Karin Pfister and Jürgen Radons contributed equally to this work.
Patient survival by Hsp70 membrane phenotype†
Association with different routes of metastasis
Article first published online: 19 JUN 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 4, pages 926–935, 15 August 2007
How to Cite
Pfister, K., Radons, J., Busch, R., Tidball, J. G., Pfeifer, M., Freitag, L., Feldmann, H.-J., Milani, V., Issels, R. and Multhoff, G. (2007), Patient survival by Hsp70 membrane phenotype. Cancer, 110: 926–935. doi: 10.1002/cncr.22864
Deutsche Forschungsgemeinschaft (MU 1238 of 7-2, G. Multhoff), European Union TRANSNET (MRTN-2003-504917, G. Multhoff), SFB-455 (R. Issels, V. Milani), Multimmune GmbH (G. Multhoff).
- Issue published online: 2 AUG 2007
- Article first published online: 19 JUN 2007
- Manuscript Accepted: 2 MAY 2007
- Manuscript Revised: 30 APR 2007
- Manuscript Received: 18 DEC 2006
- clinical trial;
- prognostic marker;
- overall survival
Heat shock proteins (HSPs) play important roles in tumor immunity. The authors prospectively investigated the correlation between the tumor-specific Hsp70 membrane expression as an independent clinicopathological marker and overall survival in tumor entities that differ in their route of metastasis.
Hsp70 membrane expression was examined by flow cytometry in 58 colon, 19 gastric, 54 lower rectal carcinoma, and 19 squamous cell carcinoma specimens and the corresponding normal tissues at time of first diagnosis. Kaplan-Meier survival curves were analyzed to determine the relation of Hsp70 expression to the patients' prognosis.
An Hsp70 membrane-positive phenotype was found in 40% (colon), 37% (gastric), 43% (lower rectal), and 42% (squamous cell) of the analyzed tumor specimens. None of the corresponding normal tissues was found to be Hsp70 membrane-positive. In patients with colon (P = .032) and gastric (P = .045) carcinomas, an Hsp70 membrane expression correlated significantly with an improved overall survival; a negative association was seen in lower rectal (P = .085) and squamous cell carcinoma (P = .048).
The authors hypothesized that differing relations between surface expression of Hsp70 on tumor cells and clinical outcomes may reflect differences in the route of metastases. Colon and gastric carcinomas metastasize into the liver where hepatic natural killer cells may have the capacity to recognize and kill Hsp70 membrane-positive tumor cells and thus account for a better overall survival. Cancer 2007; 110:926–35. © 2007 American Cancer Society.
Identification of early, molecular markers that can predict cancer-related mortality is an urgent and important goal in the development of cancer diagnostics and therapeutics. Ideally, identifying patients with high-risk carcinomas with a strong metastatic potential at primary diagnosis would provide a basis for an early and individually tailored therapy that could improve clinical outcome. Such biomarkers would be especially valuable for high-risk tumors of the gastrointestinal tract or lung because those cancers are the most common causes of cancer-related mortality. Current evidence shows that changes in the expression levels of numerous molecules including tumor promoter genes, angiogenic factors, apoptotic and cell cycle genes could be related to cancer severity and mortality.1–4 However, many of these relations were based upon retrospective studies, and the predictive value of the relations in a clinical setting has frequently been untested or insignificant. Furthermore, biomarkers that predict metastatic potential of primary tumors or that are predictive for individual patients remain to be discovered.
Heat shock proteins (HSPs) are potentially biomarkers for cancer severity or mortality. In particular, HSPs with molecular weights of 70 and 90 kDa are attractive candidates because they function in the cross-presentation of tumor-specific peptides and, thus, are potent regulators of the adaptive immune system.5–8 Even in the absence of peptides, the major stress-inducible Hsp70 can promote the innate immune responses to tumor cells by stimulating cytokine production by antigen-presenting cells.9 In addition, the expression of Hsp70 on the surface of tumor cells can also play an immunomodulatory role by increasing tumor cell lysis by natural killer (NK) cells when in the presence of proinflammatory cytokines.10–12 Nearly all tumor cells actively secrete Hsp70 in exosomes that stimulate migration and cytolytic activity of NK cells.13 The unusual membrane expression of Hsp70 occurs only on the surface of tumor cells, but not on normal tissues,14, 15 and, therefore, can be considered to be a tumor-specific marker. Furthermore, transplantation of NK cells into tumor-bearing mice can promote NK cell lysis of Hsp70-expressing tumor cells in vivo.16 These observations suggest that surface expression of Hsp70 by tumor cells could indicate a more favorable prognosis because that expression would promote NK-mediated lysis of these tumor cells. However, an Hsp70 membrane-positive phenotype has also been associated with a higher metastatic potential and an unfavorable prognosis in malignant melanoma17 and acute myeloid leukemia.15 These latter findings suggest that elevated Hsp70 membrane levels may predict a more severe cancer pathology, creating a dilemma for relating Hsp70 surface expression by tumor cells to the severity of the disease.
