The monoclonal antibody EPR1614Y against the stem cell biomarker keratin K15 lacks specificity and reacts with other keratins

Keratin 15 (K15), a type I keratin, which pairs with K5 in epidermis, has been used extensively as a biomarker for stem cells. Two commercial antibodies, LHK15, a mouse monoclonal and EPR1614Y, a rabbit monoclonal, have been widely employed to study K15 expression. Here we report differential reactivity of these antibodies on epithelial cells and tissue sections. Although the two antibodies specifically recognised K15 on western blot, they reacted differently on skin sections and cell lines. LHK15 reacted in patches, whereas EPR1614Y reacted homogenously with the basal keratinocytes in skin sections. In cultured cells, LHK15 did not react with K15 deficient NEB-1, KEB-11, MCF-7 and SW13 cells expressing only exogenous K8 and K18 but reacted when these cells were transduced with K15. On the other hand, EPR1614Y reacted with these cells even though they were devoid of K15. Taken together these results suggest that EPR1614Y recognises a conformational epitope on keratin filaments which can be reconstituted by other keratins as well as by K15. In conclusion, this report highlights that all commercially available antibodies may not be equally specific in identifying the K15 positive stem cell.


Results
The epitope locations of 4 different K15 antibodies on the K15 polypeptide are shown in Fig. 1A. Two of them, LHK15 and EPR1614Y, are monoclonal and both are commercially available. The EPR1614Y epitope is located between 400-456 residues on K15 polypeptide (Abcam, UK; www.abcam.com/cytokeratin-15-antibody-epr16 14y-ab52816.html) and therefore could overlap with the last 13 residues of the rod domain, on the other hand LHK15 has its epitope located in the last 17 residues of the tail domain 8 . We designed a series of experiments to compare the reactivities of EPR1614Y and LHK15 on cells and tissue sections. On western blotting using lysates derived from N/Tert-1, a K15+ cell line 31 and MCF-7, a K15-human breast carcinoma cell line 32 , grown in 2-dimensional (2-D) cultures, both antibodies recognised a single 50 kDa band only in N/Tert-1 but not in MCF-7 cells (Fig. 1B). This set of experiments established that on western blot both antibodies were specifically reacting with the K15 polypeptide.

Immunoreactivity of LHK15 and EPR1614Y with basal keratinocytes in skin.
To compare the reactivity of the two antibodies in skin tissues, sections of freshly frozen human foreskin and abdominal skin were immunostained with LHK15 and EPR1614Y. As shown in Fig. 2 both antibodies specifically reacted with the basal layer; EPR1614Y displayed homogenous reactivity with the basal keratinocytes whereas LHK15 showed interrupted staining in sections of both foreskin and abdominal skin. We also compared the K15 staining with K14, a known basal cell specific marker, using anti-K14 LLOO1 antibody, which reacted with basal and some suprabasal keratinocytes in both types of skin samples ( Fig. 2A,B).
LHK15 and EPR1614Y immunoreactivity with differentiating keratinocytes in organotypic cultures (OTCs). To compare reactivity of the two antibodies in differentiating keratinocytes we prepared OTCs from HaCaT, N/Tert-1, KEB-11 and NEB-1. As shown in Fig. 4, only 3 cell lines HaCaT, N/Tert-1 and NEB-1 produced stratified epithelium, which as expected, was thicker after 9 days of culturing compared with that obtained after 7 days. KEB-11 cells did not produce a stratified epithelium under these conditions (Fig. 4).
To characterise the OTCs we immunostained the cryofixed sections of 3-D cultures for K14, a marker of basal keratinocytes 1 , and K10, a marker of suprabasal layers 32 . The expression of K14 was more or less homogenous in the stratified layers of OTCs from HaCaT and N/Tert-1. In NEB-1 cultures, K14 was expressed in patches of cells located near the junction of epithelium and the collagen base (Fig. 5A), whereas K10 expression was primarily located towards the suprabasal layers in OTCs from all 3 cell lines (Fig. 5B). In OTCs of HaCaT only few cells located at the top showed K10 expression whereas in N/Tert-1 and NEB-1 strong expression was seen in all suprabasal layers (Fig. 5B).
We then investigated the reactivity of EPR1614Y and LHK15 with cryofixed sections of OTCs derived from HaCaT, N/Tert-1 and NEB-1. EPR1614Y reacted strongly with all layers of OTCs derived from the 3 different cell lines (Fig. 5C). LHK15 on the other hand reacted weakly with HaCaT OTCs but it reacted strongly with N/ Tert-1 OTCs (Fig. 5D). The most significant difference was the strong reactivity of EPR1614Y with NEB-1 derived OTCs, whereas LHK15 showed no reactivity with these OTCs (Fig. 5C,D).
K15 add back experiments to restore antibody reactivity. To further investigate the specificity, we tested reactivity of these antibodies after overexpressing K15 in cells devoid of this polypeptide. As K15 is absent in NEB-1 cells (Fig. 3B-D), we ectopically expressed K15 in these cells using retroviral transduction and immunostained the cells with LHK15 and EPR1614Y. As shown in Fig. 6A the NEB-1 cells transduced with the vector control virus immunoreacted with anti-K14 LLOO1 but did not stain with LHK15 ( Fig. 6B) whereas the K15 transduced cells stained strongly with this antibody (Fig. 6C). EPR1614Y antibody on the other hand reacted strongly with NEB-1 in the absence as well as in the presence of K15 (Fig. 6D,E). These observations suggest that EPR1614Y may be reacting non-specifically with keratin filaments in the absence of K15.
We then did a similar experiment using MCF-7 cells, which express only K8, K18 and K19 1,32 . Vector control MCF-7 cells reacted with K8 antibody A45-B/B3 (Fig. 6F), they did not react with LHK15 ( Fig. 6G) but the reactivity was induced when K15 was added into MCF-7 cells (Fig. 6H). On the other hand, EPR1614Y reacted strongly with MCF-7 vector control cells (Fig. 6I) which did not change when K15 was added into these cells (Fig. 6J). To substantiate the above observations, we transduced SW13, a cell line which does not express keratin filaments, with recombinant retroviruses expressing K8 and K18 and applied double selection with G418 and hygromycin. K8 and K18 formed IFs in SW13 cells as shown by A45-B/B3 immunostaining (Fig. 6K), these IFs did not react with LHK15 (Fig. 6L). SW13 cells expressing K8 + K18 [SW13(K8 + K18)] showed weak but detectable reactivity with EPR1614Y (Fig. 6N). The SW13(K8 + K18) cells were further transduced with K15 containing retroviruses and they were subjected to triple selection with G418, hygromycin and puromycin. As expected, the immunoreactivity of LHK15 with SW13(K8 + K18) was induced when K15 was added into these cells (Fig. 6M) and the reactivity of EPR1614Y with these cells became stronger and more filamentous in the presence of K15  6O). Taken together these results suggest that the reactivity of LHK15 is specific to K15 whereas EPR1614Y will react non-specifically with other keratins in the absence of K15 polypeptide.

