensitivity of PPD to a level that cannot be achieved with individual purified M. tuberculosis proteins predicted to provide greater specificity. Moreover, the abundance of the heat shock proteins and their chaperone activity [73] would explain the exquisite solubility of PPD; a property that would not be expected given the protein aggregation as observed by SDS-PAGE and the denaturing conditions used to generate PPD. Given the new understanding of the protein representatives that comprise PPD, these hypotheses are now being tested. Specifically, the PPD compositional data is being used to develop second generation skin test antigens that retain the qualities of PPD but are well defined in composition [74].

Abbreviations

PPD

Purified protein derivative

ICAT

isotope coded affinity tag

DTH

delayed type hypersensitivity

LC-MS/MS

liquid chromatography electrospray tandem mass spectrometry

SCX

strong cation exchange

RP

reversed phase

HPLC

high pressure liquid chromatography

RA

reduced and alkylated

RAT

reduced, alkylated and trypsin digested

RAC

reduced, alkylated and chymotrypsin digested

HSPs

heat shock proteins

ESI

eletrospray ionization

TU

tuberculin units

CYT

cytosol

UN

untreated

SSI

Statens Serum Institute

KIT

Korea Institute of Tuberculosis