In the present investigation, we examine the prognostic value of Hsp70 membrane expression in a panel of colon, gastric, lower rectal, and lung carcinomas by using an Hsp70-specific antibody that specifically detects membrane-bound Hsp70 on viable tumor cells. The tumor entities were chosen because colon, gastric, lower rectal, and squamous cell lung carcinomas differ significantly with respect to their routes of metastases. In agreement with previous findings on malignant melanoma and leukemia,15, 17 an Hsp70 membrane-positive phenotype was associated with poorer clinical outcome in lower rectal and in squamous cell carcinomas of the lung. In contrast, patients with Hsp70 membrane-positive colon and gastric carcinomas showed a significantly improved survival. These apparently contradictory results may be attributable to differences in the route of metastasis of the carcinomas that we investigated because the site of metastasis is one of the most important predicting factors for survival in cancer. Colon and gastric carcinomas, but not squamous cell and lower rectal carcinomas, preferentially metastasize to the liver from the mesenteric venous system.18 This route of metastasis may provide an opportunity for depletion of tumor cells with Hsp70 surface expression by hepatic NK cells that would not occur for metastatic cells from squamous cell or lower rectal carcinomas and that would not occur for tumor cells lacking an Hsp70 surface expression. Thus, these findings support our hypothesis that NK cells in the liver may provide an “immunological filter” for tumor cells coming from the splanchnic system.
Tumor specimens and corresponding normal tissues were obtained from patients with colon, lower rectal, and gastric carcinoma who were undergoing surgical resection at time of primary diagnosis at the University Hospitals Munich, Regensburg, and at the Hospitals in Kronach, Hemer, and Fulda in the years 1999–2002 according to the recommendation of the German Society of Surgery.19 The study cohort included 112 patients with colon and lower rectal carcinomas (74 men and 38 women ranging in age from 40.9 to 85.4 years; median, 64.3 ± 10.1), and 19 gastric cancer patients (9 men and 10 women ranging in age from 67.5 ± 11.5 years; range, 32.1–78.6). Only patients with a meticulously complete follow-up record were enrolled. The median follow-up time for colorectal cancer patients was 33.1 ± 14.4 months (range, 2.8–59.6 months); that of gastric cancer patients was 21.5 ± 11.7 months (range, 4.2–50.4 months).
The study also included 14 men and 5 women ranging in age from 44.8 to 76.2 years (median, 61.0 ± 8.2 months) that were diagnosed for squamous cell carcinomas of the lung in the Hospital Barmherzige Bruder (Regensburg, Germany) and the Lung Hospital Hemer (Hemer, Germany). The median follow-up time was 30.6 ± 10.1 months (range, 9.3–40.7 months). All patients were censored in survival analyses according to date last seen. Patients with tumor recurrence or preoperative therapy were not included in this study.
Histopathological diagnosis was made according to the World Health Organization (WHO) classification system for tumors of the colorectal, gastrointestinal tract, and the lung. In addition, the tumors were staged and graded by standard histological analysis according to guidelines of the International Union against Cancer (UICC).20 This study was performed according to the Declaration of Helsinki and approved by the Human Ethics Committee of the University of Regensburg (Germany). All patients enrolled in the study signed an informed consent form.
