Biology of Blood and Marrow Transplantation
Volume 9, Issue 9 , Pages 559-570, September 2003

An epithelial target site in experimental graft-versus-host disease and cytokine-mediated cytotoxicity is defined by cytokeratin 15 expression

  • Diana Whitaker-Menezes

      Affiliations

    • Departments of Pathology, Anatomy & Cell Biology, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
    • ,
  • Stephen C Jones

      Affiliations

    • Department of Microbiology & Immunology, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
    • ,
  • Thea M Friedman

      Affiliations

    • Department of Microbiology & Immunology, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
    • ,
  • Robert Korngold

      Affiliations

    • Department of Microbiology & Immunology, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
    • ,
  • George F Murphy

      Affiliations

    • Departments of Pathology, Anatomy & Cell Biology, Thomas Jefferson University, Philadelphia, Pennsylvania, USA
    • Corresponding Author InformationCorrespondence and reprint requests: George F. Murphy, MD, Department of Pathology, Anatomy & Cell Biology, Thomas Jefferson University, 1020 Locust St., Rm. 545 JAH, Philadelphia, PA 19107–6799, USA

Received 2 July 2003; accepted 1 August 2003.

Article Outline

Abstract 

The identity of cells within squamous epithelia that represent primary targets in acute graft-versus-host disease (GVHD) has been an enigma. Murine effector T cells implicated in the alloresponse by Vβ complementarity-determining region-3 spectratype analysis were detected with a Vβ-specific monoclonal antibody within discrete microdomains of tongue (lingual) squamous epithelium. These microdomains, termed rete-like prominences (RLPs), are similar to the rete ridges of human skin. Cells forming the basal layer of RLPs and of human skin rete ridges were shown to express a distinctive pattern of keratin expression defined by antibodies to cytokeratin 15 (K15). In experimental murine GVHD elicited across minor histocompatibility antigen barriers (miHA), early lesions involved selective apoptosis and loss of K15+ staining within lingual RLPs. An in vitro organ culture model designed to investigate target cell injury by short-term exposure to tumor necrosis factor-α and interleukin-1β, mediators relevant to GVHD, showed a similar pattern of apoptosis and loss of K15+ reactivity within RLPs. In aggregate, these findings establish a novel cytoskeletal marker for target epithelial subpopulations that should facilitate evaluation of mechanisms of host cell injury in GVHD. These data may also enable the development of therapeutic approaches to abrogate disease at the level of target cell blockade.

Keywords:  Acute graft-versus-host disease, Cytokeratin 15, Target cells

 

Back to Article Outline

Introduction 

The pathogenesis of acute graft-versus-host disease (GVHD) involves at least 3 distinct, yet interrelated, evolutionary phases: allostimulation, effector cell homing, and tissue-specific targeting [1]. Although recent cellular and molecular approaches designed to abrogate GVHD have focused on inhibiting pathways that mediate allostimulation and homing, little is known with respect to blockade at the level of host cell targeting. This is in large part related to the fact that the precise identity of host cells that are targeted in GVHD has not yet been resolved. Host cell targeting in both human and experimental GVHD involves epithelial injury elicited by specific subpopulations of allostimulated donor T cells that secrete cytokines systemically and home selectively to liver, gut, skin, and related squamous epithelia. The tissue damage that results has been shown to involve induction of epithelial apoptosis, as opposed to necrosis 2, 3. However, only certain subpopulations of host epithelial cells are vulnerable to apoptotic injury. In squamous epithelia (eg, skin and tongue), cells that reside in the basal layer are preferentially targeted in GVHD. Moreover, histopathologic evidence suggests that early targeting preferentially affects basal cell subpopulations that are concentrated in the tips of discrete downgrowths of the epidermis (rete ridges) [4], in analogous rete-like prominences (RLPs) of dorsal tongue 2, 5, 6, and in bulge regions of hair follicles 7, 8. It is interesting to note that epithelial cell populations in these microanatomical domains also have features of stem cells involved in self-renewal and repair 9, 10, 11.

Experimental murine models in which GVHD is elicited exclusively across minor histocompatibility antigen (miHA) disparities have proven both informative and relevant to human disease 12, 13, 14, 15. However, murine skin is of limited value in elucidating specific microanatomic domains targeted by effector pathways because the epidermis is devoid of the rete ridges implicated as possible primary sites of injury in human skin. Squamous epithelium of murine dorsal tongue, however, possesses structurally similar RLPs that also represent sites of selective injury in GVHD 2, 5. We have found that in disease produced across miHA barriers, RLPs are informative surrogates for epidermal rete ridges in studies focused at determining how cellular and molecular heterogeneity within the basal cell layer may relate to target cell-selective cytotoxicity [2].

A prerequisite to understanding the mechanism or mechanisms whereby specific epithelial subpopulations within the basal layer are preferentially targeted in GVHD is to identify markers selectively expressed by these cells. Recently, a monoclonal antibody (mAb; clone C8/144B) that recognizes human cytokeratin 15 (K15) was used to delineate the bulge region of human hair follicles [16], a known site of early host epithelial cell targeting in GVHD. However, other epithelial subpopulations that are targeted in GVHD (eg, epidermis) did not show K15 expression by the methods used. Although discontinuous epidermal reactivity for K15 in human, murine, and sheep epidermis has been noted by other observers 17, 18, correlation with a specific structural domain or site of immune targeting has not been made. In this study, we have identified methodologies that permit detection of K15+ epithelial domains in human and murine epidermis, as well as in murine dorsal tongue. In human skin, these K15+ regions correspond to rete ridges, and in murine tongue, they correspond to RLPs. Additionally, our data indicate that K15+ RLPs are (1) sites of infiltration by allospecific T cells in experimental murine GVHD, (2) microanatomic sites of early apoptotic injury, and (3) regions preferentially vulnerable to cytokine-mediated apoptosis in an in vitro organ culture model of GVHD-like cytotoxicity. It is concluded that basal cells are heterogeneous with regard to cytokeratin phenotype and related cytotoxic targeting characteristics. Further evaluation of K15+ RLPs should aid in elucidating and potentially abrogating targeting pathways that play critical roles in patient morbidity and mortality.

