Journal of Surgical Research 163, 47–51 (2010)doi:10.1016/j.jss.2010.04.018

ASSOCIATION FOR ACADEMIC SURGERY

Lack of NKX2.2 Expression in Bronchopulmonary Typical Carcinoid Tumors:

Implications for Patients with Neuroendocrine Tumor Metastases

and Unknown Primary Site

Yu-cheng Wang, Ph.D.,*,† Gioia Iezza, M.D.,*,‡ Marlene B. Zuraek, M.D.,*,† David M. Jablons, M.D.,*,†Pierre R. Theodore, M.D.,*,† Emily K. Bergsland, M.D.,*,§ David B. Donner, Ph.D.,*,† Robert S. Warren, M.D.,*,†

and Eric K. Nakakura, M.D., Ph.D.*,†,1

*UCSF Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, California; †Department of Surgery;‡Department of Pathology; and §Department of Medicine, University of California, San Francisco, California

Submitted for publication January 5, 2010

Background. Patients with neuroendocrine tumors(NETs) may have metastatic disease and unknown pri-mary site. NETs commonly arise from the bronchopul-monary (BP) and gastrointestinal (GI) tract. Thelargest subgroups of well-differentiated BP-NETs aretypical carcinoids (TCs). The homeodomain transcrip-tion factor NKX2.2 regulates development of gut sero-tonin cells and is a marker of GI-NETs. Previous workon a limited number of samples suggested thatBP-TCs do not express NKX2.2. We hypothesized thatlack of NKX2.2 expression in BP-TCs might be usefulto distinguish BP- from GI-NETs, and evaluatedNKX2.2 expression in a larger number of BP-TCs.

Methods. Archived formalin-fixed, paraffin-embed-ded tissues were obtained from 13 previously unde-scribed patients with BP-TCs. Expression of NKX2.2,serotonin, and the NE marker chromogranin A (CgA)were assessed by immunohistochemistry.

Results. CgA expression was robust in all 13 BP-TCs,confirming the NE phenotype. Serotonin expressionwas less frequent (9/13; 69%). Two patients withBP-TCs in which serotonin expression was absent ex-hibited Cushing’s syndrome due to ectopic ACTH pro-duction. NKX2.2 expression was not observed in anyof the 13 tumors.

Conclusions. Bronchopulmonary TCs uniformly ex-press CgA but not NKX2.2. Because most of these tu-mors express serotonin, our findings suggest thatNKX2.2 may not be required for serotonin production

1 To whom correspondence and reprint requests should beaddressed at Department of Surgery, Division of Surgical Oncology,UCSF Helen Diller Family Comprehensive Cancer Center, 1600Divisadero Street, A-724, San Francisco, CA 94143-1932. E-mail:Eric.Nakakura@ucsfmedctr.org.

47

by BP-TCs. We conclude that the presence or absenceof NKX2.2 expression may assist in the determinationof the primary tumor site in patients with NET metas-tases of unknown origin. NET metastases that areCgA-positive/NKX2.2-negative would suggest a BP pri-mary, whereas those that are CgA-positive/NKX2.2-positive would suggest a GI primary. � 2010 Elsevier Inc. All

rights reserved.

Key Words: bronchopulmonary; typical carcinoid;neuroendocrine tumor; NKX2.2.

INTRODUCTION

Neuroendocrine tumors (NETs) are a heterogeneousgroup of neoplasms that most commonly arise fromthe bronchopulmonary (BP) tree and gastrointestinal(GI) tract [1]. According to the World Health Organiza-tion, BP-NETs are classified into four subtypes: (1) typ-ical carcinoid (TC); (2) atypical carcinoid (AC); (3) largecell neuroendocrine carcinoma; and (4) small-cell lungcarcinoma [2]. The terms BP carcinoid or well-differentiated BP-NET refer to low-grade TC andintermediate-grade AC tumors. Over the last 30 y, theincidence of BP carcinoids has increased, currently oc-curring in 1.57/100,000/y in the U.S. [3]. BP carcinoidsrepresent 2% of all lung neoplasms but are the mostcommon lung neoplasm in children [3].

Although BP carcinoids are well-differentiated NETswith low- to intermediate-grade, they have the poten-tial to behave aggressively and are considered malig-nancies. The 5-y survival rate for BP carcinoids—TCand AC—is 47% [3]. BP-TCs are the largest subgroup(80%–90%) of well-differentiated BP-NETs [3].

