- Email: [email protected]
Discovery of cancer biomarkers through the use of mouse models
Cancer Letters 249 (2007) 40–48 www.elsevier.com/locate/canlet
Mini-review
Discovery of cancer biomarkers through the use of mouse models Rork Kuick a, David E. Misek a, David J. Monsma b, Craig P. Webb b, Hong Wang c, Kelli J. Peterson d, Michael Pisano d, Gilbert S. Omenn e, Samir M. Hanash c,* b
a Department of Pediatrics, University of Michigan Medical Center, Ann Arbor, MI, USA Laboratory of Tumor Metastasis and Angiogenesis, Van Andel Research Institute, Grand Rapids, MI, USA c Fred Hutchinson Cancer Research Center, Seattle, WA, USA d Proteomic Research Services, Inc., Ann Arbor, MI, USA e Internal Medicine, University of Michigan Medical Center, Ann Arbor, MI, USA
Abstract Although our understanding of the molecular pathogenesis of common types of cancer has improved considerably, the development of effective strategies for cancer diagnosis and treatment have lagged behind. Mouse models of cancer potentially represent an efficient means for uncovering diagnostic markers as genetic alterations associated with human tumors can be engineered in mice. In addition, defined stages of tumor development, breeding conditions, and blood sampling can all be controlled and standardized to limit heterogeneity. Alternatively human cancer cells can be injected into mice and tumor development monitored in xenotransplants. Mouse-based studies promise to elucidate a repertoire of protein changes that occur in blood and biological fluids during tumor development. This is illustrated in a study in which we have applied a three-dimensional intact protein analysis system (IPAS) to elucidate detectable protein changes in serum from immunodeficient mice with lung xenografts from orthotopically implanted human A549 lung adenocarcinoma cells. With sufficiently detailed protein sequence identifications, the observed protein changes can be attributed to either the host mouse or the human tumor cells. It is noteworthy that the majority of increases identified have corresponded to relatively abundant serum proteins, some of which have previously been reported as increased in the sera of cancer patients. Proteomic studies of mouse models of cancer allow assessment of the range of changes in plasma proteins that occur with tumor development and may lead to the identification of potential cancer markers applicable to humans. Ó 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: Cancer biomarkers; Mouse models for cancer; Proteinases; Mass spectrometry; Serum proteome; Proteinase inhibitors
1. Introduction * Corresponding author. Tel.: +1 206 667 7091; fax: +1 206 667 2537. E-mail address: [email protected] (S.M. Hanash).
The quantitative analysis of proteomes has reached a fairly mature stage, with a rich repertoire of technologies available [1,2]. Recent improve-
0304-3835/$ - see front matter Ó 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.canlet.2006.12.032
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
ments in proteomic profiling methods permit screening biological fluids in order to discover protein markers that are potentially useful for diagnosis, prognosis, or monitoring of disease states such as cancer [3,4]. In particular, mass spectrometry (MS) has evolved enough to detect and identify femtomols of peptides, with a dynamic range over 3–5 orders of magnitude [5]. Quantitative methods currently available for biomarker discovery can be grouped into label-free methods, which compare pure peptide ion intensities between MS analysis, and labeling methods, which are applied to intact proteins or protein digests for comparative analysis. While technical challenges may now be overcome, a major impediment to biomarker discovery is the substantial heterogeneity in disease populations and the substantial challenges involved in designing and executing well controlled studies to discover promising biomarkers [6]. As a means to simplify the process, there is interest in evaluating mouse models of cancer for a variety of applications from diagnostics to therapeutics. 2. Mouse models of cancer Refined genetically engineered mouse (GEM) models of human cancer have been shown to faithfully recapitulate the molecular, biological and clinical features of human disease. GEM models afford defined stages of tumor development, homogenized breeding and environmental conditions, and standardized blood sampling, thereby limiting biological heterogeneity [7]. The concept that plasma from GEM models of cancer contains tumor derived proteins that may be relevant as candidate markers for human cancer is attractive. Only limited data has been obtained to date, including a SELDI study of a pancreatic cancer model [8]. In a recent study, we exploited the merits of another well-characterized GEM model of pancreatic cancer [9] and of an intensive quantitative proteomic approach we have developed to determine whether proteomics technologies allow identification of protein changes associated with tumor development and whether such changes are relevant to human pancreatic cancer. Plasma was sampled from mice at early and advanced stages of tumor development and from matched controls. Among the early and advanced tumor stages, we confidently identified proteins that were distributed across seven orders of magnitude of abundance in plasma. Comparative analysis of candidate bio-
41
markers documented striking concordance of expression in human and mouse pancreatic tissue and in the blood from patients with pancreatic cancer relative to normal specimens. In addition to identifying markers of potential utility for pancreatic cancer diagnosis, our findings indicated that GEM models of cancer, in combination with proteomics, provide a rich source of candidate markers applicable to human cancer. 3. Proteomic analysis of tumor xenotransplants While the potential of GEMs to uncover potential cancer markers is substantial, these models do not readily allow a determination of the tumor origin of the protein changes detectable in blood and biological fluids. On the other hand the use of xenotransplanted human tumor cells would allow a determination of host vs. tumor cell origin of protein changes observed. We have undertaken a study to determine the host vs. tumor cell origin of protein changes observed in mouse blood which we present and review here. In this regard, we have developed methods in which two or more protein samples are labeled with different Cy dyes and separated using an orthogonal, three-dimensional (3-D) intact-protein analysis system (IPAS) to measure protein differences between the samples [10]. Subsequently, mass spectrometry is used to analyze tryptic digests of the protein bands found to differ in order to identify peptides and deduce the proteins present. We have used a cancer model in which human A549 lung adenocarcinoma cells were orthotopically implanted into nude, athymic mice and their sera compared to mice undergoing a sham implantation procedure. We assessed quantitative and qualitative differences in protein expression due to proteins shed by the tumor itself from differences due to protein alterations in the tumor microenvironment or that result from systemic responses, such as acutephase inflammatory protein responses or antibodymediated immune responses. The primary goal was to identify the proteins causing observed differences in sufficient detail to permit determination of whether the source was the human tumor cells or the host mouse. Sera from 5 mice with intrapulmonary tumors and from 8 sham-operated mice were separately pooled. The three most abundant serum/plasma proteins (albumin, IgG, and transferrin) were removed by immunodepletion chromatography. The immunodepleted fractions were concentrated
42
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
and Cy5 and Cy3 dyes were used to label the control and implanted mouse sera, respectively. The labeled samples were combined and separated in three dimensions, based successively on pI, hydrophobicity, and molecular weight. The last dimension consisted of running SDS gels for each fraction. Following gel fixation, Cy3 and Cy5 images were quantitatively analyzed. We carefully examined spots with normalized ratios of spot integrated intensities between the two images of 3.0 or greater (either up or down). 4. Serum protein changes in human tumor bearing mice are predominantly of host origin We compared sera from mice with tumors formed from orthotopically implanted cells of the A549 human lung adenocarcinoma cell line with sera of control nude athymic mice that received a sham implantation procedure. Sera were collected seven weeks after surgery, at which point all five tumor burdened mice used in this study were shown to have extensive lung carcinoma (Fig. 1). Overlaying pairs of gel scans resulted in two-color images, such as that shown in Fig. 2. We selected 46 bands of most intense green color in these images, which corresponded to proteins that were more abundant in the sera of mice with tumors by at least 3-fold, in order to identify the protein responsible for the observed band differences by mass spectrometry. We also selected five red bands, which corresponded to proteins decreased in the tumor-bearing mice. In all, we obtained identifications implicating a single protein as most likely to be responsible for the observed image difference in a total of 30 cases (27 up, and 3 down in mice with tumors), which are summarized in Table 1. The 27 identifications in bands increased in tumors represented 16 different proteins.
