Adenocarcinoma of esophageal (EAC), gastric (GAC), and colorectal carcinoma origin (CRC) are morphologically similar because of the upper gastrointestinal tumors (GITs) frequently arise through the sequence intestinal metaplasia→dysplasia→carcinoma, in the context of Barrett esophagus or chronic atrophic gastritis.1 − /CK20+ , tumors originating in the appendix and right colon frequently express CK7.2–6 7 8 9
MATERIALS AND METHODS
After approval by the Institutional Review Board (study # VAM-19-00373), immunostains for CK7 (catalog# CM061B; Biocare Medical, Concord, CA; dilution: 1:150, retrieval: Enzyme), CK20 (catalog# BSB5390; Cancer Diagnostics Inc., Santa Barbara, CA; dilution: 1:100, retrieval: EDTA), CDX2 (catalog# 235R-14; Cell Marque, Rocklin, CA; dilution: 1:300, retrieval: EDTA), SATB2 (catalog# 384R-15; Cell Marque; dilution: 1:400, retrieval: EDTA), DcR3 (catalog# ab8405; Abcam, Cambridge, MA; dilution: 1:1000, retrieval: EDTA), MUC5AC (catalog# 292M-94; Cell Marque; dilution: 1:500, retrieval: EDTA), and MUC6 (catalog# 293M-95; Cell Marque; dilution: 1:700, retrieval: EDTA), were applied to selected slides with large tumoral areas from resection specimens from 40 CRC, including 10 appendiceal, 10 right, and 20 left-sided tumors, 40 GAC, and 40 EAC. IHC stains were performed on a Leica BOND-III automated stainer (Leica Biosystems, Melbourne, Australia). CDX2 and SATB2 were considered positive only if nuclear staining was present, the remaining markers if cytoplasmic staining was present. The intensity of expression was graded as: 0=negative, 1=weak, 2=moderate, and 3=strong, relative to control cells in each preparation. Area of expression was graded as: 0=negative, 1=1% to 5%, 2=6% to 50%, 3=>50%. A semiquantitative staining score defined as the addition of the intensity and area expression [staining score (0 to 6)=area score (0 to 3)+intensity score (0 to 3)] was calculated for each marker as previously described by Fedchenko et al.10 https://links.lww.com/AIMM/A251
Data collection and statistical analysis were performed with Microsoft Excel 2016 and R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (www.R-project.org/ U test to compare 2 independent samples because this variable was not expected to have a normal distribution.
RESULTS
Individual markers analyzed as dichotomous variables (+ or −) showed a preferential expression of SATB2 (sensitivity 85%, specificity 68%, PPV 57%, NPV 90%) and DCR3 (sensitivity 98%, specificity 39%, PPV 44%, NPV 97%) in CRC, and preferential expression of CK7 (sensitivity 86%, specificity 80%, PPV 90%, NPV 74%), MUC5AC (sensitivity 75%, specificity 63%, PPV 80%, PPV 56%) and MUC6 (sensitivity 69%, specificity 85%, PPV 90%, NPV 58%) in upper GIT. Expression of CDX2 and CK20 lacked specificity and was common at both locations: all CRC expressed both markers, and in the upper GI: CDX2 and CK20 were expressed in 85% and 79% of esophageal and gastric cases respectively (Table 1 ).