0022-4804/$36.00� 2010 Elsevier Inc. All rights reserved.

JOURNAL OF SURGICAL RESEARCH: VOL. 163, NO. 1, SEPTEMBER 201048

Approximately 11%–14% of patients with NETs havemetastatic disease with unknown primary site [1].Identification of the primary site, even in asymptomaticpatients with unresectable NET liver metastases, mayimpact clinical management. For example, some stud-ies suggest the resection of a GI-NET primary may beassociated with improved survival when unresectableNET liver metastases are present [4, 5]. Because well-differentiated NETs of the BP tree and GI tract are his-tologically similar, distinguishing the site of originbetween these two common sites is not possible basedon tumor morphology [6]. Immunohistochemical analy-sis of BP- and GI-NETs using one or a panel of markers,such as cytokeratin 7 (CK7), cytokeratin 20 (CK20),thyroid transcription factor-1 (TTF-1, also known asNKX2.1), and caudal type homeobox 2 (CDX2) hasyielded conflicting results [7–16]. Thus, bettermarkers are needed that can distinguish a BP- versusGI-NET.

The homeodomain transcription factor NKX2.2 is ex-pressed in rare populations of NE cells of normal gut, is-let cells of the pancreas and newly differentiating NEcells of the small intestine [17, 18]. NKX2.2 isa critical regulator of NE differentiation in the normalGI tract (i.e., small intestine and pancreas) [17–19].In our recent study showing that NKX2.2 is a newGI-NET marker [18], most GI-NETs (24/26, 92%) ex-pressed NKX2.2, including well-differentiated NETsof the stomach, duodenum, ampulla of Vater, pancreas,ileum, and colon. Interestingly, three BP-TCs did notexpress NKX2.2. The limited number of BP-TCs pre-cluded concluding at the time of our previous workthat NKX2.2 is not expressed in this tumor type yet ispresent in most GI-NETs. We therefore evaluatedNKX2.2 expression in a larger number of BP-TCs to de-termine whether NKX2.2 might be a useful marker todistinguish GI- from BP-NETs. We hypothesized that

TABL

Patient and Tumor

Age (y) Sex Race Procedure Size (cm)Li

1 68 F White LLL wedge 12 60 F White LLL lobectomy 3.53 34 M White LLL wedge 0.84 25 M White sleeve resection 15 29 M White LLL lobectomy 3.56 48 F White RML lobectomy 27 54 M White RUL lobectomy 0.38 38 M White sleeve resection 1.89 16 F White RML wedge 1.510 25 M White Left lingula 0.611 50 F White RML lobectomy 2.712 37 M Asian R pneumonectomy 6.813 29 F White RML/LL lobectomy 4

NH ¼ none harvested; CgA ¼ chromogranin A; 5-HT ¼ serotonin; IH

lack of NKX2.2 expression in BP-TCs might be usefulto distinguish BP- from GI-NETs.

METHODS

Human Tissues

Humantissuesamples wereobtained from resection specimens intheUniversity of California San Francisco (UCSF) Pathology Departmentafter approval by the Committee on Human Research, which is the in-stitutional IRB (IRB number: H6172-26943-05A). Thirteen previouslyundescribed patients with BP-TCs were identified in the UCSF pathol-ogydatabase. Normalhumanpancreasanda well-differentiatedNETofthe pancreas and one of the ileum were used as controls.

Histology and Immunohistochemistry

Tissues were fixed in Z-fix (Anatech, Battle Creek, MI, USA) and pro-cessed for paraffin sections [17]. The following primary antibodies wereused: mouse monoclonal anti-NKX2.2 (Developmental Studies Hybrid-oma Bank, Iowa City, IA, USA), 1:25; rabbit anti-chromogranin A(Immunostar, Hudson, MA, USA), 1:1000; rabbit anti-serotonin(Immunostar), 1:4000. Biotinylated secondary antibodies (Vector,Burlingame, CA, USA) were detected with the ABC Elite immunoper-oxidase system (Vector). Slides were imaged on an Axioskop 2 micro-scope (Zeiss, Thornwood, NY, USA).

