Proteomic Analysis of Cytokeratin Isoforms Uncovers Association with Survival in Lung Adenocarcinoma¹

2D PAGE

Protein samples were solubilized in standard lysis buffer (9.5 M urea, 20 µl of nonionic detergent [Nonidet P-40], 20 µl of ampholines [pH 3.5–10; Pharmacia/LKB, Piscataway, NJ], 20 µl of 2-mercaptoethanol, and 0.89 M phenylmethylsulfonyl fluoride per milliliter of deionized water). Sample volumes of between 15 and 30 µl were immediately applied to isoelectric focusing gels containing 50 µl of ampholytes per milliliter (pH 3.5–10). First dimensional separation was performed using 700 V for 16 hours, and then 1000 V for 2 hours at room temperature. Second dimensional separation was accomplished using an 18x 18 cm gel containing an acrylamide gradient of 11.4 to 14 g/dl. Samples were run in 20 gel batches. After separation, the protein spots were visualized by a photochemical silver-based staining technique.

Detection and Quantification

Following 2D separation, each gel was scanned using a Kodak CCD camera. A 1024x1024 pixel format was used, yielding pixel widths of 163 µm where each pixel had 256 possible gray scale values (optical density). Spot detection was accomplished by Bio Image Visage System software (Bioimage, Ann Arbor, MI). Each gel generated 1600 to 2200 detectable spots, with 820 well-defined spots in most samples, chosen to obtain quantitative measurements. The integrated intensity, which is the value of each spot, was calculated as the measured optical density unitsxsquare millimeters. Then spots from each gel were matched to the 820 spots on a “master” gel [11] to allow for the identification of identical polypeptides between each gel. A total of 250 spots were chosen as ubiquitously expressed reference spots to allow adjustment for variations in protein loading and gel staining. Each of the 820 spots was then mathematically adjusted in relation to the reference spots [12].

MS

Protein spots identified for analysis were extracted from preparative 2D gels of extracts from A549 lung adenocarcinoma cell lysates (obtained from ATCC) or lung adenocarcinoma tissue lysates (obtained from patient). The conditions were identical to the analytical 2D gel, except there was a 30% greater protein loading, and were silver-stained by successive incubations in 0.02% sodium thiosulfate for 2 minutes, 0.1% silver nitrate for 40 minutes, and then 0.014% formaldehyde plus 2% sodium carbonate for 10 minutes. Identification of proteins was performed by trypsin digestion followed by either matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) in the Perseptive Voyager Biospectrometry Workstation (PerSeptive Biosystem, Framingham, MA), or nanoflow capillary liquid chromatography coupled with electrospray tandem mass spectrometry (ESI MS/MS) in a Q-TOF micro (Micromass, Manchester, UK). MALDI-TOF MS gave a “fingerprint” for each spot based on the molecular mass of trypsin-digested products and compared the resulted mass with known trypsin digest databases using the MSFit database searching (http://prospector.ucsf.edu/ucsfhtml3.2/msfit.htm). MS/MS spectra produced by ESI MS/MS were automatically processed and searched against a nonredundant database using ProteinLynx Global SERVER (www. micromass.co.uk).

RNA Isolation for Microarrays

RNA was isolated from 75 of the tumors and the 10 normal samples used in the protein studies. Two contiguous 2-mm³ samples were removed for RNA and protein isolation, respectively. Total cellular RNA was isolated using Trisol reagent (Life Technologies) and subjected to further purification using RNeasy columns (Qiagen, Valencia, CA). Five micrograms of total RNA was used as template. All protocols used for mRNA reverse transcriptase, second strand synthesis, production of cDNA and RNA amplification, hybridization, and washing conditions for the 6800 gene HUGeneFL oligonucleotide arrays are as provided by the manufacturer (Affymetrix, Santa Clara, CA). More detailed information is provided in Giordano et al. [13].

Mutational Analysis of K-ras

Genomic DNA was isolated from each tumor sample and 50 ng was subjected to PCR amplification using the primers that encompass codons 12 and 13 of the K-ras gene. The sequences of forward and reverse primers are 5′ TATAAGGCCTGCTGAAAAT 3′ and 5′ CCTGCACCAGTAATATGC 3′, respectively. Two nanograms of purified PCR products containing the exon 1 of the K-ras gene was then subjected to thermal cycle sequencing with an internal nested primer (5′ AGGCCTGCTGAAAATGACT 3′) and resolved in 8% urea PAGE gels, dried, and exposed to Phosphor-Image screens and visualized using a Phosphor-Image scanner (Molecular Dynamics, Sunnyvale, CA). The mutations were determined by comparing each tumor DNA sequences of K-ras 12th and 13th codons to its wild-type sequence GGTGGC.