Back to Article Outline

Materials and methods 

Mice 

Two major histocompatibility complex (MHC)-matched, miHA-disparate strain combinations—B10.D2/oSnJ (B10.D2) → DBA/2Ncr (DBA) (H2d) and C57BL/6ByJ (B6) → C.B10-H2b/LiMcdJ (BALB.B) (H2b)—were used to generate GVHD mediated by CD4+ T cells. B10.D2 and B6 male mice (Jackson Laboratory, Bar Harbor, ME) between the ages of 7 and 12 weeks were used as donors, and DBA (NCI Animal Production Program, Bethesda, MD) and BALB.B (Jackson Laboratory) male mice between the ages of 9 and 16 weeks were used as recipients. BALB/cAnNTac-Rag2tmlN12 (severe combined immunodeficiency disease; SCID) immunodeficient mice (Taconic, Germantown, NY) were used for initial screening of cytokeratin expression because the absence of endogenous antibody minimized background staining when mouse mAbs were used. All mice were housed in a sterile environment in microisolator cages (Lab Products, Maywood, NJ) and provided with autoclaved food and water ad libitum.

Antibodies 

Anti-Thy-1.2 (J1j; ascites fluid; rat immunoglobulin [Ig]M) and anti-CD8 (3.168; rat IgM) mAbs were used along with guinea pig complement (C; Rockland, Boyertown, PA) for cell subset depletion. Goat anti-mouse IgG antibody (ICN Immunobiologicals, Costa Mesa, CA) was used for B-cell panning.

The primary mAbs used for immunostaining were rat anti-murine CD3 (Antigenix America, Huntington Station, NY); rat anti-murine CD4, CD8, and I-Ab,d,q; biotinylated mouse anti-murine Vβ8.1,8.2 T-cell receptor (TCR) and Vβ5.1,5.2 TCR (BD Pharmingen, San Diego, CA); and mouse anti-human K14 (Biogenex, San Ramon, CA), K15, clone LHK15 (Lab Vision, Fremont, CA), and clone C8/144B (Dako, Carpinteria, CA). For evaluation of GVHD and lingual organ culture tissue samples, LHK15 mAb was biotinylated by using a BiotinTag Micro Biotinylation Kit (Sigma Chemical, St. Louis, MO). Control antibodies were rat IgG2a (BD Pharmingen), mouse IgG1 negative control (Dako), or biotinylated mouse anti-human CD45 mAb (BD Pharmingen).

For flow cytometry, rabbit polyclonal anti-mouse K14 antibody directly conjugated to fluorescein isothiocyanate (FITC; Covance Research Products, Richmond, CA) was used, along with donkey anti-mouse F(ab′)2 antibody conjugated to R-phycoerythrin (PE; Accurate Scientific, Westbury, NY) for indirect detection of K15 (clone LHK15) mAb binding.

Preparation of donor cells 

Bone marrow cells were obtained from the femurs and tibiae of donor mice by flushing with phosphate-buffered saline (PBS) supplemented with 0.1% bovine serum albumin (BSA; Sigma) and anti-T cell-depleted bone marrow (ATBM) was prepared by incubation with J1j mAb (1:100) and guinea pig complement (1:25) for 45 minutes at 37°C. T cell-enriched donor cell populations were prepared by treating pooled spleen and lymph node cells with Gey’s balanced salt lysing solution containing 0.7% NH4Cl, followed by panning on plastic petri dishes precoated with goat anti-mouse IgG antibody (40 μg/mL) for 1 hour at 4°C to remove B cells. To purify CD4+ T cells, T cell-enriched preparations were treated with anti-CD8 mAb (1:50) plus C (1:25) for 45 minutes at 37°C initially and again for 30 minutes with a wash in between. Cells were phenotyped by flow cytometry and were routinely found to be negative for CD8 expression.

Bone marrow transplantation and GVHD induction 

Recipient mice were lethally irradiated with a Gammacell (J.L. Shephard, San Fernando, CA) cesium 137 source (10 Gy at 1.3 Gy/min). Approximately 6 hours later, these recipients were injected intravenously with donor cells via the tail vein. The inoculum consisted of either ATBM cells (2 × 106) alone, as a negative disease control, or a mixture of ATBM (2 × 106) and donor purified CD4+ T cells (5 × 106 cells for B10.D2 → DBA and 2 × 107 cells for B6 → BALB.B strain combinations). Morbidity and mortality of mice were monitored daily, and weights were monitored twice weekly.

Tissue samples 

All tissue samples were snap-frozen in Optimal Cutting Temperature medium (Miles Laboratories, Elkhart, IN) for immunohistochemical or immunofluorescent analyses with 5-μm cryostat sections and/or were collected in 10% phosphate-buffered formalin (Fisher Scientific, Malvern, PA) for routine paraffin section histology. Normal human skin samples (facial, breast, and forearm) were collected from routine surgical procedures. Normal SCID back skin and tongue tissues and DBA, B6, B10.D2, and BALB.B tongue tissues were used to evaluate K15 expression. ATBM and GVHD tongue samples were collected from BALB.B recipient mice at days 3, 7 or 11, 15, 22, and 30 after bone marrow transplantation (BMT) (3 mice each per time point for 3 separate experiments) and from DBA recipient mice at days 4, 7, 14, and 21 days after BMT (3 mice each at days 4 and 7 and 2–3 mice each at days 14 and 21 for 3 separate experiments).

Vβ complementarity-determining region-3-size spectratyping 

Vβ TCR complementarity-determining region-3 (CDR3)-size spectratype analysis, a highly sensitive reverse transcriptase-polymerase chain reaction (PCR)-based technique, was used to identify T-cell expansion within specific Vβ families indicated by overrepresentation of individual CDR3-region lengths of Vβ-positive TCRs, as previously described in detail [19]. Briefly, total cellular RNA was prepared from enriched CD4+ T cells (collected 5 days after BMT) and homogenized in 1 mL of Ultraspec (Biotecx Laboratories, Houston, TX). The poly(A)+ portion of the total RNA was converted into complementary DNA by using oligo(dT) as a primer for reverse transcription. PCR was performed by using a 32P-labeled constant primer (Cβb) and a Vβ primer specific for each Vβ family to be analyzed. All the primers used have been previously described 20, 21. PCR products were electrophoresed on a 6% acrylamide sequencing gel. Sequencing gels were dried, and autoradiography was generally performed for 15 hours at room temperature (RT) without intensifying screens. Densitometric scanning of autoradiographs was performed on a Personal Densitometer SI by using ImageQuant software (Molecular Dynamics, Sunnyvale, CA).