We also obtained very confident MS search matches for 14 other bands (all up in mice with tumors), where however we obtained nearly equal quality matches to two or more distinct proteins. For these identifications, summarized in Table 2, we did not select one of the multiple identifications as more likely to be responsible for the differences observed between the sera, although such discrimination could be attempted on the basis of MS peak intensities and other criteria. Of the 41 bands increased in the sera of mice with tumors for which we obtained confident identifications, only two were for human proteins. The first of these was cystatin A (CSTA), a well-known cysteine protease inhibitor, which gave a 6.0-fold increase and was found close to its theoretical mass of 11 kDa on gel 12A (see Fig. 2) [11,12]. The second human protein identified was epidermal fatty acid binding protein 5 (FABP5, EFABP) from a relatively faint band that gave a 4.7-fold increase and was also found close to its 15 kDa theoretical mass [13–15]. The four peptides found to match our spectra were also contained in a hypothetical protein sequence predicted from the similar LOC387934 gene. It was clear from this analysis that by far the most bands found to have increased intensity in mice with tumors were from the host mice rather than the implanted human tumor cells. We found a single increased band containing Serpina1b, and several each for the Serpina1a and Serpina3k proteins, which are mouse paralogs of the human a-1antitrypsin (SERPINA1) and a-1-antichymotrypsin (SERPINA3) proteins. One of the bands found for each of these three proteins was far too low on the gel to be due to the well-known versions of these proteins, and so may have been cleaved by proteases expressed by the tumor cells. Besides these serpin protein elevations, several other proteins were identified for which previous
Fig. 1. Photographs of inflated lungs 7-weeks post sham injection (controls) or implantation of human A549 cells. In the latter mice, the lungs were virtually replaced by lung adenocarcinoma as verified by histopathological staining.
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
43
Fig. 2. Gel 12A from rotofor fraction 12, RP-HPLC fractions 13–39. Bands from approximately 4–200 kDa are visible. In this overlay of two scanned images of the gel, the image of the Cy3 scan is shown as green and the Cy5 scan as red, so that greener colors indicate increases in the sera of mice with A549 implanted cells. Labeled bands contained proteins identified by mass spectrometry as human cystatin A (band 1626), mouse apolipoprotein A-I and b-2-glycoprotein (Apo-H) (1619), mouse Serpina1b (1681), and mouse RIKEN cDNA 1300017J02, which has homology to transferrin (871). See details in Table 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)
reports of increases in cancer patients exist. We found increased anticoagulation factor 2 (F2, prothrombin). Considering the band location and peptide matches from MS searches, the particular band we observed (estimated to be 22 kDa) appears to arise from the mature thrombin heavy chain (theoretically 30 kDa). A band identified as haptoglobin (Hp) at an approximate molecular weight of 43 kDa, is most consistent with the mature b-haptoglobin. Although the mature b-haptoglobin is theoretically only 27 kDa in both human and mouse, it appears as a series of spots in 2D gels at approximately 40–45 kDa due to glycosylation or other modifications. We also identified the mouse pregnancy zone protein (Pzp), a member of the a-2-macroglobulin family, in one case where it was the best hit obtained by MS searches, and also for several other bands for which we did not obtain unique best hits. All of the bands for which Pzp identifications were obtained had much lower molecular weight than the complete Pzp monomer, which is
approximately 170 kDa. Similarly, at least 3, and perhaps all 5, of the increased bands identified as Apoa1 must be cleaved versions, as was a 9 kDa band identified as Apoa4, which is typically seen closer to its theoretically computed mass of 45 kDa. Other likely cleaved proteins identified included bands containing gelsolin (Gsn), and a hemoglobin. 5. Correlations between protein and RNA We previously published gene expression data for mouse tumor xenotransplants [16]. In order to determine if some of the increases for mouse proteins might be due to increased expression in surrounding stromal tissue rather than being acute-phase reactants produced from liver or other distant organs, we used our publicly available data to compare mRNA levels of mouse lung samples adjacent to tumors from implanted A549 cells to normal mouse lung samples. The results are shown in Table 3. Significantly, 2–3-fold increases in some
44
Table 1 Unambiguous protein identifications Mw estimate (kDa)
Finding
Digest’s ID
Gene symbol
Species
Protein title
03A_1831
12.4
Up
26742
FABP5
Human
12A_1626 05A_1194
9.9 58.2
Up Up
26764 05A_A1
CSTA Serpina3k
Human Mouse
05A_1241
52.1
Up
05A_B1
Serpina3k
Mouse
06A_1660
10.7
Up
26749
Serpina3k
Mouse
08A_1772 07A_1174 14A_1411 12A_1681 14A_1428 14A_1599 14A_1460 19A_1762 19A_1773 17A_1651 08A_1856 14A_992
15.7 48.9 26.2 7.5 26.8 10.4 24.6 9.8 8.0 6.5 9.4 53.2
Up Up Up Up Up Up Up Up Up Up Up Up
26754 26750 26777 12A_B1 26779 14A_A1 14A_C1 26792 26793 26771 26756 26774
Serpina1a Serpina1a Serpina1a Serpina1b Apoa1 Apoa1 Apoa1 Apoa1 Apoa1 Apoa2 Apoa4 Apoh
Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse
14A_1024
50.5
Up
26775
Apoh
Mouse
09A_1245 12A_871 14A_810 14A_1071
42.9 77.6 80.4 46.2
Up Up Up Up
09A_A1 26762 26773 26776
Hp 1300017J02Rik 1300017J02Rik Cpn1
Mouse Mouse Mouse Mouse
14A_1482
22.1
Up
26781
F2
Mouse
14A_1522 16A_1409 17A_1652 19A_1738 08A_1421
19.1 33.2 6.2 11.8 42.4
Up Up Up Up Down
26783 26787 26772 26791 26753
Gsn Pzp Hba Hba Hp
Mouse Mouse Mouse Mouse Mouse
Fatty acid binding protein 5 (psoriasis-associated) Cystatin A Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serpina1a protein Serpina1a protein Serpina1a protein Serpina1b protein Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-II Apolipoprotein A-IV b-2-glycoprotein I precursor(Apo-H) b-2-glycoprotein I precursor(Apo-H) Haptoglobin RIKEN cDNA 1300017J02 RIKEN cDNA 1300017J02 Carboxypeptidase N, polypeptide 1 Coagulation factor II; prothrombin Gsn protein Pregnancy zone protein Hemoglobin a, adult chain 1 Hemoglobin a, adult chain 1 Haptoglobin
Accession number (gi)
Theoretical mass (kDa)
MASCOT score
# Distinct peptides
4557581
15.5
118
4
4885165 33859636
11.0 46.9
143 155
5 4
33859636
46.9
68
2
33859636
47.0
341
6
15929675 15929675 15929675 15277553 6753096 2145139 6753096 2145139 2145139 7304897 6680702 231558
45.8 45.8 45.8 45.1 30.6 30.5 30.6 30.5 30.5 11.4 44.7 39.9
540 1103 163 131 825 198 99 171 101 95 279 848
10 20 4 3 22 5 2 6 3 3 6 22
231558
39.9
569
14
8850219 18204720 18204720 13507644
39.2 78.7 78.7 52.1
112 1621 1983 171
3 37 41 4
6753798
71.7
373
9
18606238 6680608 6680175 6680175 8850219
81.0 167.1 15.1 15.1 39.2
172 187 169 376 637
4 6 6 8 14
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
Gel and band of interest
Mouse Cpn1 26786 Down 41.7 16A_1244
Excised bands found to be either increased (up) or decreased (down) at least 3-fold in mice with implanted A549 cells, were subjected to mass spectrometry and database searches using MASCOT software. One search resulted in 5 peptide matches which could be from either the human or mouse peptidylprolyl isomerase A, which is most likely from mouse since the band was more intense in samples from unimplanted mice. More details, including search output and peaklists are available as a supplement.