TABLE 1 -
Operating Characteristics of Individual Markers by Site as Dichotomous (+/−) Variables
Markers
Sensitivity (%)
95% CI
Specificity (%)
95% CI
PPV (%)
95% CI
NPV (%)
95% CI
CRC (N=40)
SATB2
85
0.69-0.94
68
0.56-0.77
57
0.43-0.69
90
0.78-0.96
DCR3
98
0.85-0.99
39
0.28-0.50
44
0.34-0.55
97
0.82-0.99
EAC/GAC (N=80)
CK7
86
0.76-0.93
80
0.63-0.90
90
0.80-0.95
74
0.58-0.86
MUC5AC
75
0.64-0.84
62
0.46-0.77
80
0.69-0.88
56
0.40-0.70
MUC6
69
0.57-0.78
85
0.69-0.94
90
0.79-0.96
58
0.44-0.70
Sensitivity [% (95% CI)]
Specificity [% (95% CI)]
PPV [% (95% CI)]
NPV [% (95% CI)]
Marker
EAC/GAC
CRC
EAC/GAC
CRC
EAC/GAC
CRC
EAC/GAC
CRC
Upper and lower
CDX2
85 (0.75-0.92)
100 (0.89-1)
0 (0-0.11)
15 (0.08-0.25)
63 (0.53-0.72)
37 (0.28-0.47)
0 (0-0.3)
100 (0.70-1)
CK20
79 (0.68-0.87)
100 (0.89-1)
0 (0-0.11)
21 (0.13-0.32)
61 (0.51-0.70)
39 (0.29-0.49)
0 (0-0.23)
100 (0.77-1)
CI indicates confidence interval; CRC, colorectal adenocarcinoma; EAC, esophageal adenocarcinoma; GAC, gastric adenocarcinoma; NPV, negative predictive value; PPV, positive predictive value.
For CRC, comparison of new and traditional marker combinations analyzed as dichotomous variables (+ or −), showed that in general combinations of SATB2 and DCR3 with CK7 result in a minor increase in specificity (90% to 96%) at the expense of sensitivity (80% to 68%), and combinations with CDX2 or CK20 increase sensitivity (80% to 90%) but specificity decreased greatly (90% to 43%) when compared with the traditional combination of CK7, CK20, and CDX2 (Table 2 , CRC).
TABLE 2 -
Operating Characteristics of Best Gastrointestinal Specific Combinations by Site as Dichotomous (+/−) Variables
Markers
Sensitivity (%)
95% CI
Specificity (%)
95% CI
PPV (%)
95% CI
NPV (%)
95% CI
CRC
New combinations
CK7− /SATB2+
68
0.51-0.81
96
0.89-0.99
90
0.72-0.97
86
0.76-0.92
CK7− /CDX2+ /SATB2+
68
0.51-0.81
96
0.89-0.99
90
0.72-0.97
86
0.76-0.92
CK7− /DCR3+
73
0.56-0.85
93
0.84-0.97
83
0.66-0.93
87
0.78-0.93
CDX2+ /SATB2+
85
0.69-0.94
69
0.57-0.78
58
0.44-0.70
90
0.79-0.96
CK20+ /SATB2+
83
0.67-0.92
74
0.63-0.83
61
0.47-0.74
91
0.79-0.95
CDX2+ /DCR3+
90
0.75-0.97
43
0.32-0.54
44
0.33-0.55
90
0.74-0.97
CK20+ /DCR3+
88
0.72-0.95
53
0.41-0.64
48
0.36-0.60
91
0.76-0.96
Traditional combinations
CK7− /CK20+
80
0.64-0.90
89
0.79-0.94
78
0.62-0.89
89
0.81-0.95
CK7− /CDX2+
80
0.64-0.90
90
0.81-0.95
80
0.64-0.90
90
0.81-0.95
CK7− /CK20+ /CDX2+
80
0.64-0.90
93
0.84-0.97
84
0.68-0.93
90
0.81-0.95
EAC/GAC
New combinations
CDX2+ /SATB2−
54
0.42-0.65
88
0.72-0.95
90
0.77-0.96
49
0.37-0.61
CK7+ /CDX2+ /SATB2−
48
0.36-0.59
100
0.89-1
100
0.89-1
49
0.38-0.60
CK7+ /MUC5AC+
69
0.57-0.78
90
0.75-0.97
93
0.82-0.98
59
0.46-0.71
CK7+ /MUC6+
64
0.52-0.74
95
0.82-0.99
96
0.86-0.99
57
0.44-0.69
MUC5AC+ /SATB2−
55
0.44-0.66
90
0.75-0.97
92
0.79-0.97
50
0.38-0.62
MUC6+ /SATB2−
49
0.38-0.60
98
0.85-1
98
0.85-1
49
0.38-0.60
Traditional combinations
CK7+ /CDX2+
75
0.64-0.84
83
0.67-0.92
90
0.79-0.95
62
0.48-0.75
CK7+ /CK20+
68
0.56-0.77
80
0.64-0.90
87
0.76-0.94
55
0.42-0.68
CI indicates confidence interval; CRC, colorectal adenocarcinoma; EAC, esophageal adenocarcinoma; GAC, gastric adenocarcinoma; NPV, negative predictive value; PPV, positive predictive value.