Discussion
Immunochemistry, flow cytometry, enzyme-linked immunosorbent assays (ELISAs) and western blotting provide some of the most valuable, powerful, research as well as diagnostic tools in the field of biomedical sciences. However, it is well-known that the choice of the most specific and reliable antibody is of critical importance, as it determines the validity and credibility of the scientific data. Use of a non-specific antibody could lead to false results and conclusions 34 . As an example, a 2009 editorial by Michel and co-workers demonstrated that lack of selectivity appears to be the rule rather than the exception for antibodies against G-protein-coupled receptors (49 antibodies against 19 subtypes of the receptors have been tested in those studies and none of them was found to be selective) 35 . Other authors had to withdraw their publications because the antibodies they used against novel markers were later found to react in tissues from knock-out mice that lacked the markers 36 . Hence, determining the specificity of antibodies is extremely important especially when the antigen, K15, is widely used as a biomarker for adult stem cells [37][38][39][40][41][42] . Before the use of synthetic peptides for monospecific antibody production became common, most of the antibodies raised were against keratin complexes derived from cytoskeletal extracts or using purified peptides from SDS gels. Some of these antibodies recognised keratin filaments only under certain physiological conditions. For example, AE1 and COU-1 react only with tumour cells 43,44 and K G 8.13 reacts with cells undergoing mitosis 45 . A number of monoclonal antibodies such as K s 18.18, LE61 and LE65 do not react with individual polypeptide but instead their epitopes are reconstituted only when complementary keratins associate into heterotypic complexes [46][47][48] . Other antibodies, such as A45-B/B3, are able to recognise conformational changes associated with heterotypic association of keratin pairs 49 . On the other hand, most antibodies raised using synthetic peptides reacted strongly with specific keratins on western blots as well as immunochemistry which increased their usefulness in keratin expression studies. Here we have compared the reactivity of two of the most widely used monoclonal antibodies, LHK15 and EPR1614Y, for the stem cell biomarker K15 by western blotting and immunochemistry in 2-D and 3-D cultures.
The choice of K15 antibody to address a particular set of research questions should be decided primarily by the techniques to be employed. On western blotting both antibodies, LHK15 and EPR1614Y, reacted strongly with K15 polypeptides in N/Tert-1 but no reactivity was observed with any protein band in MCF-7 cells suggesting they both specifically recognise a linear or denatured epitope. Therefore, if western blotting assay for K15 expression is to be used then either of the antibodies could be employed. However, if the expression of K15 polypeptide is to be investigated in cells and tissues by immunochemistry then significant differences in the ability of LHK15 and EPR1614Y to recognise K15 were observed. In the epidermis both LHK15 and EPR1614Y reacted with the basal keratinocytes, however LHK15 gave discontinuous reactivity which was consistent with the data reported previously 8,9 , EPR1614Y on the other hand gave a homogenous reactivity with both foreskin as well as abdominal skin. This is in variance with a recent report where EPR1614Y was shown to recognise patches of basal keratinocytes in human breast skin 50 . It is therefore conceivable that patchy vs continuous staining of K15 depends on the body site where skin was taken for immunohistochemistry. Although the molecular basis for the discontinuous staining of adult skin section by LHK15 is not clear, it could be due to the heterogeneity in the basal keratinocytes reported previously [51][52][53] . This antibody has been shown to give a continuous staining in foetal skin as well as internal epithelia such as oral and oesophageal epithelia 8,9 .
On immunocytochemistry the IFs in the human epidermal keratinocyte cell lines, NEB-1, KEB-11 and N/ Tert-1, reacted very differently with LHK15 and EPR1614Y. On western blotting, both LHK15 and EPR1614Y did not react with any protein band in lysates of NEB-1 and KEB-11 suggesting they did not express K15 (Fig. 3C,D). The absence of K15 in NEB-1 and KEB-11 could be due to HPV16 induced immortalisation, perhaps involving post-transcriptional mechanisms 33 . Absence of K15 in KEB-11, which already does not express K14, would leave no type I basal keratin which explains why this cell line does not form stratified epithelia in OTCs (Fig. 4C). Surprisingly, EPR1614Y gave positive reactivity with all epithelial cell lines used in this study whereas LHK15 reacted only with HaCaT and N/Tert-1 (Figs 3, 5 and 6). Absence of immunofluorescence reactivity of LHK15 with KEB-11, NEB-1 and MCF-7 suggests that presence of K15 is an absolute requirement for this antibody to show reactivity. EPR1614Y on the other hand reacted with keratin cytoskeleton even in the absence of K15, which implies that the three-dimensional reconstitution of EPR1614Y epitope may not require presence of K15.
