Secondary Logo

Journal Logo

NordiQC Assessments of Low Molecular Weight Keratin 8/18 Immunoassays

Vyberg, Mogens MD*,†; Diernæs, Charlotte MD*; Røge, Rasmus MD*,†; Nielsen, Søren HT* ISIMM-NordiQC Immunoassay Performance Assessment Series, No. 2

Applied Immunohistochemistry & Molecular Morphology: November/December 2017 - Volume 25 - Issue 10 - p 673–678
doi: 10.1097/PAI.0000000000000603
Research Articles

This paper is number 2 in a series developed through a partnership between ISIMM and NordiQC for the purpose of reporting research assessing the performance characteristics of immunoassays in an external proficiency testing program.

*Institute of Pathology

Department of Clinical Medicine, Aalborg University Hospital, Aalborg, Denmark

The authors declare no conflict of interest.

Reprints: Mogens Vyberg, MD, Institute of Pathology, Aalborg University Hospital, DK-9000 Aalborg, Denmark (e-mail:

Received September 1, 2017

Accepted September 5, 2017

Back to Top | Article Outline


  • Select a sensitive antibody clone or clone cocktail detecting K8 or K8/18 (eg, EP17, EP17/EP30 or B22.1/B23.1) and use an optimized protocol based on heat-induced epitope retrieval (HIER).
  • Do not use clone 5D3 on the Ventana Benchmark platform.
  • Consider not to use clones, which show low performance rates (eg, CAM5.2 and 35βH11).
Back to Top | Article Outline


Ks are a large family of intermediate sized (~10 nm in diameter) cytoskeletal filaments, essential for normal structure and function of all mammalian epithelial cells. The K filaments are obligatory heterodimers each formed by one acidic (type I) and one basic-to-neutral (type II) K protein. Currently at least 54 human K types have been identified and numbered according to the nomenclature proposed by a consensus group in 2006 based on the 1982 catalog of Moll.1–4 Of these, 37 are epithelial (cytokeratins) while the rest are “hard” (hair and nail) Ks. From 2 to 10 K types are found in highly specific patterns in the individual cells related to their type and stage of differentiation. LMW Ks includes the type I K19, K18, K17, and K20, and the type II K8 and K7. K8 and K18 are most often paired, designated K8/18. They are the first to appear in embryogenesis and the most widespread Ks in the differentiated cells, representing primary (constitutive) Ks of simple epithelia. K8/18 are generally the only Ks expressed in normal parenchymatous epithelia like hepatocytes, renal proximal tubular cells, pancreatic, salivary gland and prostate acinar cells, adrenal cortical cells, and neuroendocrine cells in lung and pancreas. In other endocrine cells, additional LMW Ks may appear, for example, K7 (thyroid gland), K19 (pituitary gland, gastrointestinal tract, Merkel cells), and K20 (pituitary gland, Merkel cells).5 LMW Ks including K8 and/or K18 are furthermore found in complex epithelia (mainly the luminal compartments), urothelium, nonkeratinizing squamous epithelium (focally), mesothelium and few mesenchymal cell types, for example, fibroblastic reticulum cells, smooth muscle cells, and endothelial cells.6

Primary and metastatic carcinomas and endocrine tumors tend to retain the K profiles of the putative cell of origin. Thus, the majority of hepatocellular, clear cell renal cell, prostate and adrenal cortical carcinomas as well some neuroendocrine neoplasms solely express K8/18,2,5,7 as opposed to the plethora of other epithelial tumor types expressing a wider spectrum of K types. Therefore, immunohistochemical (IHC) mapping of K types including K8/18 is important in classification of carcinomas of unknown primary origin.

Identification of LMW K types such as K8/18 is also important in the study of neoplastic development and differentiation. In contrast to normal squamous epithelia, K8/18 may be widely expressed in squamous cell carcinoma of the uterine cervix and head and neck as well as their precursor lesions.8–10 In intraductal breast lesions and invasive ductal carcinomas, identification of the luminal cell Ks K8/18 (together with luminal cell markers K7 and K19, and basal/myoepithelial cell markers K5 and K14) are relevant for subclassification and prognostication.11,12

Neoplasms derived from simple epithelia, expressing K8/18 as the only K types, will often show a low level of K expression. IHC detection of K8/18 therefore calls for sensitive and robust immunoassays based on appropriate choice of antibodies (Abs) and protocols calibrated for the purpose. However, there is a large number of commercially available K8/18 Abs, and staining protocols can be set up in countless ways making it difficult for laboratories (Labs) to identify and implement the best immunoassays for LMW K. Less successful K Abs and suboptimal protocols may result in ambiguous or false-negative staining reactions hampering the diagnostic utility and consequently the reliability of IHC.