Immunohistochemical methods 

Immunohistochemistry was performed with a standard 3-step biotin-avidin-horseradish peroxidase method for all unconjugated rat and mouse primary mAbs. Briefly, cryostat sections or cytospin preparations were acetone-fixed for 5 minutes at −20°C, air-dried, and incubated with one of the following mAbs: rat anti-mouse CD3 (1:50), CD4 (1:400), CD8 (1:400), I-Ab,d,q (1:500), and rat IgG2a control (1:50) or mouse anti-human K14 (1:50), K15, clone LHK15 (1:50 to 1:1000), or clone C8/144B (1:5 to 1:20) and mouse IgG1 control (1:50) for 1 to 1.5 hours at RT or overnight at 4°C. Secondary antibodies (1:200) were biotinylated rabbit anti-rat, mouse-adsorbed (Vector Laboratories, Burlingame, CA) for rat primary mAb and biotinylated horse anti-mouse (Vector) for mouse mAb, followed by biotin-avidin-horseradish peroxidase as the third layer. Immunoreactivity was revealed by using 3,3′-diaminobenzidine (Sigma), Vector VIP (Vector), or NovaRED (Vector) chromagen. For immunofluorescent detection with anti-human K14 and K15 mAbs, SCID mouse tongue sections were fixed in acetone and then incubated with primary mAb (1:50) for 1.5 hours at RT followed by donkey anti-mouse Cy2 conjugate (1:25; Accurate Scientific) for 30 minutes and examined with a Nikon (Melville, NY) Microphot SA fluorescent microscope attached to an Optronics (Goleta, CA) digital camera. For immunodetection with biotinylated anti-Vβ8 and anti-Vβ5 (1:600), an avidin-biotin blocking kit (Vector) was used before primary mAb incubation, followed by amplification with the Renaissance TSA-Indirect Kit (NEN Life Science Products, Boston, MA) and 3,3′-diaminobenzidine development for 3 minutes only. For detection of K15 expression in non-SCID mouse tissues (ie, GVHD tissues and lingual organ culture), sections were fixed with 4% paraformaldehyde (PF; Electron Microscopy Sciences, Ft. Washington, PA) in PBS for 10 minutes at 4°C, washed with PBS 3×, incubated with 0.1% Triton X-100 (Sigma) in PBS for 10 minutes, washed and blocked with an avidin-biotin blocking kit for 20 to 25 minutes each (Vector), and then incubated with either biotinylated anti-human K15 mAb (LHK15; 1:50) or biotinylated anti-human CD45 irrelevant control mAb (1:5) for 1.5 hours at RT. After washing, primary antibody binding was detected with streptavidin/horseradish peroxidase conjugate (1:50; Vector) for 30 minutes at RT and developed by using NovaRED substrate. Sections were counterstained lightly with Gill No. 1 hematoxylin (Fisher) or with methyl green (Biogenex) when Vector VIP substrate was used. In experiments to assess spatial relationships between K15, CD4, and terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labeling (TUNEL; see below) expression, serial sections separated by 5 μm were used.

Detection of apoptotic cells by TUNEL 

Cryostat sections were fixed with 4% PF for 10 minutes at 4°C, washed 3 times with PBS for 5 minutes each, and permeabilized with 70% ethanol at −20°C for 5 minutes. After 3 deionized water washes, sections were incubated with terminal deoxynucleotidyl transferase enzyme in reaction buffer (Serologicals, Norcross, GA) at a ratio of 1 part enzyme to 2 parts buffer at 37°C in a humidified chamber for 1 hour. Sections were then washed and incubated with sheep anti-digoxigenin horseradish peroxidase Fab (1:50; Roche Molecular Biochemicals, Indianapolis, IN) for 30 minutes at RT, developed with NovaRED substrate, and counterstained, as described previously.

Light and electron microscopy 

All tissue samples were analyzed by routine light microscopy (hematoxylin and eosin staining). RLPs were defined as epithelial downgrowths bordered by interdigitating papillae formed by submucosal connective tissue. The border between RLP and inter-RLP basal cells was operationally designated as the midpoint of the distance between the top of the submucosal connective tissue papilla and the tip of the adjacent RLP. Apoptotic cells, as inferred by conventional staining and confirmed by TUNEL, were identified qualitatively by 2 independent observers and consisted of squamous epithelial cells with darkly stained pyknotic or fragmented nuclei surrounded by hypereosinophilic cytoplasm 2, 14.

For ultrastructural analysis, selected murine tongue samples were fixed in Karnovsky’s II fixative [22] at 4°C and stored until further processing. After washing with 0.1 M cacodylate buffer, pH 7.4, tissue samples were postfixed with 2% osmium tetroxide, dehydrated through graded ethanols with propylene oxide as the final step, and embedded in Taab Epon 812 (Marivac, Nova Scotia, Canada). One-micrometer sections were stained with 0.5% toluidine blue in 0.5% sodium borate, and areas were evaluated for further analysis. Ultrathin sections were cut on a Leica (Wetzlar, Germany) Ultracut UCT ultramicrotome, poststained with uranyl acetate followed by bismuth subnitrate, and viewed on a Hitachi (Gaithersburg, MD) H7000 electron microscope.

Isolation of murine lingual basal epithelial cells 

Tongue tissue was collected from killed mice and placed in Dulbecco’s modified Eagle’s medium 1:1 mixture with Ham’s F12 (DMEM/F12; Life Technologies, Grand Island, NY) containing 100 IU of penicillin and 100 μg of streptomycin per milliliter (P/S; Mediatech, Herndon, VA). Thin strips of tongue tissue were prepared and placed at 4°C in DMEM/F12 containing dispase 2 mg/mL (Life Technologies). After overnight incubation, epithelial sheets were collected by separating the epithelium from the underlying connective tissue with the aid of a dissecting microscope and fine forceps. Epithelial sheets were then incubated in 0.25% trypsin/0.1% EDTA (Mediatech) and agitated for 5 minutes at RT followed by an additional 5 minutes of pipetting to separate individual cells. Trypsinization was stopped by the addition of DMEM/F12 with 10% fetal bovine serum (FBS; Mediatech), and the cells were pelleted by centrifugation and washed with PBS. For cytospin preparation, 1% FBS was added to the cell suspension, and slides were prepared by using a Shandon Cytocentrifuge and stored at −80°C until use.