Immunohistochemical analysis of antigen expression levels in hu-man tissues was based on the number of positively stained tumor cellsdivided by the total number of tumor cells to calculate the percentage.Because resection specimens contained the neoplasm and surround-ing normal tissue, tumor cells were distinguished from normal cellsbased on H and E staining: monomorph cytology, rare mitoses, lowatypia, and growth pattern (i.e., glandular, trabecular, insular, undif-ferentiated, or mixed) [6]. Numeric scores were based on the percent-age of positive tumor cells: 0 (<5%), 1 (5%–25%), 2 (26%–50%),3 (51%–75%), 4 (>75%).

RESULTS

Patient and Tumor Characteristics

During the time period from September 2000 throughMay 2005, we identified 13 previously undescribed pa-tients who underwent resection of primary BP-NETs

E 1

Characteristics

ymph nodenvolvement CgA IHC 5-HT IHC NKX2-2 IHC Syndrome

NH 4þ 4þ 0NH 4þ 4þ 0

1 of 1 4þ 0 0 Cushing’sNH 4þ 4þ 0NH 4þ 4þ 0

0 of 6 4þ 4þ 00 of 13 4þ 4þ 00 of 4 4þ 4þ 00 of 7 4þ 4þ 0NH 4þ 0 0 Cushing’s

0 of 3 4þ 0 00 of 8 4þ 0 00 of 7 3þ 2þ 0

C ¼ immunohistochemistry.

FIG. 1. Lack of NKX2.2 expression in human BP-TCs. (A)–(N) Sections of normal adult human tissues and well-differentiated NETs arestained with H and E or with the antibodies indicated for immunohistochemistry. (A) H and E staining of a BP-TC shows carcinoid morphology,rare mitoses, and no necrosis. CgA (C), (G), (J) and 5-HT (D), (K) expression in the cytoplasm of NETs are shown. (M) In normal human adultpancreas, NKX2.2 expression is detected in the islets but not exocrine compartment. (B) Lack of NKX2.2 expression in BP-TCs. AbundantNKX2.2 protein is present in the nuclei of pancreas (F) and ileum (I) NETs. (Color version of figure is available online.)

WANG ET AL.: NKX2.2 AND BRONCHOPULMONARY TYPICAL CARCINOIDS 49

classified as TCs at UCSF (Table 1). These patients hada median age of 37 y (range, 16–68 y), were mostlyWhite (92%), and just over half were male (54%).

According to the WHO, BP-TCs exhibit the followingcriteria: carcinoid morphology, size > 0.5 cm, < 2mito-ses/10 HPF, and lack necrosis (Fig. 1A) [2]. The mediantumor size was 1.8 cm (range, 0.3–6.8 cm). One patientwith a well-differentiated BP NE tumorlet (0.3 cm) wasincluded in this study. Most patients (92%) did not havelymph node involvement. None had distant metastases.Lobectomy was performed in six patients (46%), wedgeresection in three (23%), sleeve resection in two (15%),lingual resection in one (8%), and pneumonectomy inone (8%).

Frequency of Immunohistochemical Markers in BP-TCs

We observed intense expression of the general NEmarker CgA in all 13 BP-TC specimens, thereby con-firming the NE phenotype (Fig. 1C). Less frequently,we detected expression of the hormone serotonin in

9/13 (69%) specimens (Fig. 1D). Interestingly, twopatients (15%) who had BP-TCs in which serotonin ex-pression was undetectable had Cushing’s syndromedue to ectopic ACTH production (Table 1). The frequen-cies of CgA and serotonin expression as well as of ec-topic ACTH production, in our patient cohort, areconsistent with those reported for BP-TCs [3], suggest-ing our findings are reflective of BP-TCs in general.

We did not observe NKX2.2 expression in any of the13 BP-TCs (Fig. 1B). In contrast, we readily detected in-tense nuclear NKX2.2 expression in well-differentiatedNETs of the pancreas and small intestine (Fig. 1Fand I). Using normal human pancreas for a control,we readily detected nuclear NKX2.2 expression in theislets but not in the exocrine compartment (Fig. 1M).

DISCUSSION

In this study, we show that BP-TCs—the most com-mon type of well-differentiated BP-NET—do not

FIG. 2. The frequency of CgA- and NKX2.2-positive BP-TCs andGI-NETs. The data represent a compilation of information from thestudy presented herein and a prior study from our group [18].