For upper GIT, combinations of SATB2, MUC5AC, and MUC6 analyzed as dichotomous variables (+ or −), increase specificity (81% to 93%) at the expense of sensitivity (72% to 56%) when compared with the traditional combination of CK7, CK20, and CDX2 (Table 2 , EAC/GAC).
Analysis of the staining score (area+intensity) using a nonparametric test of independence (Wilcoxon Mann-Whiney U test) showed significant differences by site for all markers, however, the largest differences in mean and modal expression of these markers were documented for CK7 and SATB2 in the upper and lower GIT, respectively (Table 3 ). SATB2 expression in lower GIT is usually high: staining score mean=4.5, mode=6, and in upper tumors is low: staining score mean=1, mode=0 (Fig. 1 ). CK7 expression in upper GIT is high: staining score mean=4.8, mode=6, and in lower tumors low: mean=1, mode=0. CDX2 was highly expressed at both sites: upper GIT staining score: mean=4.1, mode=5, lower GIT: staining score mean=5.7, mode=6. In summary, the markers showing the largest mean and modal differences in expression by site were CK7 for upper GIT and SATB2 for lower GIT. Although SATB2 is highly specific for CRC, CK7 lacks specificity for the GIT and for this reason it was combined with CDX2, which was the marker with the highest sensitivity for both sites. The frequencies of the 9 possible outcomes of this panel interpreted as dichotomous (+/−) variables are shown in Table 4 ; the most frequent phenotype of upper GIT in aggregate was: CK7+ /CDX2+ /SATB2− accounting for 48% of all cases. For lower GIT the most frequent phenotype was CK7− /CDX2+ /SATB2+ accounting for 68% of cases. The phenotype CK7+ /CDX2+ /SATB2+ was present in 27% of the upper and 18% of lower tumors. Other phenotypes were less common. The application of this panel to the 101 biopsies showed similar results (Table 5 ). The phenotype CK7+ /CDX2+ /SATB2− was 3 times more common in upper than lower GIT and PAC, and not observed in LAC or CHC. The phenotype CK7+ /CDX2+ /SATB2+ was twice as common in upper than lower GIT, and not observed in PAC, CHC, or LAC. The phenotype CK7− /CDX2+ /SATB2+ was 12 times more common in lower than upper GIT, and not observed in PAC, CHC, or LAC. The nonspecific phenotype CK7+ /CDX2− /SATB2− was 4 to 5 times more common in CHC, PAC, and LAC, than in gastric tumors and very rare (<5%) in esophageal and CRC. The nonspecific phenotype CK7− /CDX2− /SATB2− was most common in pancreatic tumors.
TABLE 3 -
Wilcoxon Mann-Whiney
U Test
SS (Mean±SD)
Mode
Markers
CRC
EAC/GAC
CRC
EAC/GAC
P
Mean Difference
Mode Difference
Lower GI
SATB2
4.5±2.2
1.0±1.5
6
0
<0.0001
3.5
6
DCR3
4.8±1.8
3.1±2.6
6
0
0.0001
1.8
6
Upper GI
CK7
1.0±2.0
4.8±2.0
0
6
<0.0001
−3.9
−6
MUC5AC
1.7±2.4
3.6±2.2
0
5
0.0004
−1.9
−5
MUC6
0.5±1.2
2.9±2.1
0
4
<0.0001
−2.4
−4
Upper and lower GI
CK20
5.5±0.8
3.7±2.1
6
5
<0.0001
1.7
1
CDX2
5.7±1.0
4.1±2.0
6
5
<0.0001
1.7
1
SS of individual markers by the site.
SS: 0=no expression, ≤3=low expression, 4=moderate expression, >4=high expression.
Result is significant at P <0.05.