Flow cytometry of lingual basal cells 

Isolated lingual cells were resuspended in PBS and fixed for 10 minutes at RT by the addition of an equal volume of 2% PF. The cells were washed twice with PBS, and approximately 400 000 lingual epithelial cells were aliquoted per well of a 96-well microtiter plate and incubated with rabbit anti-murine K14-fluorescein isothiocyanate conjugate (K14-F), mouse anti- human K15 (LHK15), or K14-F plus K15 for 1.5 hours at RT in PBS containing 2% BSA and 0.025% saponin (Sigma). After washing twice with PBS containing 2% BSA, cells previously incubated with K15 mAb were detected by using donkey anti-mouse PE (1:25). Control cells received either buffer alone or donkey anti-mouse PE alone. Fluorescence analysis was performed on a Beckman Coulter XL-MCL analytic cytometer (Beckman Coulter, Miami, FL) in the Kimmel Cancer Center Flow Cytometry Facility.

Lingual organ culture cytotoxicity assay 

A method to evaluate short-term effects of cytokines on explants of mouse dorsal tongue was developed by modification of an approach used previously for embryonic rat tongue organ culture [23]. DBA mouse tongues were collected in DMEM/F12 with P/S, and thin strips of the dorsal surface, approximately 3 to 5 mm long by 1 to 1.5 mm deep, were prepared with the aid of a dissecting microscope. Explants (2 per condition/time point) were then placed, connective tissue side down, on 0.45-μm HAWP millipore filter paper, 25 mm in diameter (Fisher Scientific), which was supported by a fenestrated plastic insert placed inside 60-mm petri dishes containing DMEM/F12 with 10 % FBS and P/S, in the presence or absence of murine tumor necrosis factor-α (TNF-α), murine interleukin-1β (IL-1β; Peprotech, Rocky Hill, NJ), or both. The 60-mm petri dishes were then placed inside 110 × 15 mm square petri dishes containing 20 mL of sterile water and incubated at 37°C with 5% humidity. Initial experiments were performed for 24 and 48 hours, and time-course studies were collected at 4, 8, 16, and 24 hours. Two concentrations of murine IL-1β plus murine TNF-α, as defined by previously established protocols, were evaluated 24, 25: IL-1β 25 U/mL plus TNF-α 500 U/mL and IL-1β 250 U/mL plus TNF-α 2500 U/mL, as well as IL-1β 250 U/mL alone and TNF-α 2500 U/mL alone. Replicate experiments were performed to confirm all results (5 separate explant experiments were evaluated for assessment of TNF-α plus IL-1β effects).

Back to Article Outline

Results 

Induction of acute GVHD 

Two MHC-matched miHA-disparate strain combinations, B10.D2 → DBA (H2d) and B6 → BALB.B (H2b), were used to generate GVHD mediated by CD4+ T cells. Signs of GVHD (weight loss, hunched posture, ruffled fur, and histologic confirmation in skin and tongue) were present in all animals receiving CD4+ T-cell transplants, but not in animals receiving ATBM only, as previously described 14, 19, 26. Recipient tissue from both strain combinations was comparatively evaluated by using all techniques, and pathologic alterations proved qualitatively identical. However, disease characteristically progressed more rapidly in the B10.D2 → DBA strain combination (median survival time, 11 days) [26] than in the B6 → BALB.B strain combination (median survival time, 58 days). Accordingly, the latter model was used primarily for assessment of sequential T-cell homing patterns.

Allospecific T cells selectively infiltrate RLPs in experimental GVHD 

Experiments were designed to determine whether specifically allostimulated effector T cells, as opposed to T-cell infiltrates in general, exhibit patterns of early homing consistent with RLP-directed targeting. In murine GVHD elicited across miHA barriers in which CD4+ donor T cells are known to be effectors, sequential immunohistochemical evaluation of dorsal tongue revealed progressive and selective infiltration of RLPs by CD4+ T cells during the first 3 weeks after transplantation (Figure 1A-D). Because animals that underwent transplantation received only CD4+ T cells, identical patterns of RLP infiltration were also observed with anti-CD3 mAb, and anti-CD8 staining was consistently negative.

  • View full-size image.
  • Figure 1. 

    Sequential T-cell homing patterns for dorsal lingual epithelium in experimental murine GVHD. A-D, Sequential immunohistochemistry at days 3, 11, 15, and 22, respectively, showing preferential involvement of RLP by T cells (arrowheads; identical patterns were observed with anti-CD3 and anti-CD4 mAb; CD8+ T cells were not detected). In control mouse tongue, the only immunoreactive hematopoietic cells within the epithelial layer were MHC class II-positive dendritic cells localized to RLP domains (A, inset). E, Vβ8-specific primers were used to generate PCR products from control B6 naive splenic CD4+ T cells (left) and from B6 → BALB.B GVHD-reactive CD4+ T cells (right). F, Densitometric scans of these gel films generated histograms to evaluate the size distribution of the CDR3 bands between the B6 (top) and the B6 → BALB.B (bottom) and revealed preferential use (skewing) of 2 bands (arrows; O.D. indicates optical density). G, Immunohistochemistry at the time of peak disease (day 30) with anti-Vβ8 mAb showed positive cells localized within RLPs (arrowheads), whereas staining of adjacent sections for Vβ5 (control) was negative (H).

During allostimulation, specific families of T cells defined by their TCR Vβ use become clonally expanded. These overrepresented subpopulations may be identified by Vβ TCR CDR3-size spectratype analysis [19]. In GVHD that develops in the B6 → BALB.B strain combination, spectratype analysis of CD4+ T cells after transplantation revealed overrepresentation during allostimulation (day 5) of several T-cell Vβ families, including Vβ8 (Figure 1E and 1F), but not Vβ5 (not shown; as previously observed [19]). Immunohistochemistry with anti-Vβ8.1,8.2 and anti-Vβ5.1,5.2 mAb revealed that Vβ8+, but not Vβ5+, T cells preferentially infiltrated RLPs (Figure 1G and 1H). Both Vβ8+ and CD4+ T cells were consistently localized within most (>80%) RLPs, with only rare positively stained cells (<5%) within epithelium constituting the inter-RLP regions. Of the CD4+ T-cell population at the time of peak disease activity (day 30), 60% to 75% were Vβ8+, and <5% were Vβ5+ in adjacent sections.