The epitope of LHK15 is derived from the last 17 residues of K15 polypeptide which is the end of the tail domain (Fig. 1A), a region which can vary strongly in size and sequence amongst keratins 54,55 . Therefore, an antibody raised against this region is most likely to be specific towards its target. The epitope for EPR1614Y is poorly defined. This antibody was generated by Abcam UK using their own patented technology and their website claimed the epitope was between residues 400-500, however, as there are only 456 residues in K15 polypeptide, the epitope must be between 400-456 residues. Assuming the epitope begins from residue 400 in K15 polypeptide, the peptide used to raise this antibody would contain the last 13 residues of the rod domain. This region is part of the Helix Termination Peptide (HTP), which is highly conserved and it is a high affinity association site for stabilisation of heterotypic complexes between type I and type II keratins 56 . Mutations in this region are known to destabilise the cytoskeleton and are associated with a large number of blistering syndromes [57][58][59] . The high degree of sequence conservation in this region amongst different keratins 56 suggests that heterotypic keratin complexes could assemble into a very similar three-dimensional epitope even in the absence of K15. This hypothesis is supported by the data presented in Fig. 6 which shows that EPR1614Y reactivity in the absence of K15 seems to depend on the number of polypeptides present in the cytoskeleton, with low reactivity when K8 + K18 were present and increasing strongly in MCF-7, with additional keratin K19, and in NEB-1 which would express other   60 . Interestingly the epitope of this antibody is also located in the highly conserved HTP region 49,61,62 .
The antibody LHK15 reacts with K15 from human, rat, cow and mouse (www.abcam.com/ cytokeratin-15-antibody-lhk15-ab80522.html) whereas EPR1614Y reacts only with mouse and human (www. abcam.com/cytokeratin-15-antibody-epr1614y-ab52816.html) but both reagents react strongly with keratin filaments fixed with either acetone/methanol or with formalin. As the epitope for EPR1614Y is not defined, it is conceivable that the epitope does not involve HTP and it is derived entirely from the tail domain. In that case it is possible that during heterotypic dimerisation the tail domains of type I and type II keratins associate such that they form a 3-D structure mimicking the EPR1614Y epitope.
The EPR1614Y antibody has been widely used to study K15 expression as a biomarker of epidermal stem cells in scar formation 63 , differentiation of human hair follicle cells into neurons 64 , psoriatic skin lesions 50 , ameloblastoma 65 , mammary epithelium 66 , developing ureters 67 and human cornea 37 . In view of our data that EPR1614Y would react with cell cytoskeleton even when K15 is absent, some of the published work in which this antibody has been used may require reinterpretation. Furthermore, as our understanding of the clinical relevance of K15 positive cells in cancer, tissue homeostasis and other pathology develops, it is fundamental that only highly specific antibodies are employed to correctly identify these important cells in vivo.
In conclusion, we show that both antibodies, LHK15 and EPR1614Y, recognise K15 on western blotting. However, in immunocytochemistry LHK15 showed reactivity only when K15 was present whereas EPR1614Y showed strong reactivity even when there was no K15 polypeptide. The study highlights the extreme care one should exercise in choosing and validating the specificity of an antibody for their studies. Retroviral packaging, viral transduction using spinfection. Phoenix E cells were transfected with the constructs pLPCneo, pLPChygro and pLPCpuro (control vectors) and pLPCneo_K18, pLPChygro_K8 and pLPCpuro_K15 using TransIT ® -LT1 transfection reagent (Cambridge Biosciences, UK). After 48 h culture supernatants containing ecotropic viruses were used to transduce PT67 cells in presence of 5 μg/ml polybrene (1,5-dimethyl-1,5-diazaundecamethylene polymethobromide) using the spinning method as described previously 31,75 . Briefly, the viral supernatant was treated with 5 μg/ml polybrene and the growth medium was replaced with the prepared viral supernatant. Plates were centrifuged at 1000 rpm at 32 °C for 1 h and then incubated for 24 h at 32 °C in a humidified atmosphere of 10% CO 2 . Cells were cultured in fresh complete growth medium at 37 °C for 48 h to ensure viral integration and protein expression followed by appropriate drug selection (1 mg/ ml G418; 400 µg/ml hygromycin and 1-1.5 µg/ml puromycin) for several days. Amphotropic retroviral supernatants were collected from PT67 cells three times every 24 h and snap frozen and stored at −80 °C. SW13 cells were transduced with K18 followed by K8 amphotropic retrovirus and selected with a combination of G418 and hygromycin. MCF-7, SW13 (K8 + K18) and NEB-1 cell lines were transduced with K15 (or control) amphotropic retroviral supernatant using the spinfection method as described above and selected with puromycin.