CRC indicates colorectal adenocarcinoma; EAC, esophageal adenocarcinoma; GAC, gastric adenocarcinoma; GI, gastrointestinal; SS, staining score=intensity+% area.
FIGURE 1: Examples of CK7/CDX2/SATB2 panel by the site. The first row shows examples of the expression of CK7 at different sites. Foregut derived tumors are usually positive, except for gastric tumors which were negative in >20% of the cases. The second row shows an expression of CDX2, which is frequently expressed by upper and lower gastrointestinal tumors, one third of pancreatic tumors, but is infrequent in cholangiocarcinoma and lung primaries. The third row shows the expression of SATB2. The strong and diffuse expression is typical of colorectal tumors, but the weak or focal expression is common at other sites (all micrographs were taken at ×40, except the bottom right pictures that were taken at ×400).
TABLE 4 -
Frequency of All Possible Combinations of CK7, CDX2, and SATB2
Phenotypes
CRC (N=40) (%)
EAC/GAC (N=80) (%)
OR
Additional Markers
CK7− /CDX2+ /SATB2+
68
4
18 LGI
CK7+ /CDX2+ /SATB2−
0
48
UGI
CK7+ /CDX2+ /SATB2+
18
27
1.6 UGI
MUC6+ : OR=2.3 for UGI
CK7− /CDX2+ /SATB2−
12
6
2 LGI
No change with more markers
CK7+ /CDX2− /SATB2+
0
1
UGI
CK7− /CDX2− /SATB2−
0
4
NS
CK20+ or MUC5AC+ =UGI
CK7+ /CDX2− /SATB2−
2
9
NS
CK20+ or MUC5AC+ =75% UGI
CK7− /CDX2− /SATB2+
0
0
—
CK7+ /CDX2− /SATB2+
0
1
UGI
CRC indicates colorectal carcinoma; EAC, esophageal adenocarcinoma; GAC, gastric adenocarcinoma; GI, gastrointestinal; LGI, lower gastrointestinal; NS, not specific for gastrointestinal; OR, odds ratio; UGI, upper gastrointestinal.
TABLE 5 -
CK7/CDX2/SATB2 Panel in Colorectal and Foregut Adenocarcinomas
n (%)
Esophageal
Gastric
Colorectal
Pancreatic
Cholangiocarcinoma
Lung
Phenotypes
N=17
N=57
N=17
N=57
N=19
N=57
N=18
N=15
N=15
CK7+ /CDX2+ /SATB2+
8 (47)
25 (44)
4 (23)
9 (16)
3 (16)
10 (17)
0
0
0
CK7+ /CDX2+ /SATB2−
7 (41)
28 (49)
5 (29)
22 (39)
2 (11)
2 (3)
5 (28)
0
0
CK7− /CDX2+ /SATB2+
0
0
3 (18
6 (10)
12 (63)
39 (66)
0
0
0
CK7− /CDX2+ /SATB2−
0
0
2 (12)
7 (12)
0
5 (8)
0
0
0
CK7− /CDX2− /SATB2+
0
0
0
0
0
0
0
0
0
CK7+ /CDX2− /SATB2+
0
1 (2)
0
0
0
0
0
4 (27)
5 (33)
CK7+ /CDX2− /SATB2−
2 (12)
3 (5)
3 (18)
10 (18)
0
1 (20)
11 (61)
11 (73)
10 (67)
CK7− /CDX2− /SATB2−
0
0
0
3 (5)
2 (10)
2 (3)
2 (11)
0
0
Phenotypes
Esophageal (%)
Gastric (%)
Colorectal (%)
Pancreatic (%)
Cholangiocarcinoma (%)
Lung (%)
CK7+ /CDX2+ /SATB2+
47
17
19
CK7+ /CDX2+ /SATB2−
53
41
28
CK7− /CDX2+ /SATB2+
10
73
CK7− /CDX2+ /SATB2−
13
9
CK7+ /CDX2− /SATB2+
27
33
CK7+ /CDX2− /SATB2−
19
61
73
67
CK7− /CDX2− /SATB2−
11
In the table at the top, the results of the biopsies (left) and the aggregated data of biopsy and resections specimens (right) are shown for esophageal, gastric and colorectal adenocarcinomas. In the table at the bottom, the phenotypes show only aggregated data after excluding phenotypes with frequencies ranging from 0% to ≤5% (black boxes).