In the Nordic immunohistochemical Quality Control (NordiQC) external quality assessment program, a high proportion of LMW K (specified as K8/18) stains submitted by the participants for central assessment has been marked insufficient. The aim of the present paper is to give an overview of external quality assessment results for LMW K tests able to detect K8/18 in the latest 3 runs 33, 38, and 49, accomplished in 2011, 2013, and 2017, respectively. The detailed results are available on the NordiQC homepage (

Back to Top | Article Outline


The assessments were performed as described on, the principles being outlined in a recent paper.13 In brief, tissue microarrays for each of the runs were constructed, comprising 6 to 7 cores, each 3 mm in diameter. Tissues included were normal appendix, liver and esophagus, breast ductal carcinoma, clear cell renal cell carcinoma, and endocrine carcinoma (eg, small cell lung carcinoma). In one run (run 33) also hepatocellular carcinoma was included. All tissues were selected to be able to fulfil the criteria for optimal staining results defined by the NordiQC assessor group in compliance with international recommendations as to reflect the purpose of the immunoassay to demonstrate K LMW in carcinomas with different expression levels of the target analyte.14 Optimal staining required a strong, distinct cytoplasmic staining reaction of the appendiceal columnar epithelial cells and bile duct epithelial cells, an at least weak to moderate predominantly membranous staining reaction of the large majority of hepatocytes, a moderate to strong, distinct cytoplasmic staining reaction of virtually all neoplastic cells in the breast ductal carcinoma and hepatocellular carcinoma, and an at least weak to moderate cytoplasmic staining reaction of the majority of neoplastic cells in the renal clear cell carcinoma and neuroendocrine carcinoma. A faint reaction in some endothelial cells, reticulum cells and smooth muscle cells was expected while other nonepithelial cells should be negative.

All participants submitted their assay protocol on the NordiQC homepage and received a set of slides cut from the tissue microarray block above, stained them according to their protocol and returned one of them for central assessment performed in blinded manner by an expert panel of pathologists and technicians examining each slide on a microscope linked to a monitor. Each slide was by consensus marked as “optimal,” “good,” “borderline” or “poor” based on the technical quality and adherence to the staining criteria outlined above. Optimal and good were grouped as “sufficient,” borderline and poor grouped as “insufficient.” Only protocols based on an Ab (or Ab cocktail) directed against K8 and/or K18 were accepted. Abs, which also react with another LMW K were accepted (namely, CAM5.2, reacting weakly with K7).

Back to Top | Article Outline


The number of Labs participating in the 3 runs was 141, 161, and 213, respectively. For the present overview, the results for all 515 slides are pooled. Totally, 195 (38%) were assessed as optimal, 161 (31%) good, 102 (20%) borderline and 57 (11%) poor. Up to 11 different clones were used in concentrated formats (Conc), while the number of ready-to-use formats (RTU) increased from seven in run 33 to 14 in run 49. Table 1 shows the most commonly used clones/formats (≥10 slides, some products with closely related results are grouped together, see details below).



Clone EP17 [single (Epitomics) or in a cocktail with EP30 (Agilent/Dako), Fig. 1, row A] gave the best performance both used as Conc (Epitomics) and in an RTU system (Agilent/Dako). All of 33 submitted stains were sufficient, the large majority optimal (for the RTU product all of 16 were optimal). Pretreatment was in all cases HIER in an alkaline buffer.



Clone cocktail B22.1/B23.1 (Conc: Roche/Ventana, CellMarque; RTU: Roche/Ventana, CellMarque, Master Diagnostica), clone DC10 (Conc: several vendors, RTU: Agilent/Dako) and clone TS1 (Conc: several vendors) all gave good performance both when used as Conc and in RTU formats. Most insufficient staining reactions were due to use of suboptimal protocols (too diluted Ab, insufficient or missing HIER, less sensitive detection systems) while the proportion of sufficient results were 90% to 100% with optimal protocol settings. For the RTU formats, the proportion of sufficient assays were >80%. However, for the clone cocktail B22.1/23.1 in the RTU system from Roche/Ventana, optimal results in run 49 were only seen in Labs modifying the recommended protocol in order to increase the sensitivity. Moreover, DC10, which does not detect K8, gave weak reaction in carcinomas with very low expression of K18 (Fig. 1, row B).

Clone 5D3 (Conc.: several vendors; RTU: Biocare, Diagnostics BioSystems, Leica/Novocastra) gave, when used as Conc, an overall low proportion of sufficient results (55%). However, with optimized protocols the proportion of sufficient results increased to almost 100%. One reason for the poor performance was the use of proteolytic pretreatment (which is recommended by Leica/Novocastra and Thermo Fisher/Neomarkers) instead of HIER (as recommended by NordiQC). Moreover, clone 5D3 used on the Benchmark XL/Ultra platform (Roche/Ventana) failed in all cases (Fig. 2). When used as an RTU system, clone 5D3 gave excellent results (as for the RTU system, Leica recommends HIER).



