John D. Beatty, Antigen Presentation Research Group, Imperial College London, Northwick Park and St. Mark's Campus, Harrow, UK. e-mail: firstname.lastname@example.org
To test the hypothesis that dendritic cells (DC), antigen-presenting cells with the potential to stimulate primary T-cell responses, may appear in the urine of patients with bladder cancer, and that their characteristics may reflect those of DC in cancer tissue.
PATIENTS AND METHODS
Cells from digested tissue of transurethral resection specimens from eight patients and urine from 18 with bladder cancers were analysed using flow cytometry, immunohistochemistry and electron microscopy. Urine samples from 12 patients were also analysed during intravesical bacillus Calmette-Guérin (BCG) therapy.
Immature DC positive for major histocompatibility complex class II antigens, negative for markers of other leukocyte lineages and with low levels of co-stimulatory markers, were identified in CD45-positive cells isolated immediately from cancer tissue or amongst cells migrating from tissue fragments after overnight culture. Immature-phenotype DC were also identified in the urine of patients with bladder cancer. Their identity was confirmed by immunohistochemistry and electron microscopy. Using these methods, DC were monitored from the bladder during BCG installation for bladder cancer in 12 patients for a mean of 10 months. Of six patients who developed a recurrence of their bladder cancer over this period, all but one showed a lower percentage of DC in their urine at the end of their initial treatment.
We identified DC in the urine of patients with bladder cancer for the first time. We speculate that variability in the percentage of urinary DC may reflect changes in immunological activity at the tumour site; prospective studies are required to evaluate the relevance of these DC counts and characteristics to clinical outcome.
The presence of high numbers of leukocytes in the urine of patients undergoing BCG therapy correlated with a reduction in superficial bladder cancer recurrence . We hypothesized that cells arising from abnormal tumour epithelium may contain dendritic cells (DC), and that these cells migrate/leak into the urine. DC act as ‘sentinels’ of the immune system in the tissues where they acquire antigens. They can process the antigens, become activated, mature and migrate into lymphoid tissue, where they present antigen to lymphocytes to produce a specific immune reaction; this response may be either tolerogenic or immunogenic  DC differ from macrophage and monocyte antigen-presenting cells because they are potent initiators of primary naïve T-cell responses.
To recognize tumours as an immunological ‘foreign’ threat, DC must be able to promote immune responses to tumour antigen. Tumours arise from self-tissues and may evade detection because they lack ‘non-self’ antigens and have genetic instability . Despite these escape mechanisms, the immune system can recognize proteins or peptides expressed differentially by tumour and normal cells, known as tumour-associated antigens. In advanced cancers, cytotoxic lymphocytes generated against specific tumour-associated antigens can be identified and have been used in the development of tumour vaccines . Specific cytotoxic lymphocytes can be induced by using autologous DC which have been matured, stimulated and manipulated with different cocktails of cytokines, and then pulsed with tumour cells, their lysates or tumour-associated antigens . However, no DC vaccine has been reproducibly successful in preventing tumour progression and ensuring survival in humans. This failure may be because DC from patients with advanced cancer have a decreased ability to stimulate a normal T-cell response . Despite limited clinical success, there are sporadic cases of spontaneous and engineered tumour regression, and even cure . This finding underlines the importance of understanding, measuring and monitoring immune mechanisms.
BCG is the most effective immunotherapy for bladder cancer but the precise mechanism by which intravesical BCG reduces bladder cancer recurrence remains unknown, even though it has been used as immunotherapy for almost 30 years . The effectiveness may be a result of a change in the immunological activity of cells, and in this respect DC are critical in initiating and maintaining primary immune responses. In vitro studies have shown that BCG activates natural killer cells  and that BCG-activated DC are potent activators of T cells . Could BCG therapy in bladder cancer be exerting its anticancer therapeutic effects via DC? BCG can induce maturation and activation of DC, possibly via its interaction with pattern recognition receptors such as Toll-like receptor (TLR)-2 and -4 found on the surface of these cells [11–13]. By characterizing DC within untreated tumours and monitoring changes during therapy it might be possible to identify potential mechanisms for stimulating or blocking tumour immunity. Recently, new approaches to treating bladder cancer using genetically modified BCG and DC have been used in an attempt to skew the immune system to a cytotoxic rather than a tolerogenic response [14,15]. With new immunotherapies there is a need to monitor and understand the effect of immunotherapy at the cancer/urine interface. Thus we describe, for the first time, a noninvasive technique for monitoring immune responses in bladder cancers by identifying and characterizing DC within the urine of patients with bladder cancer.
PATIENTS AND METHODS
Samples of TUR specimens from eight patients with bladder cancer were obtained to identify DC in bladder cancer and to investigate their migratory capacity. The voided urine from 18 patients with endoscopically or histologically confirmed bladder cancer was collected before surgery or therapy with intravesical BCG. Of the 18 patients, 12 had their urine DC monitored during intravesical BCG therapy by obtaining a further specimen before their sixth and final installation of BCG. Blood samples were taken from 24 patients with bladder cancer to compare urine and blood DC. A two-tailed Student's t-test was used to analyse differences in the results in the cells from patients treated with BCG. The hospital ethics committee approved the study and written consent was obtained from patients.
Tissue specimens obtained at the time of surgery were collected in complete medium: RPMI 1640 (Sigma Chem Co., Poole, Dorset) supplemented with 10% fetal calf serum (FCS), 2 mmol/L l-glutamine and penicillin/streptomycin (100 U/mL). The tissue was dissected with surgical scalpels into 1–2 mm pieces and digested using 1 mg/mL collagenase D (Roche Molecular Products, Welwyn Garden City, UK) in RPMI 1640 containing 20 µg/mL DNase 1 (Roche) and 2% FCS. Digestion was continued on a shaker at 37 °C for 2 h. Mononuclear cells were separated (700 g, 30 min, room temperature) on a low-density gradient (Ficoll-Paque; Amersham Pharmacia, UK) and washed in complete medium.