CRC indicates colorectal adenocarcinoma; EAC, esophageal adenocarcinoma; GAC, gastric adenocarcinoma; SS, staining score=intensity+% area.
Tabulating the aggregated data from resection and biopsy specimens for upper and lower GIT showed similar results (Table 5 , top). The phenotype CK7+ /CDX2+ was present in 93% EAC, 54% GAC, and 20% CRC. The phenotype CK7− /CDX2+ was present in 74% CRC, 21% GAC, and 0% EAC. Excluding phenotypes with frequencies ≤5% (Table 5 , bottom) showed that with this panel all tumors by the site have relatively homogenous phenotypes, except gastric tumors which show greater phenotypic variability (Fig. 2 ).
FIGURE 2: Frequency (%) of phenotypes by primary site. The figure shows the results of aggregated data after excluding very infrequent phenotypes (≤5%). The upper 3 phenotypes are most frequent in pancreatobiliary and lung adenocarcinomas. The remaining phenotypes are more characteristics of tumors arising in the esophagus, stomach, and colon/rectum, with some phenotypes strongly favoring an upper or lower gastrointestinal (GI) origin. Expression of SATB2 in colorectal tumors is usually strong, and in the other tumors is usually weak or focal.
DISCUSSION
EAC, GAC, and CRC frequently show overlapping morphology because the upper GIT typically evolve from pathologies involving intestinal metaplasia.1 5,11,12
In our study, the marker with the overall best operating characteristics and largest differences in expression for upper GIT was CK7, however, in isolation, this marker lacks specificity, being expressed by most normal mucosal epithelia, mesothelium, genitourinary, and Müllerian-derived tissues, endocrine organs, and neoplasms derived from these tissues. It is also expressed by a subset of renal neoplasms, and about one fourth of nonkeratinizing squamous cell carcinomas.13 + /CDX2+ with a sensitivity of 93% and 54% for EAC and GAC, respectively, but was also present in 20% of randomly selected CRC biopsies.
The marker with the overall best operating characteristics and the largest differences in expression for lower GIT was SATB2. In contrast to CK7, this marker in isolation is both sensitive and specific for CRC.7,11 14
We combined the markers with the best individual operating characteristics for segregating upper from lower GIT (CK7 and SATB2) with the marker with the highest sensitivity for both sites (CDX2) as a panel and it applied to a set of biopsies of tumors derived from the foregut and large intestine. CK7− /CDX2+ /SATB2+ was moderately sensitive (66%) and very specific (96%) for CRC. The phenotype CK7+ /CDX2+ showed good sensitivity for upper GIT: EAC 93%, GAC 54%, overall 74%; adding SATB2 increased specificity from 83% to nearly 100% if negative (ie, CK7+ /CDX2+ /SATB2− ). However, a positive result (ie, CK7+ /CDX2+ /SATB2+ ) was common at both sites (UGI: 30%, LGI 17%). The results in resection specimens translated well to biopsies despite the much smaller tumoral areas available for evaluation. Pancreatobiliary and LAC are most commonly CK7+ /CDX2− /SATB2− (67%), however, one third of CHC and LAC showed weak and/or focal SATB2 expression (ie, CK7+ /CDX2− /SATB2+ ) but this phenotype was very uncommon in upper GIT (1%) and not observed in CRC. Similarly, about one third of pancreatic tumors were positive for CDX2 (ie, CK7+ /CDX2+ /SATB2− ), overlapping with upper GIT. This phenotype has also been reported in most intestinal neuroendocrine tumors.14 15 16
CONCLUSIONS
Among the recently developed markers for GIT, SATB2 added to CK7 and CDX2 provides high-discriminative power for differentiating upper from lower GIT and tumors derived from the foregut. The phenotype CK7+ /CDX2+ /SATB2− is highly specific of upper GIT. The phenotype CK7− /CDX2+ /SATB2+ is highly specific for lower GIT. The phenotypes CK7+ /CDX2− /SATB2− and CK7+ /CDX2− /SATB2+ favor a non-GI (eg, pancreatobiliary or lung) primary. The use of this panel enhances diagnostic accuracy compared with the traditional CK7/CK20/CDX2 panel. Less frequent phenotypes show substantial overlap, however, using the frequencies derived from our study may help optimize subsequent diagnostic steps. An important caveat about SATB2 is that while strong diffuse expression is characteristic of CRC, weak and/or focal expression is present in one third or more of upper GI, CHC, and LAC. DcR3, MUC5AC and MUC6 improve specificity at the expense of sensitivity and may be used as second-tier markers.