Tumor and normal tissues derived from the same patient were minced mechanically in phosphate-buffered saline (PBS). After washing, homogenates were passed through a sterile mesh in RPMI 1640 medium supplemented with 10% fetal calf serum (FCS), 1 mM sodium pyruvate, antibiotics (all from Gibco-BRL, Eggenstein, Germany), and 2 mM L-glutamine (PAN Systems, Aidenbach, Germany). For flow cytometry, 0.2–1 × 106 cells were incubated for 30 minutes at 4°C in the dark either with fluorescein 5-isothiocyanate (FITC)-labeled antihuman Hsp70 mAb (cmHsp70.1; Multimmune GmbH, Munich, Germany) or with FITC-labeled IgG1 (BD Pharmingen, Heidelberg, Germany) as an isotype-matched control antibody, and with a fibroblast-specific antibody ASO2 (Dianova, Hamburg, Germany). After washing in PBS conditioned with 2% FCS, the cells were analyzed directly by flow cytometry performed on a FACSCalibur instrument (Becton Dickinson, Franklin Lakes, NJ) in the presence of 1 μg/mL propidium iodide (PI; Sigma-Aldrich, Deisenhofen, Germany). Fluorescence of 10,000 gated events was measured on a 1024 channel, 4 decades, log scale through forward light scatter (FSC) and linear scale through right angle scatter (SSC). The fluorescence histograms were generated by using gated data. Data acquisition and analysis were performed automatically by CellQuest Pro software (Becton Dickinson). Only viable, PI-negative cells were analyzed.
For statistical analysis, the SPSS software for Windows Release 12.0 (SPSS, Chicago, Ill) was used. Two groups were compared by using the multiple proportional hazard model (Cox regression). Differences were regarded significant at P < .05. The survival times within each subgroup were estimated from Kaplan-Meier curves and compared by log-rank test.21 Continuous variables are given as means ± standard deviation as indicated. A walking cutoff was run from the lowest to the highest level of Hsp70 expression on the single-cell suspensions of the solid tumor samples as shown previously for leukemia.22
Between August 1999 and June 2002, this study enrolled 112 patients with newly diagnosed colon and lower rectal carcinoma, 19 patients with gastric carcinoma, and 19 patients with squamous cell carcinoma of the lung without any preoperative therapy. Patients resected with positive margins or with evidence of local recurrence directly after surgery were excluded from the study.
The clinical course of our patient cohort was analyzed by calculating Kaplan-Meier curves for overall survival of colorectal carcinoma (n = 112) and stratifying the data for their UICC stages. As expected, overall survival was associated with the clinicopathological UICC stage in colorectal (P = .001) (Fig. 1). Tumor invasion, lymph node status, and metastases also have been found to be consistent with these observations (Tables 1 and 3).
|Factors||Category||Total No. of patients||Hsp70 membrane-positive tumors (n = 23)||P|
|Age||Total||58||66.1 ± 9.9|
|Hsp70−||35||67.2 ± 10.3||NS|
|Hsp70+||23||64.5 ± 9.1||NS|
We also examined the Hsp70 membrane phenotype on single-cell suspensions of the different tumor samples by flow cytometry. For these studies, we used mAb cmHsp70.1 (Multimmune GmbH, Munich, Germany), which specifically recognizes the major stress-inducible Hsp70 and does not cross-react with the constitutive Hsc70, as determined by Western blot analysis, flow cytometry, and dot blot analysis. By pepscan analysis, the epitope of this antibody was identified as being TKDNNLLGRFELSG (amino acids 450–463), a sequence located within the C-terminus of the Hsp70 protein, which is exposed to the extracellular milieu of tumor cells.23 A sample was regarded as being Hsp70 membrane-positive when greater than 10% of the cells were stained positively after subtraction of the control fluorescence. This cutoff value was evaluated by walking cutoff analysis (P < .05), as described previously.22 Apart from the normal connective tissue, which was determined by using the ASO2 antibody, all tumor cells were found to be Hsp70 membrane-positive at comparable cell surface densities. Figure 2 summarizes 4 typical examples of an Hsp70 membrane-positive phenotype (gray graph) on colon, gastric, lower rectal, and squamous cell carcinomas compared with corresponding normal tissue (white graph). The mean fluorescence intensity (mfi) differed significantly between Hsp70 membrane-positive (46 ± 15) and Hsp70 membrane-negative (22 ± 10) samples, but it did not differ significantly between the different tumor entities.