6 226 52.1 13507644
170 20.0 14A_1511
Down
26782
Ppia or PPIA
Both
Peptidylprolyl isomerase A; cyclophilin A Carboxypeptidase N, polypeptide 1
6679439
18.1
5
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
45
probe-sets for Serpina3k and Serpina1b were observed, though not for Serpina1a, which gave very weak signals on the arrays. Three fold increases for complement component C3 were also observed. mRNA levels for the 1200017J02Rik protein were increased 1.5-fold, but the significance of the difference was low (p = .07). For other proteins that were increased in the sera of mice with tumors in the current study, the differences in mRNA levels were small. 6. Conclusions and future directions It is important to note that in the xenotransplant studies presented, an orthotopic model of lung carcinoma was utilized, in which cells were implanted into their tissue of origin. It is quite likely that implantation of tumor cells at different sites, such as in an artificial sub-cutaneous ectopic site, in the chest wall during implantation or in other organs as occurs in the formation of metastases, may result in the generation of different biomarker signatures. Protein expression in the host and tumor cells may also be influenced by the athymic status of the mice, which have no detectable cell-mediated immune function [17]. In all, it may be concluded that the vast majority of protein changes that occur in serum or plasma associated with tumor development do not represent tumor cell derived proteins, but represent secondary changes of host origin. However, it should be noted that as depth of analysis and sensitivity increases, an increased repertoire of tumor cell derived proteins should be expected. A good number of the host protein changes observed were previously reported as occurring in serum from cancer patients pointing to similarities between mouse and human. Of course the goal is not to confirm what has been previously discovered but to extend discovery to novel markers. In our current studies of a GEM model of pancreatic cancer, we have identified such candidate markers for pancreatic cancer in the mouse model and have demonstrated a similar pattern of expression of these markers in human pancreatic cancer. These findings lead us to conclude that mouse models of cancer have substantial utility in complementing similar studies of human cancer and encourage further effort to apply proteomics to mouse models of cancer not only for diagnostic biomarker identification but also for identifying critical targets for innovative therapies and for monitoring disease progression or response.
46
Table 2 Multiple identifications from the same digest Mw estimate (kDa)
Finding
Digest’s ID
Gene symbol
Species
Protein title
Accession number (gi)
Theoretical mass (kDa)
MASCOT score
# Distinct peptides
07A_1340 07A_1340 07A_1434 07A_1434 11A_1346 11A_1346 11A_1346 11A_1417 11A_1417 11A_1716 11A_1716 11A_1779 11A_1779 12A_1619 12A_1619
35.9 35.9 30.8 30.8 38.7 38.7 38.7 35.2 35.2 14.4 14.4 10.7 10.7 10.6 10.6
Up Up Up Up Up Up Up Up Up Up Up Up Up Up Up
26751 26751 26752 26752 26758 26758 26758 26759 26759 26760 26760 26761 26761 26763 26763
Kng1 Orm2 Pzp Hpxn C3 Apoa4 Pzp PzP Apoe Hbb Pzp Alb1 Apoe Apoa1 Apoh
Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse
12963497 387176 34785996 22022646 23956044 109575 34785996 34785996 192005 4760590 34785996 33859506 6753102 2145139 231558
49.0 16.0 167.1 31.5 187.9 44.5 167.1 167.1 33.2 15.8 167.1 70.7 35.9 30.5 39.9
174 172 172 138 582 558 458 713 675 386 235 594 498 190 133
5 5 4 3 17 15 13 16 16 7 5 12 11 4 3
14A_1425
26.8
Up
26778
Apoh
Mouse
231558
39.9
219
5
14A_1425 14A_1551
26.8 16.0
Up Up
26778 26784
Hp Hbb
Mouse Mouse
8850219 31982300
39.2 15.8
174 368
4 7
14A_1551
16.0
Up
26784
Serpina3k
Mouse
33859636
47.0
337
7
16A_1432 16A_1432
32.0 32.0
Up Up
26788 26788
PzP Mbl1
Mouse Mouse
6680608 6754654
167.1 25.8
332 330
9 8
16A_1499
28.6
Up
26789
Hbb
Mouse
31982300
15.8
366
7
16A_1499 17A_1604 17A_1604 17A_1636 17A_1636 19A_1804 19A_1804
28.6 10.1 10.1 7.3 7.3 5.7 5.7
Up Up Up Up Up Up Up
26789 26768 26768 26770 26770 26794 26794
Plg Pzp Apoa1 PzP Cxcl7 PzP Apoa1
Mouse Mouse Mouse Mouse Mouse Mouse Mouse
Kininogen 1 a-1-Acid glycoprotein Pzp protein Hemopexin Complement component 3 Apolipoprotein A-IV precursor Pzp protein Pzp protein Apolipoprotein E b-1-Globin Pzp protein Serum albumin precursor Apolipoprotein E Apolipoprotein A-I b-2-Glycoprotein I precursor (Apo-H) b-2-Glycoprotein I precursor(Apo-H) Haptoglobin Hemoglobin, b adult major chain Serine (or cysteine) proteinase inhibitor, clade A, member 3K Pregnancy zone protein Mannose binding lectin, liver (A) Hemoglobin, b adult major chain Plasminogen Pzp protein Apolipoprotein A-I Pzp protein Pro-platelet basic protein Pzp protein Apolipoprotein A-I
31982113 34785996 2145139 34785996 12963823 34785996 2145135
93.5 167.1 30.5 167.1 12.6 167.1 30.5
239 202 187 148 137 156 105
6 4 5 3 3 3 3
This table is organized similarly to Table 1 except that 2 or 3 rows correspond to multiple mass spectrometry search hits for single gel plugs. We present the results without claiming which protein is responsible for the observed intensity difference between sera of mice with and without A549 implants. See the supplement for more detailed search output and the peaklists.