Immunohistochemical Staining

Diagnoses for all primary lung adenocarcinomas used in this study were confirmed by a board-certified pathologist. A tissue microarray block was assembled based on the best morphological areas of the tumors used in this study according to the method of Kononen et al. [14]. Deparaffinized sections of the pulmonary adenocarcinoma tissue microarray were microwave-pretreated in citric acid to retrieve antigenicity. The sections were incubated with 1% hydrogen peroxide for 60 minutes to inhibit endogenous peroxidase activity at room temperature. Following blocking to reduce nonspecific binding, the sections were incubated with primary antibodies overnight at 4°C. Antibodies included anti-p53 (M-7001; 1:500), anti-CK19 (M0772; 1:500), and anti-CK7 (OV-TL 12/30; 1:500) from DAKO (Carpenteria, CA); anti-CK7 (K72.7; 1:500; Neomarkers, Fremont, CA); and anti-CK18 (DC-10; 1:500) and anti-CK8 (TS1; 1:500) from Novo Castra Laboratories (Newcastle, UK). The immuno-complex was visualized by the immunoglobulin enzyme bridge technique using a Vector ABC peroxidase kit (Vector Laboratories, Burlingame, CA). The enzyme substrate was 3,3′ diaminobenzidine tetrachloride, resulting in a brown reactant. The sections were lightly counterstained with hematoxylin.

2D Western Blotting

Protein extracts of A549 lung adenocarcinoma cells were run on 2D gels using the identical conditions as used for the analytical 2D gels. The separated proteins were transferred onto polyvinylidene fluoride membranes and incubated for 2 hours at room temperature with a blocking buffer consisting of TBST (Tris-buffered saline, 0.01% Tween 20) and 5% nonfat dry milk. Individual membranes were washed and incubated with monoclonal antibodies listed above for immunohistochemistry and used at 1:500 dilution. After additional washes with TBST, the membranes were incubated with a secondary antibody conjugated with horseradish peroxidase (HRP) at a 1:5000 dilution for 1 hour, further washed, and then incubated for 1 minute with enhanced chemiluminescence (ECL) (Pierce, Rockford, IL).

CK Isoforms Associated with Patient Survival and Other Clinical-Pathologic Variables

Univariate Cox proportional hazards regression analysis revealed that one CK8, one CK19, and all five CK7 isoforms identified by MS were significantly associated (P<.05) with patient survival. Interestingly, the five CK7 isoforms had estimated molecular mass ranging from 36.7 to 41.7 kDa, and estimated pIs ranging from 4.6 to 4.9 that are quite different from the reported native CK7 molecular mass of 51.3 kDa and pI of 5.5. The estimated pI and molecular mass for CK 8, 18, and 19 isoforms were relatively similar to their reported standards [3].

Of the 21 CK isoforms examined, only a subset showed a significant correlation to other clinical-pathological variables (Table 3). Relationships between these variables may provide some insight into the role of these CKs in lung cancer. F-tests showed that some CK isoforms such as CK7 (no. 691), which had a relationship to survival, were decreased in bronchial-derived tumors and those with a positive lymphocytic response. None of the other CK7 isoforms showed a relationship to any clinical-pathological variables. However, three CK8 isoforms showed relationships to differentiation, positive lymphocytic response, or K-ras mutation status. One CK18 isoform was also increased in tumors with a positive lymphocytic response, and of the three CK19 isoforms, one was increased in tumors with p53 and K-ras mutations, and one was decreased in poorly differentiated tumors.

2D Western Blot Analysis of CKs

We determined the patterns of reactivity of CKs with 2D Western blot analysis (Figure 2) of the regions of 2D gels demonstrating the immunoreactivity shown in Figure 1, boxes A–D. Outlined box B indicated the region from which the five overexpressed CK7 proteins identified by MS were located. The CK8 antibody identified at least 15 separate immunoreactive proteins (Figure 2B) including the eight isoforms that were identified by MS (Figure 1, box A), and for which quantitative analysis was performed (Figure 2A). Two different monoclonal antibodies to CK7 representing two different clones were examined (Figure 2, C and D). Both antibodies detected some isoforms in common as well as distinct CK7 isoforms. The location of the immunoreactive CK7 isoforms is very close to the location to the CK8 fragments (Figure 2B), and three of the isoforms (436, 444, 446) appear to overlap between CK7 and CK8 fragments. All other fragments are clearly distinct from each other, although their location is similar. The redundancy of the immunostaining may be the result of the extensive sequence similarity of CK7 and CK8 [16,17].