JOURNAL OF SURGICAL RESEARCH: VOL. 163, NO. 1, SEPTEMBER 201050

express the homeodomain transcription factor NKX2.2.Because most BP-TCs express serotonin, our resultssuggest that NKX2.2 may not be required for serotoninproduction by BP-TCs and that distinct mechanismsunderlie NE differentiation in BP- and GI-NETs [18].Our findings build on our recent observation thatNKX2.2 is expressed in most (24/26, 92%) well-differentiated GI-NETs, including NETs of the stom-ach, duodenum, ampulla of Vater, pancreas, ileum,and colon, but not in 3 BP-TCs [18]. Therefore, todate, we have not detected NKX2.2 expression in 16BP-TCs. Figure 2 summarizes the expression patternsof CgA and NKX2.2 in BP-TCs and GI-NETs, includingdata from this study and our prior one [18]. All BP-TCsand GI-NETs expressed CgA, reflecting the NE pheno-type. Although most (92%) GI-NETs expressedNKX2.2; none of the BP-TCs expressed this marker.

NETs comprise a heterogeneous group of neoplasmsand most commonly arise from the BP tree and GI tract[1]. Currently, NETs are diagnosed based on morphol-ogy, expression of general NE markers (e.g., CgA),and detection of hormone excess. However, based onthese findings alone, the particular NET type often can-not be identified. Clinically, this issue is important inpatients with NET liver metastases and an unknownprimary tumor site. Immunohistochemical analyses us-ing various markers to discern a GI versus BP etiologyhave yielded conflicting results.

Studies of CK7 and CK20 expression have demon-strated utility in the differential diagnosis of some car-cinomas of epithelial origin [7]. These observations ledto evaluations in NETs. However, although CK7 andCK20 expression have been reported to be specific forBP carcinoids and GI-NETs, respectively [8], a largerstudy did not corroborate these findings [7]. Moreover,CK7 and CK20 expression have low sensitivity for

BP- and GI-NETs. When expression of CK7 and CK20is detected, it is usually patchy and present in lessthan 10% of cells [8].

Others have studied the utility of immunohistochem-ical analysis of developmental transcription factor ex-pression in distinguishing BP- from GI-NETs. Thyroidtranscription factor-1 (TTF-1, also known as NKX2.1)participates in normal thyroid, brain, and lung develop-ment [9]. Although TTF-1 expression appears to be spe-cific for BP carcinoids [9–11, 13–15], its sensitivity islow and variable, depending on the antibody used[12]. Caudal type homeobox 2 (CDX2) is involved inintestinal development and is specific for GI-NETs[13–16], but has low and variable sensitivity for GI-NETs. In particular, CDX2 expression does not appearto be a good marker for pancreatic NETs. Only 0%–18%pancreatic NETs stain positive for CDX2 [13–15].

CONCLUSION

In conclusion, we show here that BP-TCs do not ex-press NKX2.2. Based on this finding reported hereinand our previous findings that most GI-NETs expressNKX2.2 [18], we suggest that NKX2.2 is a highlysensitive and specific marker for well-differentiatedGI-NETs. Immunohistochemical staining for NKX2.2using the antibody in this study is robust, restrictedto the nucleus, and exhibits minimal background.Therefore, we believe staining for NKX2.2 could bereadily adopted into clinical practice. For patientswith a well-differentiated NET metastasis of an un-known primary, metastases that are CgA-positive/NKX2.2-negative would suggest a BP-NET primary,whereas those that are CgA-positive/NKX2.2-positivewould suggest a GI-NET primary. Further study ofNKX2.2 expression in NETs is warranted to validateour findings and to determine its utility in identifyingprimary and metastatic NETs from diverse primarysites.

ACKNOWLEDGMENTS

This work was supported by the James Ewing Oncology FellowshipAward (to EKN), American College of Surgeons Faculty Research Fel-lowship (to EKN), University of California Cancer Research Coordi-nating Committee (to EKN), UCSF Research Evaluation andAllocation Committee (REAC) Grant (to EKN), Novartis Oncologyprovided through the Cancer and Leukemia Group B Foundation(to EKN), and The V Foundation for Cancer Research (to EKB,EKN). The authors thank the UCSF Mouse Pathology Core for helpwith tissue processing, and Pamela Derish in the UCSF Departmentof Surgery for editorial assistance.