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
Gel and band of interest
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
47
Table 3 RNA assays of mouse lung tissue adjacent to A549 xenografts (4 samples) and of normal mouse lung tissue (3 samples), using Affymetrix MOE_430A arrays [16] Gene symbol
Probe set ID
Serpina3k Serpina3k
1423866_at 1423867_at
Serpina1a
Means Adjacent tissue
Normal lung
t-test on logs
253 62
108 64
0.0057 0.8095
1420553_x_at
12
9
0.2368
Serpina1b Serpina1b
1418282_x_at 1451513_x_at
1792 1259
609 350
0.0377 0.0335
Apoa1 Apoa1 Apoa1 Apoa1
1419232_a_at 1419233_x_at 1438840_x_at 1455201_x_at
2 168 162 75
7 35 22 32
0.9489 0.1399 0.0620 0.2056
Apoa2
1417950_a_at
324
241
0.0056
Apoa4 Apoa4
1417761_at 1436504_x_at
95 362
99 485
1.0000 0.4302
Apoh
1416677_at
419
453
0.8928
Hp
1448881_at
20,677
25,311
0.0173
1300017J02Rik
1424722_at
531
340
0.0744
Cpn1
1417745_at
25
5
0.4050
F2
1418897_at
361
399
0.6662
Gsn Gsn Gsn Gsn Gsn Gsn
1415812_at 1436991_x_at 1437171_x_at 1456312_x_at 1456568_at 1456569_x_at
34,671 19,232 22,072 19,860 14 8554
32,099 18,571 22,435 21,757 28 8307
0.1424 0.6778 0.9697 0.7976 0.3427 0.7092
Pzp
1417246_at
174
242
0.2462
Hba-a1 Hba-a1 Hba-a1 Hba-a1
1417714_x_at 1428361_x_at 1452757_s_at 1453574_at
78,064 66,413 52,537 683
60,175 54,776 40,908 509
0.0001 0.0039 0.0203 0.1320
Ppia
1417451_a_at
27,644
32,631
0.0155
C3
1423954_at
24,777
7787
0.00003
P-values from two-sample t-tests on log-transformed data are given in the last column.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.canlet.2006.12.032. References [1] S.E. Ong, M. Mann, Mass spectrometry-based proteomics turns quantitative, Nat. Chem. Biol. 1 (2005) 252–262. [2] M. Bronstrup, Absolute quantification strategies in proteomics based on mass spectrometry, Expert Rev. Proteomics 1 (2004) 503–512.
[3] S. Hanash, Disease proteomics, Nature 422 (2003) 226–232. [4] L.F. Steel, B.B. Haab, S. Hanash, Methods of comparative proteomic profiling for disease diagnostics, J. Chromatogr. B: Analytical Technologies in the Biomedical & Life Sciences 815 (2005) 275–284. [5] J.T. Stults, D. Arnott, Proteomics, Meth. Enzymol. 402 (2005) 245–289. [6] D.F. Ransohoff, Bias as a threat to the validity of cancer molecular-marker research, Nat. Rev. Cancer 5 (2005) 142– 149. [7] N.E. Sharpless, R.A. Depinho, The mighty mouse: genetically engineered mouse models in cancer drug development, Nat. Rev. Drug Discov. 5 (2006) 741–754. [8] S.R. Hingorani, E.F. Petricoin, A. Maitra, V. Rajapakse, C. King, M.A. Jacobetz, S. Ross, T.P. Conrads, T.D. Veenstra,
48
[9]
[10]
[11] [12]
R. Kuick et al. / Cancer Letters 249 (2007) 40–48 B.A. Hitt, Y. Kawaguchi, D. Johann, L.A. Liotta, H.C. Crawford, M.E. Putt, T. Jacks, C.V. Wright, R.H. Hruban, A.M. Lowy, D.A. Tuveson, Preinvasive and invasive ductal pancreatic cancer and its early detection in the mouse, Cancer Cell 4 (2003) 437–450. A.J. Aguirre, N. Bardeesy, M. Sinha, L. Lopez, D.A. Tuveson, J. Horner, M.S. Redston, R.A. DePinho, Activated Kras and Ink4a/Arf deficiency cooperate to produce metastatic pancreatic ductal adenocarcinoma, Genes Dev. 17 (2003) 3112–3126. H. Wang, S.G. Clouthier, V. Galchev, D.E. Misek, U. Duffner, C.-K. Min, R. Zhao, J. Tra, G.S. Omenn, J.L. Ferrara, S.M. Hanash, Intact-protein based high-resolution three-dimensional quantitative analysis system for proteome profiling of biological fluids, Mol. Cell. Proteomics 4 (2005) 618–625. M. Abrahamson, M. Alvarez-Fernandez, C.M. Nathanson, Cystatins, Biochem. Soc. Symp (2003) 179–199. J. Kos, M. Krasovec, N. Cimerman, H.J. Nielsen, I.J. Christensen, N. Brunner, Cysteine proteinase inhibitors stefin A, stefin B, and cystatin C in sera from patients with colorectal cancer: relation to prognosis, Clin. Cancer Res. 6 (2000) 505–511.
[13] M. Ruse, A.M. Broome, R.L. Eckert, S100A7 (psoriasin) interacts with epidermal fatty acid binding protein and localizes in focal adhesion-like structures in cultured keratinocytes, J. Investig. Dermatol. 121 (2003) 132–141. [14] G. Hagens, I. Masouye, E. Augsburger, R. Hotz, J.H. Saurat, G. Siegenthaler, Calcium-binding protein S100A7 and epidermal-type fatty acid-binding protein are associated in the cytosol of human keratinocytes, Biochem. J. 339 (Pt 2) (1999) 419–427. [15] M.C. Brouard, J.H. Saurat, G. Ghanem, G. Siegenthaler, Urinary excretion of epidermal-type fatty acid-binding protein and S100A7 protein in patients with cutaneous melanoma, Melanoma Res. 12 (2002) 627–631. [16] C. Creighton, J.L. Bromberg-White, D.E. Misek, D.J. Monsma, F. Brichory, R. Kuick, T.J. Giordano, W. Gao, G. Omenn, C.P. Webb, S. Hanash, Analysis of tumor-host interactions by gene expression profiling of lung adenocarcinoma xenografts identifies genes involved in tumor formation, Mol. Cancer Res. (2005) 119–129. [17] S.V. Szapiel, N.A. Elson, J.D. Fulmer, G.W. Hunninghake, R.G. Crystal, Bleomycin-induced interstitial pulmonary disease in the nude, athymic mouse, Am. Rev. Respiratory Dis. 120 (1979) 893–899.