Lingual RLPs prominently express K15 

Epithelial cells within the bulge region of human hair follicles express K15 [16], and these regions are targeted in GVHD 7, 8. It was thus reasoned that other epithelial target sites (eg, murine RLPs) also may possess a similar or distinctive cytokeratin phenotype. C8/144B mAb, which recognizes K15 [16], was used with a modified immunohistochemical staining protocol. This new protocol differed from those previously reported by (1) use of cryostat sections, (2) fixation in cold acetone or 4% PF, and (3) overnight primary antibody incubation at 4°C. With this protocol, basal cells that formed rete ridges of the epidermal layer in control human skin showed strong cytoplasmic reactivity (Figure 2A and 2B), whereas cells that formed interrete regions were either weakly positive or negative. In addition, basal cells in the outer root sheath and bulge region of hair follicles were also positive for K15, as previously reported [16]. Evaluation of normal mouse tissue (SCID; see Materials and Methods) with this modified technique revealed a similar pattern of K15 reactivity concentrated in basal cells forming RLPs of dorsal tongue (Figure 2C and 2D). However, K15 staining of murine RLPs was less intense and more variable than that observed for rete ridges of human skin. Isolated clusters of basal cells (2–4 cells) within the murine epidermis, as well as bulge regions of hair follicles, were also positive for K15 (Figure 2C, insets).

  • View full-size image.
  • Figure 2. 

    K15 expression in normal human skin and murine lingual epithelium. A and B, K15 staining in normal human skin with clone C8/144B mAb, which recognizes human cytokeratin 15 [16]. Strong preferential basal cell reactivity for K15 was detected in epidermal rete ridges (A, arrows; rare individual cells in the dermis and epidermis represent normally trafficking T cells expressing CD8 that are also recognized as a K15-independent epitope by this mAb [16]). C and D, In murine lingual epithelium, immunoreactivity was detected by C8/144B mAb in the basal cells of RLPs (C, arrows). Apparent faint suprabasal staining proved in serial sections to be an artifact of the plane of sectioning through RLPs. In murine epidermis (C; inset, top left) and hair follicles (C; inset, top right), isolated basal cells and the follicular bulge region were also positive (arrowheads). E, Immunofluorescent detection of K14 showed uniform expression throughout the lingual basal cell layer, whereas K15 immunoreactivity detected with anti-cytokeratin 15 mAb (clone LHK15, which identifies K15 exclusively) demonstrated strong positivity concentrated at the tips of RLPs (F). G, Cytocentrifuge preparation of isolated basal epithelial cells stained with LHK15 mAb revealed a spectrum of reactivity from strongly positive to negative. H, This spectrum of K15 reactivity within the K14+ population was also demonstrated by flow cytometric analysis, where 60.8% of the K14+ epithelial cells showed variable fluorescent intensity for K15. FITC indicates fluorescein isothiocyanate; PE indicates phycoerythrin.

To improve detection of preferential K15 immunoreactivity within the basal cell layer, an additional anti-K15 mAb was screened in normal mouse tissue. LHK15, an antibody generated against the last 17 amino acids of the K15 polypeptide [17], demonstrated markedly enhanced and consistent immunoreactivity for basal epithelial cells restricted to RLPs of mouse tongue (Figure 2F). This pattern was confirmed in all murine strains evaluated (data not shown). K15+ basal cells consistently showed strong staining of uniform intensity concentrated at the tips of RLPs. This staining dissipated in basal cells forming the uppermost edges of RLPs and inter-RLP domains. In contrast, antibodies to K14 showed continuous reactivity involving the entire basal cell layer (Figure 2E). Trypsinized basal cell suspensions of both human skin and murine lingual mucosa revealed a spectrum of K15 immunoreactivity among cells in cytospin preparations ranging from strong to negative (Figure 2G). Flow cytometric analysis confirmed this spectrum of K15 reactivity within the K14+ basal cell layer population (Figure 2H).

K15-positive RLPs are altered by experimental GVHD 

A mainstay for evaluating epithelial GVHD pathology in the miHA murine model has been the detection of dyskeratotic (now known to be apoptotic) target cells 2, 13, 14, 27, 28, 29. By routine light microscopy, apoptotic epithelial cells, distinguished by pyknotic or fragmented nuclei and hypereosinophilic cytoplasm, could be detected within RLPs, but not inter-RLP regions, by days 7 to 11 (Figure 3A). These injured cells were confirmed to be apoptotic and were consistently demonstrated to be restricted to the basal or immediate suprabasal layers of RLPs by TUNEL immunohistochemistry (Figure 3B). Previous studies using immunoultrastructural approaches for correlation of TUNEL with intermediate filament expression have shown such apoptotic cells to be of epithelial origin [3]. In addition, transmission electron microscopy revealed lymphocyte apposition to epithelial cells within RLPs (Figure 3C). Immunohistochemistry of the lingual epithelium at days 4 and 7 revealed progressive infiltration of RLPs by CD4+ T cells (Figure 3D and 3E) associated with focal loss of K15+ staining (Figure 3F and 3G). By day 7 (Figure 3E), foci where CD4+ T cells surrounded larger nonreactive cells were detected within RLPs. In adjacent sections, the centrally located cells in such foci proved to be strongly reactive for K15 (Figure 3G), thus correlating with the cytoarchitecture of satellitosis. In animals receiving only ATBM transplants, tongue samples were indistinguishable from those of normal controls, with no evidence of epithelial T-cell infiltration or apoptosis by the various techniques used.

  • View full-size image.
  • Figure 3. 

    Evaluation of murine GVHD lingual tissue for apoptosis, T-cell infiltration, and K15 expression. A, Light microscopy of hematoxylin and eosin-stained tissue from GVHD at day 11 after BMT reveals lymphocytic infiltration and apparent apoptosis in RLPs (arrow). B, Apoptosis restricted to RLPs was confirmed by TUNEL immunohistochemistry (arrows). C, Transmission electron microscopy demonstrates RLP infiltrated by a lymphocyte (left arrow) in apposition to an epithelial cell undergoing mitotic division (right arrow). D and E, Immunohistochemistry for CD4 at days 4 (D) and 7 (E) after BMT with progressive infiltration of RLPs by CD4+ T cells. Note the focus where CD4+ T cells (E; arrow) surround a centrally located nonreactive cell (E; inset). F and G, K15 expression in sections adjacent to (D) and (E), respectively, reveals focal loss of K15 immunoreactivity (F; arrow). Note the cell strongly positive for K15 and surrounded by nonreactive cells (G; arrow and inset), a pattern reciprocal to the inset in (E) and suggestive of satellitosis.