REFERENCES
1. Bosman FT, Carneiro F, Hruban RH, et al. WHO Classification of Tumours of the Digestive System. Lyon, France: IARC; 2010:25–31.
2. Werling RW, Yaziji H, Bacchi CE, et al. CDX2, a highly sensitive and specific marker of adenocarcinomas of intestinal origin: an immunohistochemical survey of 476 primary and metastatic carcinomas. Am J Surg Pathol. 2003;27:303–310.
3. Kaimaktchiev V, Terracciano L, Tornillo L, et al. The homeobox intestinal differentiation factor CDX2 is selectively expressed in gastrointestinal adenocarcinomas. Mod Pathol. 2004;17:1392–1399.
4. Kende AI, Carr NJ, Sobin LH. Expression of cytokeratins 7 and 20 in carcinomas of the gastrointestinal tract. Histopathology. 2003;42:137–140.
5. Gurzu S, Jung I. Aberrant pattern of the cytokeratin 7/cytokeratin 20 immunophenotype in colorectal adenocarcinomas with BRAF mutations. Pathol Res Pract. 2012;208:163–166.
6. Gurzu S, Szentirmay Z, Toth E, et al. Serrated pathway adenocarcinomas: molecular and immunohistochemical insights into their recognition. PLoS One. 2013;8:e57699.
7. Dragomir A, de Wit M, Johansson C, et al. The role of SATB2 as a diagnostic marker for tumors of colorectal origin. Am J Clin Pathol. 2014;141:630–638.
8. Uhlén M, Fagerberg L, Hallström BM, et al. Proteomics. Tissue-based map of the human proteome. Science. 2015;347:1260419.
9. Lee HK, Kim HS, Yang HK, et al. MUC1, MUC2, MUC5AC, and MUC6 expressions in gastric carcinomas. Their roles as prognostic indicators. Cancer. 2001;92:1427–1434.
10. Fedchenko N, Reifenrath J. Different approaches for interpretation and reporting of immunohistochemistry analysis results in the bone tissue—a review. Diagn Pathol. 2014;9:221.
11. Mesa H, Manivel JC, Larson W, et al. Immunophenotypic comparison of neoplasms of the appendix, right and left colon; in search of a site-specific phenotypic signature. Int J Surg Pathol. 2020;28:20–30.
12. Bayrak R, Yenidünya S, Haltas H. Cytokeratin 7 and cytokeratin 20 expression in colorectal adenocarcinomas. Pathol Res Pact. 2011;207:156–160.
13. Chu P, Wu E, Weiss LM. Cytokeratin 7 and cytokeratin 20 expression in epithelial neoplasms: a survey of 435 cases. Mod Pathol. 2000;13:962–972.
14. Moskaluk CA, Zhang H, Powell SM, et al. Cdx2 protein expression in normal and malignant human tissues: an immunohistochemical survey using tissue microarrays. Mod Pathol. 2003;16:913–919.
15. Aulakh KS, Chisholm CD, Smith DA, et al. TTF-1 and napsin A do not differentiate metastatic lung adenocarcinomas from primary esophageal adenocarcinomas: proposal of a novel staining panel. Arch Pathol Lab Med. 2013;137:1094–1098.
16. Mazziotta RM, Borczuk AC, Powell CA, et al. CDX2 immunostaining as a gastrointestinal marker: expression in lung carcinomas is a potential pitfall. Appl Immunohistochem Mol Morphol. 2005;13:55–60.