Mini-review
Discovery of cancer biomarkers through the use of mouse models Rork Kuick a, David E. Misek a, David J. Monsma b, Craig P. Webb b, Hong Wang c, Kelli J. Peterson d, Michael Pisano d, Gilbert S. Omenn e, Samir M. Hanash c,* b
a Department of Pediatrics, University of Michigan Medical Center, Ann Arbor, MI, USA Laboratory of Tumor Metastasis and Angiogenesis, Van Andel Research Institute, Grand Rapids, MI, USA c Fred Hutchinson Cancer Research Center, Seattle, WA, USA d Proteomic Research Services, Inc., Ann Arbor, MI, USA e Internal Medicine, University of Michigan Medical Center, Ann Arbor, MI, USA
Abstract Although our understanding of the molecular pathogenesis of common types of cancer has improved considerably, the development of effective strategies for cancer diagnosis and treatment have lagged behind. Mouse models of cancer potentially represent an efficient means for uncovering diagnostic markers as genetic alterations associated with human tumors can be engineered in mice. In addition, defined stages of tumor development, breeding conditions, and blood sampling can all be controlled and standardized to limit heterogeneity. Alternatively human cancer cells can be injected into mice and tumor development monitored in xenotransplants. Mouse-based studies promise to elucidate a repertoire of protein changes that occur in blood and biological fluids during tumor development. This is illustrated in a study in which we have applied a three-dimensional intact protein analysis system (IPAS) to elucidate detectable protein changes in serum from immunodeficient mice with lung xenografts from orthotopically implanted human A549 lung adenocarcinoma cells. With sufficiently detailed protein sequence identifications, the observed protein changes can be attributed to either the host mouse or the human tumor cells. It is noteworthy that the majority of increases identified have corresponded to relatively abundant serum proteins, some of which have previously been reported as increased in the sera of cancer patients. Proteomic studies of mouse models of cancer allow assessment of the range of changes in plasma proteins that occur with tumor development and may lead to the identification of potential cancer markers applicable to humans. Ó 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: Cancer biomarkers; Mouse models for cancer; Proteinases; Mass spectrometry; Serum proteome; Proteinase inhibitors
1. Introduction * Corresponding author. Tel.: +1 206 667 7091; fax: +1 206 667 2537. E-mail address: [email protected] (S.M. Hanash).
The quantitative analysis of proteomes has reached a fairly mature stage, with a rich repertoire of technologies available [1,2]. Recent improve-
0304-3835/$ - see front matter Ó 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.canlet.2006.12.032
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
ments in proteomic profiling methods permit screening biological fluids in order to discover protein markers that are potentially useful for diagnosis, prognosis, or monitoring of disease states such as cancer [3,4]. In particular, mass spectrometry (MS) has evolved enough to detect and identify femtomols of peptides, with a dynamic range over 3–5 orders of magnitude [5]. Quantitative methods currently available for biomarker discovery can be grouped into label-free methods, which compare pure peptide ion intensities between MS analysis, and labeling methods, which are applied to intact proteins or protein digests for comparative analysis. While technical challenges may now be overcome, a major impediment to biomarker discovery is the substantial heterogeneity in disease populations and the substantial challenges involved in designing and executing well controlled studies to discover promising biomarkers [6]. As a means to simplify the process, there is interest in evaluating mouse models of cancer for a variety of applications from diagnostics to therapeutics. 2. Mouse models of cancer Refined genetically engineered mouse (GEM) models of human cancer have been shown to faithfully recapitulate the molecular, biological and clinical features of human disease. GEM models afford defined stages of tumor development, homogenized breeding and environmental conditions, and standardized blood sampling, thereby limiting biological heterogeneity [7]. The concept that plasma from GEM models of cancer contains tumor derived proteins that may be relevant as candidate markers for human cancer is attractive. Only limited data has been obtained to date, including a SELDI study of a pancreatic cancer model [8]. In a recent study, we exploited the merits of another well-characterized GEM model of pancreatic cancer [9] and of an intensive quantitative proteomic approach we have developed to determine whether proteomics technologies allow identification of protein changes associated with tumor development and whether such changes are relevant to human pancreatic cancer. Plasma was sampled from mice at early and advanced stages of tumor development and from matched controls. Among the early and advanced tumor stages, we confidently identified proteins that were distributed across seven orders of magnitude of abundance in plasma. Comparative analysis of candidate bio-
41
markers documented striking concordance of expression in human and mouse pancreatic tissue and in the blood from patients with pancreatic cancer relative to normal specimens. In addition to identifying markers of potential utility for pancreatic cancer diagnosis, our findings indicated that GEM models of cancer, in combination with proteomics, provide a rich source of candidate markers applicable to human cancer. 3. Proteomic analysis of tumor xenotransplants While the potential of GEMs to uncover potential cancer markers is substantial, these models do not readily allow a determination of the tumor origin of the protein changes detectable in blood and biological fluids. On the other hand the use of xenotransplanted human tumor cells would allow a determination of host vs. tumor cell origin of protein changes observed. We have undertaken a study to determine the host vs. tumor cell origin of protein changes observed in mouse blood which we present and review here. In this regard, we have developed methods in which two or more protein samples are labeled with different Cy dyes and separated using an orthogonal, three-dimensional (3-D) intact-protein analysis system (IPAS) to measure protein differences between the samples [10]. Subsequently, mass spectrometry is used to analyze tryptic digests of the protein bands found to differ in order to identify peptides and deduce the proteins present. We have used a cancer model in which human A549 lung adenocarcinoma cells were orthotopically implanted into nude, athymic mice and their sera compared to mice undergoing a sham implantation procedure. We assessed quantitative and qualitative differences in protein expression due to proteins shed by the tumor itself from differences due to protein alterations in the tumor microenvironment or that result from systemic responses, such as acutephase inflammatory protein responses or antibodymediated immune responses. The primary goal was to identify the proteins causing observed differences in sufficient detail to permit determination of whether the source was the human tumor cells or the host mouse. Sera from 5 mice with intrapulmonary tumors and from 8 sham-operated mice were separately pooled. The three most abundant serum/plasma proteins (albumin, IgG, and transferrin) were removed by immunodepletion chromatography. The immunodepleted fractions were concentrated
42
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
and Cy5 and Cy3 dyes were used to label the control and implanted mouse sera, respectively. The labeled samples were combined and separated in three dimensions, based successively on pI, hydrophobicity, and molecular weight. The last dimension consisted of running SDS gels for each fraction. Following gel fixation, Cy3 and Cy5 images were quantitatively analyzed. We carefully examined spots with normalized ratios of spot integrated intensities between the two images of 3.0 or greater (either up or down). 4. Serum protein changes in human tumor bearing mice are predominantly of host origin We compared sera from mice with tumors formed from orthotopically implanted cells of the A549 human lung adenocarcinoma cell line with sera of control nude athymic mice that received a sham implantation procedure. Sera were collected seven weeks after surgery, at which point all five tumor burdened mice used in this study were shown to have extensive lung carcinoma (Fig. 1). Overlaying pairs of gel scans resulted in two-color images, such as that shown in Fig. 2. We selected 46 bands of most intense green color in these images, which corresponded to proteins that were more abundant in the sera of mice with tumors by at least 3-fold, in order to identify the protein responsible for the observed band differences by mass spectrometry. We also selected five red bands, which corresponded to proteins decreased in the tumor-bearing mice. In all, we obtained identifications implicating a single protein as most likely to be responsible for the observed image difference in a total of 30 cases (27 up, and 3 down in mice with tumors), which are summarized in Table 1. The 27 identifications in bands increased in tumors represented 16 different proteins.