TNF-α and IL-1β in vitro mimic GVHD effects on K15-positive RLPs 

To determine whether RLP-restricted apoptotic injury and loss of K15 staining could result from cytotoxic mediators of relevance to target cell injury in GVHD (eg, TNF-α and IL-1β) [30], an in vitro organ culture model for murine tongue explantation was developed (Figure 4; also see Materials and Methods). The first alterations were detected 8 hours after culture with TNF-α plus IL-1β with concentrations previously established to produce biological effects in similar explant systems 24, 25, 31, 32. These alterations were focal and subtle (<10% of RLPs affected), consisting of isolated TUNEL+ cells and nonuniform K15 reactivity restricted to the basal layer of RLPs (Figure 4 insets). By 24 hours, clusters of apoptotic cells primarily within the immediate suprabasal layer of most RLPs (>80%) were observed by routine histology and confirmed by TUNEL staining (Figure 4). This was associated with markedly diminished immunoreactivity for RLP K15 in adjacent sections. With TNF-α alone, TUNEL positivity was restricted to single cells rather than cell clusters, whereas TNF-α and IL-1β in combination produced more prominent clusters of TUNEL+ cells within RLPs. RLP-associated patterns of apoptosis and K15 loss were remarkably similar to those documented in GVHD animals (Figure 3). Explants cultured in IL-1β or medium alone failed to show pathologic alterations in K15 distribution or TUNEL staining within RLPs at any time point evaluated.

  • View full-size image.
  • Figure 4. 

    Cytokine-mediated RLP cytotoxicity using an in vitro lingual organ culture model (incubation time, 24 hours). Note the loss of K15 immunoreactivity in RLPs after incubation with TNF-α and with IL-1β + TNF-α (inset in IL-1β + TNF-α/K15 panel, represents an 8-hr time point; note at this earlier time point the subtle diminution of K15 immunoreactivity). TUNEL-positive cells in adjacent sections of explants incubated with TNF-α and IL-1β + TNF-α were restricted to suprabasal regions of RLPs (arrows) (inset in IL-1β + TNF-α/TUNEL panel; 8-hour time point demonstrating early apoptosis affecting a single cell within the basal layer [arrow]). Whereas explants incubated with IL-1β or medium alone were indistinguishable from normal lingual mucosa by K15 and TUNEL immunohistochemistry, those exposed to TNF-α and IL-1β + TNF-α showed patterns of immunoreactivity remarkably similar to those of experimental GVHD (see Figure 3). H&E indicates hematoxylin and eosin.

Back to Article Outline

Discussion 

We have used sequential evaluation of murine tongue to better define specific sites of epithelial targeting in experimental GVHD. RLPs, structures similar to human epidermal rete ridges, were found to be sites of selective infiltration by allostimulated CD4+ T cells, as defined by Vβ8-specific mAb (in the B6 → BALB.B model). Basal epithelium forming RLPs, but not intervening cells, was shown for the first time to express high levels of immunohistochemically detectable K15, a feature shared with GVHD target cells in the basal layer of human epidermal rete ridges. Moreover, epithelial target cell injury in early experimental GVHD results in loss of K15 reactivity and induction of apoptosis preferentially within RLPs. Finally, in a model of murine lingual organ culture, cytokines of established importance in GVHD pathogenesis were capable of inducing GVHD-like alterations within RLP target sites.

In 1985, Sale et al. [4], using morphometric approaches, made the seminal observation that rete ridge keratinocytes may represent preferred targets in cutaneous GVHD. They subsequently hypothesized that similar cell populations restricted to filiform papillae of the tongue [5] may also constitute GVHD targets. Although it is generally accepted that tissue-associated homing to skin, liver, and gut is characteristic of GVHD, the notion of target cell heterogeneity within these tissues is only beginning to be elucidated. Squamous epithelium contains cells representative of both hematopoietic (eg, Langerhans cells) and nonhematopoietic (eg, epithelial cells) lineages. BMT recipient mice chimeric for selected compartments of the epithelial layer (hematopoietic versus nonhematopoietic) have been used in an effort to address this issue. Korngold and Sprent [33] used a CD8+ T cell-mediated B10.BR → CBA miHA-disparate model in which reirradiated chimeric transplant recipients demonstrated significantly diminished GVHD when donor miHAs were expressed exclusively by the hematopoietic compartment. The importance of the nonhematopoietic compartment in GVHD development also has been observed recently in a CD4-mediated miHA model (B6 → BALB.B; S.C. Jones, unpublished data, 2003). Teshima et al. [30], however, evaluated bone marrow chimeras in which either MHC class I or II alloantigen was expressed exclusively by hematopoietic cells, including antigen-presenting cells, and found that acute GVHD does not require alloantigen expression on host target epithelium. Although these results may reflect targeting differences involving miHA versus MHC, they provide evidence that early cellular targeting is likely to be focused at specific cell types within the epithelial layer.

RLPs of murine dorsal tongue are potentially relevant to mechanisms involving both primary hematopoietic and nonhematopoietic targeting. This is in part because the nonhematopoietic elements (RLP epithelial cells) are now shown to be (1) phenotypically distinctive and separable with respect to K15 expression, (2) sites of preferential infiltration by an allospecific Vβ subpopulation of effector T cells, and (3) target regions for apoptotic injury and loss of K15 immunoreactivity. However, it is also of interest that MHC class II-positive dendritic cells of hematopoietic origin are often localized within the suprabasal layers of the RLP, but not within the intervening epithelial domains (Figure 1A inset). In this regard, the possibility that RLPs may serve as microenvironments conducive to localized immune-mediated injury is fortified by recent documentation that in GVHD, these regions selectively express vascular cell adhesion molecule-1 [6], a molecule that mediates both adhesive and co-stimulatory events relevant to allostimulated T cells 34, 35, 36.

Bickenback [11] has demonstrated that basal cells in murine lingual RLPs have features of epithelial stem cells. Lyle et al. [16] have shown that K15 may define cells with characteristics of stem cells in the bulge region of human follicular epithelium. Although speculation exists regarding whether epithelial stem cells in epidermis, hair follicles, and tongue epithelium are primarily targeted in GVHD [37], their localization in GVHD target sites 11, 38 now known in murine tissues to harbor potentially overlapping subpopulations of K15-reactive cells adds additional strength to this possibility. On stimulation or injury, epithelial stem cells are capable of a rapid proliferative burst 10, 39. Proliferating (activated) epithelial cells show diminished expression for K15 17, 40, and epithelial proliferation is an early response in GVHD [41]. In addition, TNF-α has been shown to suppress K15 messenger RNA expression in an immortalized human keratinocyte line [40]. Thus, the loss of K15 expression observed by immunohistochemistry, although it indicates immunopathology that preferentially affects RLP, may represent the effects of apoptosis, epithelial activation/proliferation in response to injury, or cytokine-mediated suppression—either singly or in combination.