Clone CAM5.2 (Conc: Cell Marque, Zytomed, Immunologic; RTU: Becton Dickinson, Master Diagnostica, Roche/Ventana, Fig. 1C) generally gave less satisfactory results: For Conc 44% sufficient, 3% optimal, and for RTU 37% sufficient, 8% optimal. The RTU product sold by Becton Dickinson, which is not designed for a specific platform, was used by 21% of the Labs, giving sufficient results in 29% and optimal in 3% only. In contrast, the new Roche/Ventana RTU system, only used in run 49, gave a better performance (60% sufficient). Proteolytic pretreatment with Proteinase K consistently gave better results than other proteases as well as HIER (which for Conc was used by about 30% of the Labs).

Clone 35βH11 (RTU: Roche/Ventana/Cell Marque, Monosan, ZS, Fig. 1D) gave, when grouped together, sufficient results in only 5%, none of which were optimal. A few Labs used this clone as a Conc, also obtaining unsatisfactory results.

Back to Top | Article Outline


Robust and sensitive Abs and methods are of great importance in diagnostic IHC. Many Labs do not have the capacity to validate and implement immunoassays but will select an Ab and protocol known from the literature and/or rely on the protocol settings recommended by the Ab producer and/or described in studies where a profound validation may not have been carried out or specified. In order to support Labs to implement immunoassays for diagnostic use, NordiQC publishes on its open website ( detailed protocols for all Abs providing an optimal result in the assessment for the specific target analyte. Access to these protocols and information about best practice tissue controls is an indispensable tool to implement and use immunoassays in diagnostics. The performance for Labs that have participated more than once in a NordiQC assessment for a specific marker is on the average 20% higher than for Labs participating for the first time, which suggest an effect of guiding the Labs. Yet, the over-all rate of insufficient results among participating Labs is consistently in the region of 30%. This also counts for LMW K and may—apart from new participants—be ascribed to new (and usually better) Abs, reagents and platforms, which increase demands to the Labs.13

A prevailing feature of the NordiQC results for LMW K detection is too weak or false-negative staining reactions, particularly in normal tissues and tumors with low-level expression of Ks, while overstaining or false-positive results almost never occur. While most Labs could demonstrate LMW Ks in the high-level expressing cells (appendix enterocytes, bile duct epithelium and breast carcinoma), many failed in detecting the Ks in low-level expressing cells (hepatocytes, neuroendocrine carcinoma, and clear cell renal cell carcinoma). This emphasizes the tendency for Labs to base protocol calibration on tissues with high-level expression of Ks, giving too low analytical sensitivity. Many Labs use proteolytic pretreatment for Abs that perform much better with HIER. One reason for this may be data sheets which still recommend proteolysis. Among the Abs mentioned in this paper, only clone CAM5.2 requires proteolytic pretreatment. Clone CAM5.2 has in the NordiQC tests given relatively poor results. Partly because HIER has been used, partly because the clone CAM5.2 RTU product from Becton Dickinson Biosciences, which is not designed for any particular platform, is used inappropriately, sometimes even diluted.

The large majority of published studies of K8 expression in tumors are based on Ab clones 35βH11 and CAM 5.2 while clone DC10 is mostly used for detection of K18. Noteworthy, even though K8 and K18 are paired, K8 is in many studies detected less frequently than K18. Thus, with Ab clone 35βH11 Skinnider et al7 found only 6/15 (40%) K8 positive clear cell renal cell carcinoma while 15/15 (100%) were K18 positive with clone DC10. However, NordiQC has revealed a poor performance of clone 35βH11. In a study of K distribution in endothelial cells and vascular tumors, Miettinen and Fetsch6 reported normal endothelial cells negative for K8 with Ab clone CAM 5.2 while K18 positive with Ab clone DC10. Likewise, they detected K8 in only 10% of epithelioid angiosarcomas but K18 in 100%. However, in that study, K8 was retrieved with pepsin for clone CAM 5.2 (for which proteinase K would give better results) while K18 was retrieved with HIER for clone DC10. Studying renal cell carcinomas, Langner et al15 reported a significantly higher proportion of K18 positive (as detected with Ab clone DC10) than K8/18 positive cases (as detected with Ab clone 5D3). However, they used proteolytic pretreatment for clone 5D3 (for which HIER would give better results).