We also obtained very confident MS search matches for 14 other bands (all up in mice with tumors), where however we obtained nearly equal quality matches to two or more distinct proteins. For these identifications, summarized in Table 2, we did not select one of the multiple identifications as more likely to be responsible for the differences observed between the sera, although such discrimination could be attempted on the basis of MS peak intensities and other criteria. Of the 41 bands increased in the sera of mice with tumors for which we obtained confident identifications, only two were for human proteins. The first of these was cystatin A (CSTA), a well-known cysteine protease inhibitor, which gave a 6.0-fold increase and was found close to its theoretical mass of 11 kDa on gel 12A (see Fig. 2) [11,12]. The second human protein identified was epidermal fatty acid binding protein 5 (FABP5, EFABP) from a relatively faint band that gave a 4.7-fold increase and was also found close to its 15 kDa theoretical mass [13–15]. The four peptides found to match our spectra were also contained in a hypothetical protein sequence predicted from the similar LOC387934 gene. It was clear from this analysis that by far the most bands found to have increased intensity in mice with tumors were from the host mice rather than the implanted human tumor cells. We found a single increased band containing Serpina1b, and several each for the Serpina1a and Serpina3k proteins, which are mouse paralogs of the human a-1antitrypsin (SERPINA1) and a-1-antichymotrypsin (SERPINA3) proteins. One of the bands found for each of these three proteins was far too low on the gel to be due to the well-known versions of these proteins, and so may have been cleaved by proteases expressed by the tumor cells. Besides these serpin protein elevations, several other proteins were identified for which previous
Fig. 1. Photographs of inflated lungs 7-weeks post sham injection (controls) or implantation of human A549 cells. In the latter mice, the lungs were virtually replaced by lung adenocarcinoma as verified by histopathological staining.
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
43
Fig. 2. Gel 12A from rotofor fraction 12, RP-HPLC fractions 13–39. Bands from approximately 4–200 kDa are visible. In this overlay of two scanned images of the gel, the image of the Cy3 scan is shown as green and the Cy5 scan as red, so that greener colors indicate increases in the sera of mice with A549 implanted cells. Labeled bands contained proteins identified by mass spectrometry as human cystatin A (band 1626), mouse apolipoprotein A-I and b-2-glycoprotein (Apo-H) (1619), mouse Serpina1b (1681), and mouse RIKEN cDNA 1300017J02, which has homology to transferrin (871). See details in Table 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)
reports of increases in cancer patients exist. We found increased anticoagulation factor 2 (F2, prothrombin). Considering the band location and peptide matches from MS searches, the particular band we observed (estimated to be 22 kDa) appears to arise from the mature thrombin heavy chain (theoretically 30 kDa). A band identified as haptoglobin (Hp) at an approximate molecular weight of 43 kDa, is most consistent with the mature b-haptoglobin. Although the mature b-haptoglobin is theoretically only 27 kDa in both human and mouse, it appears as a series of spots in 2D gels at approximately 40–45 kDa due to glycosylation or other modifications. We also identified the mouse pregnancy zone protein (Pzp), a member of the a-2-macroglobulin family, in one case where it was the best hit obtained by MS searches, and also for several other bands for which we did not obtain unique best hits. All of the bands for which Pzp identifications were obtained had much lower molecular weight than the complete Pzp monomer, which is
approximately 170 kDa. Similarly, at least 3, and perhaps all 5, of the increased bands identified as Apoa1 must be cleaved versions, as was a 9 kDa band identified as Apoa4, which is typically seen closer to its theoretically computed mass of 45 kDa. Other likely cleaved proteins identified included bands containing gelsolin (Gsn), and a hemoglobin. 5. Correlations between protein and RNA We previously published gene expression data for mouse tumor xenotransplants [16]. In order to determine if some of the increases for mouse proteins might be due to increased expression in surrounding stromal tissue rather than being acute-phase reactants produced from liver or other distant organs, we used our publicly available data to compare mRNA levels of mouse lung samples adjacent to tumors from implanted A549 cells to normal mouse lung samples. The results are shown in Table 3. Significantly, 2–3-fold increases in some
44
Table 1 Unambiguous protein identifications Mw estimate (kDa)
Finding
Digest’s ID
Gene symbol
Species
Protein title
03A_1831
12.4
Up
26742
FABP5
Human
12A_1626 05A_1194
9.9 58.2
Up Up
26764 05A_A1
CSTA Serpina3k
Human Mouse
05A_1241
52.1
Up
05A_B1
Serpina3k
Mouse
06A_1660
10.7
Up
26749
Serpina3k
Mouse
08A_1772 07A_1174 14A_1411 12A_1681 14A_1428 14A_1599 14A_1460 19A_1762 19A_1773 17A_1651 08A_1856 14A_992
15.7 48.9 26.2 7.5 26.8 10.4 24.6 9.8 8.0 6.5 9.4 53.2
Up Up Up Up Up Up Up Up Up Up Up Up
26754 26750 26777 12A_B1 26779 14A_A1 14A_C1 26792 26793 26771 26756 26774
Serpina1a Serpina1a Serpina1a Serpina1b Apoa1 Apoa1 Apoa1 Apoa1 Apoa1 Apoa2 Apoa4 Apoh
Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse
14A_1024
50.5
Up
26775
Apoh
Mouse
09A_1245 12A_871 14A_810 14A_1071
42.9 77.6 80.4 46.2
Up Up Up Up
09A_A1 26762 26773 26776
Hp 1300017J02Rik 1300017J02Rik Cpn1
Mouse Mouse Mouse Mouse
14A_1482
22.1
Up
26781
F2
Mouse
14A_1522 16A_1409 17A_1652 19A_1738 08A_1421
19.1 33.2 6.2 11.8 42.4
Up Up Up Up Down
26783 26787 26772 26791 26753
Gsn Pzp Hba Hba Hp
Mouse Mouse Mouse Mouse Mouse
Fatty acid binding protein 5 (psoriasis-associated) Cystatin A Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serine (or cysteine) proteinase inhibitor, clade A, member 3K Serpina1a protein Serpina1a protein Serpina1a protein Serpina1b protein Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-I Apolipoprotein A-II Apolipoprotein A-IV b-2-glycoprotein I precursor(Apo-H) b-2-glycoprotein I precursor(Apo-H) Haptoglobin RIKEN cDNA 1300017J02 RIKEN cDNA 1300017J02 Carboxypeptidase N, polypeptide 1 Coagulation factor II; prothrombin Gsn protein Pregnancy zone protein Hemoglobin a, adult chain 1 Hemoglobin a, adult chain 1 Haptoglobin
Accession number (gi)
Theoretical mass (kDa)
MASCOT score
# Distinct peptides
4557581
15.5
118
4
4885165 33859636
11.0 46.9
143 155
5 4
33859636
46.9
68
2
33859636
47.0
341
6
15929675 15929675 15929675 15277553 6753096 2145139 6753096 2145139 2145139 7304897 6680702 231558
45.8 45.8 45.8 45.1 30.6 30.5 30.6 30.5 30.5 11.4 44.7 39.9
540 1103 163 131 825 198 99 171 101 95 279 848
10 20 4 3 22 5 2 6 3 3 6 22
231558
39.9
569
14
8850219 18204720 18204720 13507644
39.2 78.7 78.7 52.1
112 1621 1983 171
3 37 41 4
6753798
71.7
373
9
18606238 6680608 6680175 6680175 8850219
81.0 167.1 15.1 15.1 39.2
172 187 169 376 637
4 6 6 8 14
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
Gel and band of interest
Mouse Cpn1 26786 Down 41.7 16A_1244
Excised bands found to be either increased (up) or decreased (down) at least 3-fold in mice with implanted A549 cells, were subjected to mass spectrometry and database searches using MASCOT software. One search resulted in 5 peptide matches which could be from either the human or mouse peptidylprolyl isomerase A, which is most likely from mouse since the band was more intense in samples from unimplanted mice. More details, including search output and peaklists are available as a supplement.