In murine GVHD models, both TNF-α and IL-1β are believed to play important roles in disease evolution 30, 42. Inhibition of TNF-α and IL-1β, or of TNF-α alone, may significantly ameliorate or even prevent disease 15, 30, 43. TNF-α interacts with the TNF-RI receptor that contains a death domain capable of inducing apoptosis in many cell types [44], including murine squamous epithelium [45]. In this study, we examined the hypothesis that one or both of these cytokines could produce pathologic lesions that preferentially affect RLPs and, thus, resemble those documented in experimental GVHD. Our data suggest that TNF-α alone, and, to a greater extent, in combination with IL-1β, is capable of producing RLP-directed pathology that has significant qualitative overlap with the phenotype of epithelial target injury seen in experimental disease. Factors other than cytokines, including the Fas/Fas ligand and perforin/granzyme pathways, may also play roles in RLP-directed target cell injury in GVHD, and these possibilities require further study. However, our findings support an important role for TNF-α and IL-1β in inducing RLP-directed epithelial target cell injury and thus are in keeping with similar in vivo results implicating these cytokines as key mediators of epithelial pathology [30].

In summary, microanatomically and antigenically distinctive lingual epithelial cells represent preferred sites of targeting in experimental murine GVHD. Similar targeting sites based on K15 expression within epidermis may also exist, accounting for focality of apoptotic injury. Cytokines relevant to GVHD are capable of selectively affecting the same epithelial domains that are vulnerable in experimental disease. The ability to differentiate target epithelial cells and to study them in in vitro models relevant to human disease should now enable translational approaches that will define and inhibit pathways of target cell apoptosis in GVHD.

Back to Article Outline

Acknowledgements 

This work was supported by National Institutes of Health grant nos. CA 40358 and HL 55593. We are grateful to Dr. Judith C. Kim, Brigette S. Adair, Danielle Castor, Dana Telem, and Paul Hallberg for technical assistance in various aspects of this work.