3-dimensional tissue culture model of keratinocyte differentiation.
HaCaT, N/Tert-1, KEB-11 and NEB-1 keratinocytes were seeded on a human primary dermal fibroblast containing disc of rat-tail collagen type I (Corning, UK) in a Millipore ® Millicell ® (Sigma-Aldrich, UK). The cells were cultured for 24 h at 37 °C in a humidified atmosphere of 10% CO 2 and maintained in FAD medium. After 24 h the collagen disc was raised to the air-liquid interphase to induce keratinocyte differentiation. The cells were cultured for 7 or 9 more days after which they were either fixed with 4% (w/v) paraformaldehyde, embedded in paraffin or cryofixed and sectioned for immunostaining. Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare, Little Chalfont, UK). Peroxidase activity was detected with Amersham Hyperfilm ™ ECL autoradiography film (GE Healthcare).

SDS electrophoresis and western blotting.
Immuno-cytochemistry and -histochemistry. Cells were cultured on collagen-coated coverslips for 24 h, fixed in a mixture of ice-cold acetone and methanol (1:1) for 10 min and dried in air. After blocking the non-specific antibody reactivity with 10% (v/v) normal goat serum, the coverslips were incubated with primary antibody for 2 h at RT. After washing with phosphate-buffered saline (PBS) the coverslips were incubated with Alexa Fluor ® 488-labelled anti-mouse or anti-rabbit secondary antibodies in the dark for 1 h at RT. The coverslips were washed again and mounted on glass slides using VECTASHIELD HardSet Antifade Mounting Medium with DAPI (Vector Laboratories, UK) and visualised with a Leica Epi DM5000B or DM4000B microscope equipped with a DFC350 FX digital camera and 20x/0.5 NA and 40x/0.75 NA objective lenses. The 3-D reconstructed plugs or human tissues were cryo-sectioned and attached to glass slides, fixed in ice-cold methanol for 15 min and washed with PBS. The sections were treated with primary antibodies at 4 °C overnight, washed, stained with secondary antibody and mounted in DAPI containing mounting medium as described above for cultured cells.

Data Availability
Any supporting data not included in this manuscript or reagents used in this study, which are not commercially available, will be provided to readers following a written request to the corresponding author.