6 226 52.1 13507644
170 20.0 14A_1511
Down
26782
Ppia or PPIA
Both
Peptidylprolyl isomerase A; cyclophilin A Carboxypeptidase N, polypeptide 1
6679439
18.1
5
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
45
probe-sets for Serpina3k and Serpina1b were observed, though not for Serpina1a, which gave very weak signals on the arrays. Three fold increases for complement component C3 were also observed. mRNA levels for the 1200017J02Rik protein were increased 1.5-fold, but the significance of the difference was low (p = .07). For other proteins that were increased in the sera of mice with tumors in the current study, the differences in mRNA levels were small. 6. Conclusions and future directions It is important to note that in the xenotransplant studies presented, an orthotopic model of lung carcinoma was utilized, in which cells were implanted into their tissue of origin. It is quite likely that implantation of tumor cells at different sites, such as in an artificial sub-cutaneous ectopic site, in the chest wall during implantation or in other organs as occurs in the formation of metastases, may result in the generation of different biomarker signatures. Protein expression in the host and tumor cells may also be influenced by the athymic status of the mice, which have no detectable cell-mediated immune function [17]. In all, it may be concluded that the vast majority of protein changes that occur in serum or plasma associated with tumor development do not represent tumor cell derived proteins, but represent secondary changes of host origin. However, it should be noted that as depth of analysis and sensitivity increases, an increased repertoire of tumor cell derived proteins should be expected. A good number of the host protein changes observed were previously reported as occurring in serum from cancer patients pointing to similarities between mouse and human. Of course the goal is not to confirm what has been previously discovered but to extend discovery to novel markers. In our current studies of a GEM model of pancreatic cancer, we have identified such candidate markers for pancreatic cancer in the mouse model and have demonstrated a similar pattern of expression of these markers in human pancreatic cancer. These findings lead us to conclude that mouse models of cancer have substantial utility in complementing similar studies of human cancer and encourage further effort to apply proteomics to mouse models of cancer not only for diagnostic biomarker identification but also for identifying critical targets for innovative therapies and for monitoring disease progression or response.
46
Table 2 Multiple identifications from the same digest Mw estimate (kDa)
Finding
Digest’s ID
Gene symbol
Species
Protein title
Accession number (gi)
Theoretical mass (kDa)
MASCOT score
# Distinct peptides
07A_1340 07A_1340 07A_1434 07A_1434 11A_1346 11A_1346 11A_1346 11A_1417 11A_1417 11A_1716 11A_1716 11A_1779 11A_1779 12A_1619 12A_1619
35.9 35.9 30.8 30.8 38.7 38.7 38.7 35.2 35.2 14.4 14.4 10.7 10.7 10.6 10.6
Up Up Up Up Up Up Up Up Up Up Up Up Up Up Up
26751 26751 26752 26752 26758 26758 26758 26759 26759 26760 26760 26761 26761 26763 26763
Kng1 Orm2 Pzp Hpxn C3 Apoa4 Pzp PzP Apoe Hbb Pzp Alb1 Apoe Apoa1 Apoh
Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse Mouse
12963497 387176 34785996 22022646 23956044 109575 34785996 34785996 192005 4760590 34785996 33859506 6753102 2145139 231558
49.0 16.0 167.1 31.5 187.9 44.5 167.1 167.1 33.2 15.8 167.1 70.7 35.9 30.5 39.9
174 172 172 138 582 558 458 713 675 386 235 594 498 190 133
5 5 4 3 17 15 13 16 16 7 5 12 11 4 3
14A_1425
26.8
Up
26778
Apoh
Mouse
231558
39.9
219
5
14A_1425 14A_1551
26.8 16.0
Up Up
26778 26784
Hp Hbb
Mouse Mouse
8850219 31982300
39.2 15.8
174 368
4 7
14A_1551
16.0
Up
26784
Serpina3k
Mouse
33859636
47.0
337
7
16A_1432 16A_1432
32.0 32.0
Up Up
26788 26788
PzP Mbl1
Mouse Mouse
6680608 6754654
167.1 25.8
332 330
9 8
16A_1499
28.6
Up
26789
Hbb
Mouse
31982300
15.8
366
7
16A_1499 17A_1604 17A_1604 17A_1636 17A_1636 19A_1804 19A_1804
28.6 10.1 10.1 7.3 7.3 5.7 5.7
Up Up Up Up Up Up Up
26789 26768 26768 26770 26770 26794 26794
Plg Pzp Apoa1 PzP Cxcl7 PzP Apoa1
Mouse Mouse Mouse Mouse Mouse Mouse Mouse
Kininogen 1 a-1-Acid glycoprotein Pzp protein Hemopexin Complement component 3 Apolipoprotein A-IV precursor Pzp protein Pzp protein Apolipoprotein E b-1-Globin Pzp protein Serum albumin precursor Apolipoprotein E Apolipoprotein A-I b-2-Glycoprotein I precursor (Apo-H) b-2-Glycoprotein I precursor(Apo-H) Haptoglobin Hemoglobin, b adult major chain Serine (or cysteine) proteinase inhibitor, clade A, member 3K Pregnancy zone protein Mannose binding lectin, liver (A) Hemoglobin, b adult major chain Plasminogen Pzp protein Apolipoprotein A-I Pzp protein Pro-platelet basic protein Pzp protein Apolipoprotein A-I
31982113 34785996 2145139 34785996 12963823 34785996 2145135
93.5 167.1 30.5 167.1 12.6 167.1 30.5
239 202 187 148 137 156 105
6 4 5 3 3 3 3
This table is organized similarly to Table 1 except that 2 or 3 rows correspond to multiple mass spectrometry search hits for single gel plugs. We present the results without claiming which protein is responsible for the observed intensity difference between sera of mice with and without A549 implants. See the supplement for more detailed search output and the peaklists.