Back to Article Outline

References 

  1. Gilliam AC, Murphy GF. Cellular pathology of cutaneous graft-versus-host disease. In:  Ferrara JLM,  Deeg HI,  Burakoff SJ editor. Graft-vs-Host Disease.: Immunology, Pathophysiology, and Treatment. New York, NY: Marcel Dekker; 1996;p. 291–313
  2. Gilliam A, Whitaker-Menezes D, Korngold R, Murphy GF. Apoptosis is the predominant form of epithelial target cell injury in acute experimental graft-versus-host disease. J Invest Dermatol. 1996;107:377–383
  3. Yoo YH, Gilliam A, Whitaker-Menezes D, Korngold R, Murphy GF. Experimental induction and ultrastructural characterization of apoptosis in murine acute graft-versus-host disease. Arch Dermatol Res. 1997;289:389–398
  4. Sale GE, Shulman HM, Gallucci BB, Thomas ED. Young rete ridge keratinocytes are preferred targets in cutaneous graft-vs-host disease. Am J Pathol. 1985;118:278–287
  5. Sale GE, Raff RF, Storb R. Stem cell regions in filiform papillae of tongue as targets of graft-vs-host disease. Transplantation. 1994;58:1273–1287
  6. Kim JC, Whitaker-Menezes D, Deguchi M, Adair B, Korngold R, Murphy GF. Novel expression of VCAM-1 (CD106) by squamous epithelium in experimental acute graft-versus-host disease. Am J Pathol. 2002;161:763–769
  7. Sale GE, Beauchamp MD, Akiyama M. Parafollicular bulges, but not hair bulb keratinocytes, are attacked in graft-vs-host disease of human skin. Bone Marrow Transplant. 1994;14:411–413
  8. Murphy GF, Lavker RM, Whitaker D, Korngold R. Cytotoxic folliculitis in GVHD. J Cutan Pathol. 1990;18:309–314
  9. Lavker RM, Sun T-T. Heterogeneity in epidermal basal keratinocytes (morphological and functional correlations). Science. 1982;215:1239–1241
  10. Cotsarelis G, Sun T-T, Lavker RM. Label-retaining cells reside in the bulge area of pilo-sebaceous unit (implications for follicular stem cells, hair cycle, and skin carcinogenesis). Cell. 1990;61:1329–1337
  11. Bickenback JR. Identification and behavior of label-retaining cells in oral mucosa and skin. J Dent Res. 1981;60:1611–1620
  12. Korngold R, Sprent J. Lethal graft-vs-host disease following bone marrow transplantation across minor histocompatibility barriers in mice. Prevention by removing mature T cells from marrow. J Exp Med. 1978;148:1687–1698
  13. Ferrara J, Guillen FJ, Sleckman B, Burakoff SJ, Murphy GF. Cutaneous acute graft-vs-host disease to minor histocompatibility antigens in a murine model (histologic analysis and correlation to clinical disease). J Invest Dermatol. 1986;86:371–375
  14. Murphy GF, Whitaker D, Sprent J, Korngold R. Characterization of target injury of murine acute graft-vs-host disease directed to multiple minor histocompatibility antigens elicited by either CD4+ or CD8+ effector cells. Am J Pathol. 1991;138:983–990
  15. Murphy GF, Sueki H, Teuscher C, Whitaker D, Korngold R. Role of mast cells in early epithelial target cell injury in experimental acute graft-versus-host disease. J Invest Dermatol. 1994;102:451–461
  16. Lyle S, Christofidou-Solomidou M, Liu Y, Elder D, Albelda S, Cotsarelis G. The C8/144B monoclonal antibody recognizes cytokeratin 15 and defines the location of human hair follicle stem cells. J Cell Sci. 1998;111:3179–3188
  17. Waseem A, Dogan B, Tidman N, et al.  Keratin 15 expression in stratified epithelia (downregulation in activated keratinocytes). J Invest Dermatol. 1999;112:362–369
  18. Whitbread IA, Powell BC. Expression of the intermediate filament keratin gene, K15, in the basal cell layers of epithelia and the hair follicle. Exp Cell Res. 1998;244:448–459
  19. Friedman TM, Statton D, Jones SC, Berger MA, Murphy GF, Korngold R. Vβ spectratype analysis reveals heterogeneity of CD4+ T cell responses to minor histocompatibility antigens involved in graft-versus-host disease (correlations with epithelial tissue infiltrate). Biol Blood Marrow Transplant. 2001;7:2–13
  20. Friedman TM, Gilbert M, Briggs C, Korngold R. Repertoire analysis of CD8+ T cell responses to minor histocompatibility antigens involved in graft-versus-host disease. J Immunol. 1998;161:41–48
  21. Casanova JL, Romero P, Widmann C, Kourilsky P, Maryanski JL. T cell receptor genes in a series of class I major histocompatibility complex-restricted cytotoxic T lymphocyte clones specific for a Plasmodium berghei nonapeptide (implications for T cell allelic exclusion and antigen-specific repertoire). J Exp Med. 1991;174:1371–1383
  22. Karnovsky MJ. A formaldehyde-glutaraldehyde fixative of high osmolarity for use in electron microscopy. J Cell Biol. 1965;27:137A; (abstr.)
  23. Mbiene J-P, MacCallum DK, Mistretta CM. Organ cultures of embryonic rat tongue support tongue and gustatory papilla morphogenesis in vitro without intact sensory ganglia. J Comp Neurol. 1997;377:324–340
  24. Stoitzner P, Zanella M, Ortner U, et al.  Migration of Langerhans cells and dermal dendritic cells in skin organ cultures (augmentation by TNFα and IL-1β). J Leukoc Biol. 1999;66:462–470
  25. Messadi DV, Pober JD, Fiers W, Gimbrone MA, Murphy GF. Induction of an activation antigen on postcapillary venular endothelium in human skin organ culture. J Immunol. 1987;139:1557–1562
  26. Korngold R, Jameson B, McDonnell JM, et al.  Peptide analogs which inhibit IgE-FcϵRIα interactions ameliorate the development of lethal graft-versus-host disease. Biol Blood Marrow Transplant. 1997;3:187–193
  27. Van den Brink MRM, Moore E, Horndasch KJ, Crawford JM, Murphy GF, Burakoff SJ. Fas ligand-deficient gld mice are more susceptible to graft-versus-host disease. Transplantation. 2000;70:184–191
  28. Van den Brink MRM, Moore E, Crawford JM, Hoffman J, Murphy GF, Burakoff SJ. Fas-deficient lpr mice are more susceptible to graft-versus-host disease. J Immunol. 2000;164:469–480
  29. Schmaltz C, Alpdogan O, Kappel BJ, et al.  T cells require TRAIL for optimal graft-versus-tumor activity. Nat Med. 2002;8:1433–1437
  30. Teshima T, Ordemann R, Reddy P, et al.  Acute graft-versus-host disease does not require alloantigen expression on host epithelium. Nat Med. 2002;8:575–581
  31. Klein LM, Lavker RM, Matis WL, Murphy GF. Degranulation of human mast cells induces an endothelial antigen central to leukocyte adhesion. Proc Natl Acad Sci U S A. 1989;86:8972–8976
  32. Walsh LJ, Trinchieri G, Waldorf HA, Whitaker D, Murphy GF. Human dermal mast cells contain and release tumor necrosis factor-α which induces endothelial leukocyte adhesion molecule-1. Proc Natl Acad Sci U S A. 1991;88:4220–4224
  33. Korngold R, Sprent J. Features of T cells causing H-2 restricted lethal graft-versus-host disease across minor histocompatibility barriers. J Exp Med. 1982;155:872–883
  34. Elices MJ, Osborn L, Takada Y, et al.  VCAM-1 on activated endothelium interacts with the leukocyte integrin VLA-4 at a site distinct from the VLA/fibronectin binding site. Cell. 1990;60:577–584
  35. Honda S, Campbell JJ, Andrew DP, et al.  Ligand-induced adhesion to activated endothelium and to vascular cell adhesion molecule-1 in lymphocytes transfected with the N-formyl peptide receptor. J Immunol. 1994;152:4026–4035
  36. Damle NK, Klussman K, Linsley PS, Aruffo A. Differential costimulatory effects of adhesion molecules B7, ICAM-1, LFA-3, and VCAM-1 on resting and antigen-primed CD4+ T lymphocytes. J Immunol. 1992;148:1985–1992
  37. Sale GE. Are stem cells attacked in graft-versus-host-disease?. J Cutan Pathol. 1991;18:305–306
  38. Lavker RM, Miller SJ, Sun T-T. Epithelial stem cells, hair follicles and tumor formation. Recent Results Cancer Res. 1993;128:31–43
  39. Cotsarelis G, Cheng S-Z, Dong G, Sun T-T, Lavker RM. Existence of slow-cycling limbal epithelial basal cells that can be preferentially stimulated to proliferate (implications on epithelial stem cells). Cell. 1998;57:201–209
  40. Werner S, Werner S, Munz B. Suppression of keratin 15 expression by transforming growth factor β in vitro and by cutaneous injury in vivo. Exp Cell Res. 2000;254:80–90
  41. Jakic-Razumovic J, Browne MD, Sale GE. Proliferation rates in epidermis of patients with graft-versus-host disease, non-specific inflammation and normal skin. Bone Marrow Transplant. 1992;10:27–31
  42. Piguet PF, Grau GE, Allet B, Vassalli P. Tumor necrosis factor/cachectin is an effector of skin and gut lesions of the acute phase of graft-vs-host disease. J Exp Med. 1987;166:1280–1289
  43. Korngold R, Marini J, de Baca ME, Murphy GF, Giles-Komar J. Role of tumor necrosis factor-α and graft-versus-host disease and graft-versus-leukemia responses. Biol Blood Marrow Transplant. 2003;9:292–303
  44. Tartaglia LA, Goeddel DV. Two TNF receptors. Immunol Today. 1992;13:151–153
  45. Ruckert R, Linder G, Bulfone-Paus S, Paus R. High-dose proinflammatory cytokines induce apoptosis of hair bulb keratinocytes in vivo. Br J Dermatol. 2000;143:1036–1039

PII: S1083-8791(03)00288-X

doi:10.1016/S1083-8791(03)00288-X

Biology of Blood and Marrow Transplantation
Volume 9, Issue 9 , Pages 559-570, September 2003

Article Tools

* Image Usage & Resolution