REFERENCES

1. Hauso O, Gustafsson BI, Kidd M, et al. Neuroendocrine tumorepidemiology: Contrasting Norway and North America. Cancer2008;113:2655.

WANG ET AL.: NKX2.2 AND BRONCHOPULMONARY TYPICAL CARCINOIDS 51

2. Beasley MB, Brambilla E, Travis WD. The 2004 World HealthOrganization classification of lung tumors. Semin Roentgenol2005;40:90.

3. Gustafsson BI, Kidd M, Chan A, et al. Bronchopulmonary neuro-endocrine tumors. Cancer 2008;113:5.

4. GiviB,PommierSJ,ThompsonAK,etal.Operativeresectionofpri-mary carcinoid neoplasms in patients with liver metastases yieldssignificantlybettersurvival.Surgery2006;140:891.discussion897.

5. Hellman P, Lundstrom T, Ohrvall U, et al. Effect of surgery onthe outcome of midgut carcinoid disease with lymph node andliver metastases. World J Surg 2002;26:991.

6. Kulke MH, Mayer RJ. Carcinoid tumors. N Engl J Med 1999;340:858.

7. Chu P, Wu E, Weiss LM. Cytokeratin 7 and cytokeratin 20 ex-pression in epithelial neoplasms: A survey of 435 cases. ModPathol 2000;13:962.

8. Cai YC, Banner B, Glickman J, Odze RD. Cytokeratin 7 and 20and thyroid transcription factor 1 can help distinguish pulmo-nary from gastrointestinal carcinoid and pancreatic endocrinetumors. Hum Pathol 2001;32:1087.

9. Agoff SN, Lamps LW, Philip AT, et al. Thyroid transcriptionfactor-1 is expressed in extrapulmonary small cell carcinomasbut not in other extrapulmonary neuroendocrine tumors. ModPathol 2000;13:238.

10. Oliveira AM, Tazelaar HD, Myers JL, et al. Thyroid transcrip-tion factor-1 distinguishes metastatic pulmonary from well-differentiated neuroendocrine tumors of other sites. Am J SurgPathol 2001;25:815.

11. Du EZ, Goldstraw P, Zacharias J, et al. TTF-1 expression is spe-cific for lung primary in typical and atypical carcinoids: TTF-1-positive carcinoids are predominantly in peripheral location.Hum Pathol 2004;35:825.

12. Matoso A, Singh K, Jacob R, et al. Comparison of thyroidtranscription factor-1 expression by 2 monoclonal antibodies inpulmonary and nonpulmonary primary tumors. Appl Immuno-histochem Mol Morphol 2009;18:142.

13. Saqi A, Alexis D, Remotti F, et al. Usefulness of CDX2 and TTF-1in differentiating gastrointestinal from pulmonary carcinoids.Am J Clin Pathol 2005;123:394.

14. Lin X, Saad RS, Luckasevic TM, et al. Diagnostic value of CDX-2and TTF-1 expressions in separating metastatic neuroendocrineneoplasms of unknown origin. Appl Immunohistochem Mol Mor-phol 2007;15:407.

15. Srivastava A, Hornick JL. Immunohistochemical staining forCDX-2, PDX-1, NESP-55, and TTF-1 can help distinguishgastrointestinal carcinoid tumors from pancreatic endocrineand pulmonary carcinoid tumors. Am J Surg Pathol 2009;33:626.

16. Mazziotta RM, Borczuk AC, Powell CA, et al. CDX2 immunos-taining as a gastrointestinal marker: Expression in lung carcino-mas is a potential pitfall. Appl Immunohistochem Mol Morphol2005;13:55.

17. Sussel L, Kalamaras J, Hartigan-O’Connor DJ, et al. Mice lack-ing the homeodomain transcription factor Nkx2.2 have diabetesdue to arrested differentiation of pancreatic b cells. Development11998;25:2213.

18. Wang YC, Gallego-Arteche E, Iezza G, et al. Homeodomaintranscription factor NKX2.2 functions in immature cells to con-trol enteroendocrine differentiation and is expressed in gastro-intestinal neuroendocrine tumors. Endocr Relat Cancer 2009;16:267.

19. Desai S, Loomis Z, Pugh-Bernard A, S, et al. Nkx2.2 regulatescell fate choice in the enteroendocrine cell lineages of the intes-tine. Dev Biol 2008;313:58.