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
Gel and band of interest
R. Kuick et al. / Cancer Letters 249 (2007) 40–48
47
Table 3 RNA assays of mouse lung tissue adjacent to A549 xenografts (4 samples) and of normal mouse lung tissue (3 samples), using Affymetrix MOE_430A arrays [16] Gene symbol
Probe set ID
Serpina3k Serpina3k
1423866_at 1423867_at
Serpina1a
Means Adjacent tissue
Normal lung
t-test on logs
253 62
108 64
0.0057 0.8095
1420553_x_at
12
9
0.2368
Serpina1b Serpina1b
1418282_x_at 1451513_x_at
1792 1259
609 350
0.0377 0.0335
Apoa1 Apoa1 Apoa1 Apoa1
1419232_a_at 1419233_x_at 1438840_x_at 1455201_x_at
2 168 162 75
7 35 22 32
0.9489 0.1399 0.0620 0.2056
Apoa2
1417950_a_at
324
241
0.0056
Apoa4 Apoa4
1417761_at 1436504_x_at
95 362
99 485
1.0000 0.4302
Apoh
1416677_at
419
453
0.8928
Hp
1448881_at
20,677
25,311
0.0173
1300017J02Rik
1424722_at
531
340
0.0744
Cpn1
1417745_at
25
5
0.4050
F2
1418897_at
361
399
0.6662
Gsn Gsn Gsn Gsn Gsn Gsn
1415812_at 1436991_x_at 1437171_x_at 1456312_x_at 1456568_at 1456569_x_at
34,671 19,232 22,072 19,860 14 8554
32,099 18,571 22,435 21,757 28 8307
0.1424 0.6778 0.9697 0.7976 0.3427 0.7092
Pzp
1417246_at
174
242
0.2462
Hba-a1 Hba-a1 Hba-a1 Hba-a1
1417714_x_at 1428361_x_at 1452757_s_at 1453574_at
78,064 66,413 52,537 683
60,175 54,776 40,908 509
0.0001 0.0039 0.0203 0.1320
Ppia
1417451_a_at
27,644
32,631
0.0155
C3
1423954_at
24,777
7787
0.00003
P-values from two-sample t-tests on log-transformed data are given in the last column.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.canlet.2006.12.032. References [1] S.E. Ong, M. Mann, Mass spectrometry-based proteomics turns quantitative, Nat. Chem. Biol. 1 (2005) 252–262. [2] M. Bronstrup, Absolute quantification strategies in proteomics based on mass spectrometry, Expert Rev. Proteomics 1 (2004) 503–512.
[3] S. Hanash, Disease proteomics, Nature 422 (2003) 226–232. [4] L.F. Steel, B.B. Haab, S. Hanash, Methods of comparative proteomic profiling for disease diagnostics, J. Chromatogr. B: Analytical Technologies in the Biomedical & Life Sciences 815 (2005) 275–284. [5] J.T. Stults, D. Arnott, Proteomics, Meth. Enzymol. 402 (2005) 245–289. [6] D.F. Ransohoff, Bias as a threat to the validity of cancer molecular-marker research, Nat. Rev. Cancer 5 (2005) 142– 149. [7] N.E. Sharpless, R.A. Depinho, The mighty mouse: genetically engineered mouse models in cancer drug development, Nat. Rev. Drug Discov. 5 (2006) 741–754. [8] S.R. Hingorani, E.F. Petricoin, A. Maitra, V. Rajapakse, C. King, M.A. Jacobetz, S. Ross, T.P. Conrads, T.D. Veenstra,
48
[9]
[10]
[11] [12]
R. Kuick et al. / Cancer Letters 249 (2007) 40–48 B.A. Hitt, Y. Kawaguchi, D. Johann, L.A. Liotta, H.C. Crawford, M.E. Putt, T. Jacks, C.V. Wright, R.H. Hruban, A.M. Lowy, D.A. Tuveson, Preinvasive and invasive ductal pancreatic cancer and its early detection in the mouse, Cancer Cell 4 (2003) 437–450. A.J. Aguirre, N. Bardeesy, M. Sinha, L. Lopez, D.A. Tuveson, J. Horner, M.S. Redston, R.A. DePinho, Activated Kras and Ink4a/Arf deficiency cooperate to produce metastatic pancreatic ductal adenocarcinoma, Genes Dev. 17 (2003) 3112–3126. H. Wang, S.G. Clouthier, V. Galchev, D.E. Misek, U. Duffner, C.-K. Min, R. Zhao, J. Tra, G.S. Omenn, J.L. Ferrara, S.M. Hanash, Intact-protein based high-resolution three-dimensional quantitative analysis system for proteome profiling of biological fluids, Mol. Cell. Proteomics 4 (2005) 618–625. M. Abrahamson, M. Alvarez-Fernandez, C.M. Nathanson, Cystatins, Biochem. Soc. Symp (2003) 179–199. J. Kos, M. Krasovec, N. Cimerman, H.J. Nielsen, I.J. Christensen, N. Brunner, Cysteine proteinase inhibitors stefin A, stefin B, and cystatin C in sera from patients with colorectal cancer: relation to prognosis, Clin. Cancer Res. 6 (2000) 505–511.
[13] M. Ruse, A.M. Broome, R.L. Eckert, S100A7 (psoriasin) interacts with epidermal fatty acid binding protein and localizes in focal adhesion-like structures in cultured keratinocytes, J. Investig. Dermatol. 121 (2003) 132–141. [14] G. Hagens, I. Masouye, E. Augsburger, R. Hotz, J.H. Saurat, G. Siegenthaler, Calcium-binding protein S100A7 and epidermal-type fatty acid-binding protein are associated in the cytosol of human keratinocytes, Biochem. J. 339 (Pt 2) (1999) 419–427. [15] M.C. Brouard, J.H. Saurat, G. Ghanem, G. Siegenthaler, Urinary excretion of epidermal-type fatty acid-binding protein and S100A7 protein in patients with cutaneous melanoma, Melanoma Res. 12 (2002) 627–631. [16] C. Creighton, J.L. Bromberg-White, D.E. Misek, D.J. Monsma, F. Brichory, R. Kuick, T.J. Giordano, W. Gao, G. Omenn, C.P. Webb, S. Hanash, Analysis of tumor-host interactions by gene expression profiling of lung adenocarcinoma xenografts identifies genes involved in tumor formation, Mol. Cancer Res. (2005) 119–129. [17] S.V. Szapiel, N.A. Elson, J.D. Fulmer, G.W. Hunninghake, R.G. Crystal, Bleomycin-induced interstitial pulmonary disease in the nude, athymic mouse, Am. Rev. Respiratory Dis. 120 (1979) 893–899.