Patients' characteristics, including tumor size and/or invasion, lymph node status, metastases, grading, and UICC stage of all tumor types associated with the Hsp70 status, are summarized in Tables 1–4. In total, 40% of the colon (23 of 58) and 37% (7 of 19) of the gastric tumor biopsies were found to be Hsp70 membrane-positive. In colon carcinomas, an Hsp70 membrane-positive phenotype was found to be associated with an N0 lymph node status (P = .039), lack of metastases (P = .045), and a lower UICC stage (P = .039) at diagnosis (Table 1). Adjuvant chemotherapy was given only to colon carcinoma patients in stage pN1; none of the patients received radiotherapy. In gastric carcinomas (n = 19), none of the clinicopathological parameters correlated significantly with an Hsp70 membrane-positive phenotype (Table 2); none of these patients received adjuvant chemotherapy and/or radiotherapy.
|Factors||Category||Total No. of patients||Hsp70 membrane-positive tumors (n = 7)||P|
|Age||Total||19||67.5 ± 11.5|
|Hsp70−||12||69.9 ± 7.0||NS|
|Hsp70+||7||63.4 ± 16.6||NS|
Forty-three percent (23 of 54) of the lower rectal and 42% (8 of 19) of the squamous cell tumor biopsies of the lung were found to be Hsp70 membrane-positive. It is worth mentioning that in our squamous cell carcinoma panel, 56% were diagnosed as grade 3 and 44% as grade 2. In lower rectal carcinomas (n = 54), an Hsp70 membrane-positive phenotype was found to be associated with a lack of lymph node metastases (P = .044) and lower UICC stage (P = .017) (Table 3) similar to colon carcinomas (Table 1). Adjuvant chemo-radiotherapy was given to lower rectal carcinoma patients in stage pN1. Squamous cell carcinomas of the lung (n = 19) did not show any significant correlation between the Hsp70 phenotype and clinicopathological parameters (Table 4). Postoperative squamous cell carcinoma patients in UICC stages <2 and ≥2 were treated following standard adjuvant chemotherapy guidelines.
|Factors||Category||Total No. of patients||Hsp70 membrane-positive tumors (n = 23)||P|
|Age||Total||54||62.4 ± 10.0|
|Hsp70−||31||60.4 ± 8.6||NS|
|Hsp70+||23||65.0 ± 11.3||NS|
|Factors||Category||Total No. of patients||Hsp70 membrane-positive tumors (n = 8)||P|
|Age||Total||19||61.9 ± 8.2|
|Hsp70−||11||61.6 ± 9.0||NS|
|Hsp70+||8||62.3 ± 7.5||NS|
We next analyzed overall survival in Hsp70 membrane-positive and membrane-negative tumor patients by using the multiple proportional hazard model (Cox regression). Patients with Hsp70 membrane-positive colon (Fig. 3A; n = 58, P = .032) and gastric (Fig. 3B; n = 19, P = .045) carcinomas showed a significantly better overall survival when compared with their Hsp70 membrane-negative counterparts.
These findings were not expected because previous results indicated that in malignant melanoma and leukemia, an Hsp70 membrane-positive phenotype was associated with poor prognosis.15, 17 In line with these previous findings, but in contrast to colon and gastric cancer patients, patients with Hsp70 membrane-positive lower rectal (Fig. 4A, n = 54; P = .085) and squamous cell carcinomas of the lung (Fig. 4B, n = 19; P = .048) had a poorer clinical outcome with respect to overall survival when compared with their Hsp70 membrane-negative counterparts. The latter data are consistent with previous observations showing membrane-expressed Hsp70 in 50% of endometrial carcinomas with poor clinical prognosis and significantly decreased survival rates.24
Both the Hsp70 phenotype and the UICC stages could be determined as significant independent negative prognostic markers for the overall survival in lower rectal carcinomas (UICC, P = .021, hazard rate 6.684; Hsp70, P = .012; hazard rate 5.633) in the multiple proportional hazard model (Fig. 5A). The P-value for the equality of survival distributions for the different levels of UICC and Hsp70 was .007.
In colon carcinoma patients, the UICC stage also served as a significant, independent, negative prognostic marker for overall survival (UICC, P = .001; hazard rate 13.882). In this tumor entity, the Hsp70 membrane expression was found to be associated with a better overall survival and, thus, was predictive for a positive clinical outcome (Fig. 5B). Because of the low patient number, a detailed analysis of UICC stages and Hsp70 for gastric (n = 19) and squamous cell carcinomas (n = 19) was not performed.