Open in a separate window

Figure 2

(A) 2D gel digital image of the silver-stained region shown in Figure 1, Box A, containing eight CK8 spots expressed in a primary lung adenocarcinoma. (B) 2D Western blot analysis using anti-CK8 monoclonal antibody clone TS1 showing the eight isoforms quantified of the approximately 20 that are immunoreactive in A549 cells. The region is shown in Figure 1, Box A. (C) 2D Western blot analysis using anti-CK7 monoclonal antibody clone OV-TL 12/30. The region is shown in Figure 1, Box A. (D) 2D Western blot analysis using anti-CK7 monoclonal antibody clone K72.7. Three of the spots (nos. 436, 444, and 446) appear to be at very similar location to CK8 spots and are detected with both CK7 antibodies. None of the five CK7 isoforms detected by mass spectrometry reacted with either antibody. The unnumbered spots, although detected in the A549, were not abundant enough to be quantified in the lung and tumor samples. The region is shown in Figure 1, Box A. (E) 2D Western blot analysis using anti-CK18 monoclonal antibody clone DC-10. The region is shown in Figure 1, Box C. Five individual isoforms were quantified. (F) 2D Western blot analysis using anti-CK19 monoclonal antibody clone M0772. The region is shown in Figure 1, Box D. Three CK19 isoforms were quantified.

The immunoreactive CK7 proteins have the same pI and molecular mass predicted from their sequence; however, neither of the two CK7 antibodies directed against distinct clones reacted with the five isoforms identified by MS, suggesting that these smaller CK7 proteins represented cleavage products that lacked the epitopes recognized by these monoclonal antibodies. Quantitative values for the immunoreactive CK7 isoforms could not be obtained from the silver-stained gels due to low abundance, although these forms are relatively abundant in the A549 lung cell line. All five CK7 isoforms were significantly associated with survival (P<.05), significantly increased in the adenocarcinomas, and frequently expressed (58.1% vs 39.8%) relative to normal lung (Table 2). None of the immunoreactive CK7/CK8 isoforms showed a significant relationship with survival; however, spot nos. 444 and 446 were significantly increased in tumors relative to normal lung. In addition, spot no. 436 showed a significant increased expression in stage III relative to stage I tumors, and significant decreased expression in tumors showing lymphocytic response.

Five CK18 and three CK19 isoforms were detected using MS and 2D Western blots (Figure 2, E and F) and their protein abundance quantitatively analyzed (Table 2). Unlike CK7 and CK8, all CK18 and CK19 isoforms demonstrated similar molecular mass but varied in pI. Interestingly, all five CK18 isoforms were significantly increased in the adenocarcinomas and detected frequently in these tumors relative to normal lung (15–60%), but an association with survival was not observed. The isoform nos. 529 and 523 are the most abundant in both normal lung and lung adenocarcinomas. All three CK19 isoforms were significantly increased in the adenocarcinoma but only one isoform (no. 1955), the most positively charged isoform, was associated with survival and demonstrated the greatest differential expression between normal lung and the adenocarcinomas. This isoform was significantly associated with the CK19 mRNA levels, suggesting that the level of transcription in part can influence its abundance.

CK mRNA Levels and Correlation to Other Genes

Because the same lung adenocarcinomas and nonneoplastic lung samples that were quantitatively analyzed for protein expression were examined at the level of mRNA abundance using oligonucleotide arrays, correlation coefficients between these CK proteins and the levels of their corresponding mRNA could be determined.

A positive correlation was observed between the CK protein expression values for two CK7 isoforms (nos. 165 and 2091) and CK7 mRNA levels (Table 4). One CK8 isoform (no. 439), two CK18 isoforms (nos. 523 and 2324), and one CK19 isoform (no. 1955) were also significantly associated with mRNA levels of corresponding genes. Importantly, univariate Cox proportional hazards regression analysis revealed that CK7, CK8, CK18, and CK19 mRNA showed a significant relationship to survival and all four were significantly increased at the mRNA level in the adenocarcinomas relative to normal lung (Table 5). Further, correlation analysis showed that the mRNA levels for all four CK genes were significantly correlated (P<.05) to one another (Table 6), suggesting that these increases may reflect a common mechanism. Further analysis of 4966-expressed genes examined in these samples revealed two genes encoding liver-specific bHLH-Zip transcription factor and smooth and nonmuscle myosin light chain polypeptide to be significantly correlated to all four CKs.

The authors acknowledge the contributions from Lin Lin, Charles Miller, Melissa Krause, David Wood, Robert Hinderer, Rork Kuick, and Eric Puravs.