In an assessment organized by Canadian Immunohistochemical Quality Control, minimal differences between participating Labs was found when tissues with high expression of LMW Ks (basically K8/18) were analyzed. However, when tissues with low-level expression were analyzed, 9/11 (81%) Labs failed to demonstrate LMW Ks in normal hepatocytes or proximal tubules of the kidney or both.16 LMW Ks were detected in renal cell carcinomas by 5/13 participants. The 8 participants with false-negative results also failed to demonstrate LMW Ks in the mentioned normal epithelial cells. These findings support that protocol calibration may often have been performed on high expressing tissues instead of low expressing. However, Information about Abs and protocols in the Canadian Immunohistochemical Quality Control challenge was not given.

In conclusion, the NordiQC results for LMW K assays demonstrate marked differences in performance between different Abs, protocols, and stainer platforms. We suggest that Labs should select a sensitive clone or clone cocktail detecting K8 or K8/18 (eg, EP17, EP17/EP30 or B22.1/B.23.1), and optimize the protocol according to our directions based on low expressing tissues. Labs should be aware of issues with clone DC10 (which does not detect K8) and clone CAM5.2 (which requires proteolytic pretreatment). Insensitive clones like 35βH11 should not be used. Labs should also be aware of misleading data sheets, which for example, recommends proteolytic pretreatment for LMW K Abs (except CAM5.2), when optimal performance requires HIER.

Back to Top | Article Outline


1. Moll R, Franke WW, Schiller DL, et al. The catalog of human cytokeratins: patterns of expression in normal epithelia, tumors and cultured cells. Cell. 1982;31:11–24.
2. Chu PG, Weiss LM. Keratin expression in human tissues and neoplasms. Histopathology. 2002;40:403–439.
3. Schweizer J, Bowden PE, Coulombe PA, et al. New consensus nomenclature for mammalian keratins. J Cell Biol. 2006;174:169–174.
4. Moll R, Divo M, Langbein L. The human keratins: biology and pathology. Histochem Cell Biol. 2008;129:705–733.
5. Chu PG, Lau SK, Weiss LM. Keratin expression in endocrine organs and their neoplasms. Endocr Pathol. 2009;20:1–10.
6. Miettinen M, Fetsch JF. Distribution of keratins in normal endothelial cells and a spectrum of vascular tumors: implications in tumor diagnosis. Hum Pathol. 2000;31:1062–1067.
7. Skinnider BF, Folpe AL, Hennigar RA, et al. Distribution of cytokeratins and vimentin in adult renal neoplasms and normal renal tissue: potential utility of a cytokeratin antibody panel in the differential diagnosis of renal tumors. Am J Surg Pathol. 2005;29:747–754.
8. Carrilho C, Alberto M, Buane L, et al. Keratins 8, 10, 13, and 17 are useful markers in the diagnosis of human cervix carcinomas. Hum Pathol. 2004;35:546–551.
9. Smedts F, Ramaekers F, Robben H, et al. Changing patterns of keratin expression during progression of cervical intraepithelial neoplasia. Am J Pathol. 1990;136:657–668.
10. Matthias C, Mack B, Berghaus A, et al. Keratin 8 expression in head and neck epithelia. BMC Cancer. 2008;8:267.
11. Böcker W, Hungermann D, Weigel S, et al. Immunohistochemistry in breast pathology: differential diagnosis of epithelial breast lesions. Pathology. 2009;30:13–19.
12. Mackinder MA, Evans CA, Chowdry J, et al. Alteration in composition of keratin intermediate filaments in a model of breast cancer progression and the potential to reverse hallmarks of metastasis. Cancer Biomark. 2012;12:49–64.
13. Vyberg M, Nielsen S. Proficiency testing in immunohistochemistry—experiences from Nordic Immunohistochemical Quality Control (NordiQC). Virchows Arch. 2016;468:19–29.
14. Torlakovic EE, Nielsen S, Francis G, et al. Standardization of positive controls in diagnostic immunohistochemistry: recommendations from the International Ad Hoc Expert Committee. Appl Immunohistochem Mol Morphol. 2015;23:1–18.
15. Langner C, Wegscheider BJ, Ratschek M, et al. Keratin immunohistochemistry in renal cell carcinoma subtypes and renal oncocytomas: a systematic analysis of 233 tumors. Virchows Arch. 2004;444:127–134.
16. Copete M, Garratt J, Gilks B, et al. Inappropriate calibration and optimisation of pan-keratin (pan-CK) and low molecular weight keratin (LMWCK) immunohistochemistry tests: Canadian Immunohistochemistry Quality Control (CIQC) experience. J Clin Pathol. 2011;64:220–225.

low molecular weight keratins; external quality assessment; NordiQC

Copyright 2017 Wolters Kluwer Health, Inc. All rights reserved.