The present investigation findings show that surface expression of Hsp70 by tumor cells may have clinical value as a predictive biomarker for metastatic carcinomas. However, a surprising outcome of our study is that an Hsp70 surface-positive phenotype can be associated with a positive or negative prognosis, depending on the cancer entity examined. We found that an Hsp70 membrane-positive phenotype was associated with a better overall survival in patients with colon and gastric carcinomas but predicted worse overall survival in lower rectal and squamous cell carcinomas of the lung. Thus, our results emphasize that Hsp70 may serve multiple roles in the pathophysiology of metastatic diseases and that specific cancers in individuals can differ in their surface expression of Hsp70 for reasons that are not yet understood.
Our current understanding of the role of NK cells in cancer immunobiology provides a mechanistic context to interpret some of our findings because a growing body of evidence indicates that NK cells can preferentially target and kill tumor cells that express Hsp70 on their cell surfaces. Experimental findings, using both in vitro and in vivo assays, show that when Hsp70 is administered in combination with proinflammatory cytokines, it can modulate the response of NK cells to tumor cells.10, 16 Incubation of NK cells with soluble Hsp70 protein12 or with the 14 mer Hsp70 peptide TKD25 enhances the cytolytic activity of NK cells and the IFN-γ secretion by NK cells when administered in the presence of IL-2. Recent finding that these immune modulatory functions of Hsp70 can also be achieved by release of Hsp70 surface-positive exosomes from tumor cells13 indicated that these exosomes can provide signals to NK cells remote from tumors, to attract them to locate and lyse tumors. This has now been confirmed experimentally in studies that demonstrated that Hsp70 membrane-positive tumors release Hsp70 surface-positive exosomes that are chemoattractive and activate NK-cell cytolytic activity. Antibodies to the Hsp70 peptide TKD have the capacity to block chemoattraction of NK cells and prevent granzyme B-mediated killing of tumor cells.13, 26, 27
Our findings that Hsp70 surface expression by tumor cells is associated with a positive prognosis in colon and gastric cancers are consistent with data that show NK cells selectively target and lyse Hsp70 membrane-positive tumor cells. However, this interpretation is apparently contradicted by findings that Hsp70 surface expression is not associated with a positive outcome in lower rectal or squamous cell carcinomas of the lung. Although experimental data to interpret these findings are not available, we propose that differences in routes of metastasis of squamous cell and lower rectal carcinomas to colon and gastric cancers may underlie the different relations between Hsp70 expression and clinical outcome. Whereas colon and gastric carcinomas metastasize preferentially to the liver,18 metastasis of lower rectal tumors occurs directly by way of the vena cava to the lung, bypassing the liver.28 In addition, metastases of squamous cell carcinomas of the lung do not pass the liver; instead, they metastasize to regional lymph nodes and to the brain.29, 30 Thus, a potentially important feature in determining whether tumor cells with Hsp70 surface expression are correlated with a positive or negative prognosis may be the route of their metastases.
The passage of metastasizing tumor cells through the liver may be important in determining whether these cells are eliminated because the liver functions as an “immunological filter” for circulating tumor cells and infectious organisms in portal venous blood that are constantly drained from the splanchnic venous system.18 The liver also plays a key role in the innate immune response by influencing the homing and differentiation of NK cells.31–34 Up to 65% of all lymphocytes present in the normal liver consists of NK cells, NKT cells, and γδ T cells, and the frequency of NK cells among human intrahepatic lymphocytes ranges between 25% and 30%.31, 32 Furthermore, NK cells are located in the liver sinusoids, adherent to the liver sinusoidal endothelial cells and, thus, are well situated to eliminate arriving metastasizing tumor cells.35 Hepatic NK cells have a higher cytolytic capacity against tumor cells than blood or spleen NK cells. They also express high levels of the C-type lectin receptor CD94,36 which is involved in the interaction of NK cells with tumor membrane-bound Hsp70.37
Hsp70-specific immune responses in NK cells require proinflammatory cytokines (unpublished observation). Among these cytokines, IL-15 has emerged as a key regulator for activation, recruitment, proliferation, differentiation, and survival of NK cells.30, 38, 39 This may be important in the modulation of the immune response by the liver because Kupffer cells are a rich source of IL-15.40, 41 Thus, the liver can provide an environment for NK cells in which they can selectively eliminate colon and gastric carcinoma cells that express Hsp70 on their surfaces. The better overall survival demonstrated by our Kaplan-Meier analyses may reflect NK cell-mediated killing of tumor cells as they transit the liver.
Although our hypothetical model is sufficient to explain the positive relation of an Hsp70 membrane-positive phenotype in metastasizing colon and gastric carcinomas, the model does not explain why Hsp70 surface expression reflects a worse outcome in lower rectal and squamous cell carcinomas. Previous investigators have also shown that Hsp70 expression was found to correlate significantly with poorer clinical outcome in endometrial carcinomas,24 in which the route of metastasis is entirely extrahepatic. Presumably, the predictive value of an Hsp70 surface expression for a worse clinical outcome in cancers that have an extrahepatic route of metastasis reflects the diverse and complex roles played by Hsp70.
HSPs are over-expressed in a broad range of human tumors and play crucial roles in tumor invasion, metastasis, cell proliferation, differentiation, and cell death (for review see ref. 8).8 Depending on their location, HSPs either mediate protection against lethal damage induced by exogenous stress or exert immune activation as danger signals in cancer immunity. In the cytosol, HSPs function as molecular chaperones supporting folding and transport of a great variety of polypeptides and proteins under both physiological conditions and after chemical or physical stress stimuli. Hsp70, the major stress-inducible HSP, is able to protect cells from a wide range of apoptotic and necrotic stimuli.42–45 In the absence of Hsp70 and cytokine-activated NK cells, any of these numerous Hsp70 functions may dominate the immunostimulatory role by Hsp70 in tumor cells. Unpublished data from our own laboratory further indicate that tumor cells exhibiting an Hsp70 surface-positive phenotype are significantly more resistant to chemically and physically induced stress stimuli when compared with their Hsp70 surface-negative counterparts. We hypothesize that, in the absence of cytokine/TKD-activated NK cells, membrane-bound Hsp70 helps the tumor to be protected from lethal damage induced by chemotherapy and/ or radiotherapy and, thus, results in a worse clinical outcome. Comparative analysis of the lipid composition of tumors with a differential Hsp70 membrane expression revealed a higher membrane rigidity in Hsp70 membrane-positive tumors, which may cause better chemotherapy resistance and/or radiotherapy resistance. In summary, we hypothesize that chemotherapy resistance and/or radiotherapy resistance may not correlate with resistance to an NK cell-mediated immunity,
Although many questions remain concerning the immunobiology of NK cell interactions with tumor cells that express Hsp70 on their cell surface, recent and continuing studies indicate that these interactions can be productively exploited for diagnostic and therapeutic goals. The observation that the adoptive transfer into tumor-bearing mice of Hsp70-stimulated NK cells caused the eradication of primary tumors and metastases16, 46, 47 provides strong support for the therapeutic potential of this approach.
Cytokine-activated NK cells can induce regression of established lung and liver tumors.48–51 On the basis of these findings in mouse models, we tested the effects of delivering ex vivo cytokine and Hsp70 preactivated NK cells to patients with therapy-refractory, multiple, metastasized colon and nonsmall cell lung carcinomas. As demonstrated in this phase I clinical trial, reinfusion of ex vivo-stimulated, autologous NK cells is safe, feasible, and well tolerated.52 Furthermore, we showed that it is possible to stimulate an Hsp70-reactivity in NK cells in multiple metastasized and pretreated patients. Although refractory to standard chemotherapy, 2 of 5 patients who received more than 4 NK cell reinfusions showed clinical responses. These promising immunological results and clinical responses justify additional studies in patients with lower tumor burden and an established Hsp70 membrane-positive phenotype.
- 4TNM independent prognostic factors in lung cancer. Rays. 2004; 29: 373–376., , , et al.
- 20SobinLH, WittekindCh, editors. TNM Classification of Malignant Tumours, 6th Edition. New York: Wiley; 2005.
- 28Magen-Darm-Trakt. In: BraunJ, DormannA, editors. Klinikleitfaden Innere Medizin. Munchen: Urban & Fischer; 2001: 256–312., , .
- 29Tumoren der Bronchien und der Lunge. In: BerdelWE, BohmM, ClassenM, DiehlV, KochsiekK, SchmiegelW, editors. Innere Medizin. Munchen: Urban & Fischer; 2004: 542–551., , .
- 30Lung. In: KumarV, AbbasAK, FaustoN, editors. Robbins and Cotran's Pathologic Basis of Disease. Oxford: Elsevier Saunders; 2005: 757–766., , .