Migratory cells were recovered from tissue samples based on the method described by Mahida et al.. Tissue specimens were prepared as before, placed in a Petri dish, covered with 5 mL of complete medium and incubated for 16 h at 37 °C in 5% CO2. Tissue pieces were then removed and migratory cells recovered in three separate experiments.
Mid-stream or sterile-catheter urine specimens were collected, centrifuged, the pellet of cells resuspended in medium and separated over a Ficoll-Paque gradient, as described for cells from tissue. Peripheral blood samples were collected in Vacutainer tubes containing heparin (Becton Dickinson, Plymouth, UK).
Antibodies to HLA-DR (clones L243, Immu357), CD3 (SK7), CD8 (SK1), CD11c (S-HCL-3), CD45 (2D1) and CD123 (945) were purchased from Becton Dickinson (Oxford, UK). DC-SIGN (CD209) was purchased from R&D Systems (Minneapolis, USA). CD3 (UCHT1), CD8 (B9.11), CD14 (RMO52), CD16 (3G8), CD19 (J4.119), CD34 (Immu133), CD45 (J.33), CD56 (N901 (NKH1)), DC-LAMP (104.G4) and Langerin (DCGM4) were purchased from ImmunoTech (Westbrook, ME, USA). HLA-DR (DK22) was purchased from DAKO (Carpinteria, CA, USA). Biotin 10 nm gold was purchased from British Biocell International (Cardiff, UK). Isotype-control antibodies were obtained from the same manufacturers.
Cells suspended in complete medium were labelled with antibody and left for 30 min at 4 °C in the dark and then washed in PBS supplemented with 1 mmol/L EDTA, 0.02% sodium azide and 2% FCS (FACS buffer) and resuspended in 4% paraformaldehyde in PBS. Using a unique combination of techniques for quantitative flow cytometry, data were acquired uncompensated using a Coulter EPICS® XL flow cytometer (Beckman Coulter, UK). To compensate samples, a region was drawn around the lymphocyte population based on their forward/side-scatter properties. Lymphocytes were stained with anti-CD45, anti-CD3 or anti-CD8 surface markers and evaluated in each channel using the WinListTM 5.0 program (Verity Software House, USA) to compensate the test sample. Because there were few cells and problems with autofluorescence and staining in the urine specimens, blood samples were taken from patients at the time of urine collection and used for compensation. This approach was based on the whole-blood technique described by Sewell et al.. Briefly, samples of blood were collected in Vacutainer tubes containing heparin, and 50-µL aliquots placed into four test tubes and labelled with the anti-CD8 surface marker for each channel and left at 4 °C for 30 min. To compare the efficacy of detecting the few DC using compensation on urine or blood cells, urine was labelled with anti-CD45 as above, in conjunction with the blood labelling in five experiments. Red blood cells were then lysed using Optilyse C (ImmunoTech). By comparing urine and blood compensation using identical analytical methods, there was a 14% mean decrease in the number of DC detected using urine-compensated specimens, probably because there were fewer cells available and differences in staining with the anti-CD45 antibody, leading to inadequate compensation. Therefore, urine specimens were compensated with the compensation values obtained using the blood technique in subsequent analyses. Super-enhanced Dmax statistics were used for channel-by-channel analysis of flow cytometry data using the WinListTM 5.0 package. The Dmax and Kolmogorov-Smirnov statistics were used to assess the significance of a positive subtraction result by calculating the Dcrit. Only Dcrit values with P < 0.05 were accepted and lower values were regarded as zero .
To show that urine DC were not blood DC contaminants, blood samples from 24 patients with bladder cancer were collected and labelled with monoclonal antibodies. Blood DC were identified as above for urine and tissue DC. Percentages of mononuclear cells that were identified as DC are given as the mean (sd) %.
ELECTRON MICROSCOPY AND IMMUNOHISTOCHEMISTRY
Between 100 000 and 500 000 of the Ficoll-Paque separated urine cells were labelled with anti-HLA-DR, followed by goat anti-biotin 10 nm gold, fixed with 3% glutaraldehyde in 0.1 mol/L sodium phosphate buffer and processed for electron microscopy. Immunohistochemistry was performed on cytospins of the separated urine cells. Protein expression was identified using a standard avidin-biotin peroxidase complex method . Cells were incubated with the primary antibodies (HLA-DR, DC-SIGN, DC-LAMP and Langerin) for 1 h at room temperature, followed by further incubation in biotinylated monoclonal antimouse IgG (1/200, Vector) for 30 min, and a final layer of avidin-biotin peroxidase complex (VectorStain Elite kit, Vector) also for 30 min. Positively labelled cells stained brown when 3,3-diaminobenzidine was used to develop the peroxidase.
In single-cell suspensions of bladder cancer, HLA-DR positive cells were identified in the large cell population, based on their forward/side-scatter profile. This indicated that there might be DC in the tissue, and therefore further experiments used multi-channel flow cytometry. Cells were characterized as DC using three-channel flow cytometry; DC were identified as CD45+, HLA-DR+ and negative for markers of other lineages (Fig. 1). Furthermore, we identified bladder cancer DC in the supernatants of tissue pieces left overnight in complete medium; in samples from six patients, 4.7 (2.7)% of mononuclear cells were DC. The co-stimulatory molecule CD86 and the maturation marker CD83 were not significantly expressed in tissue DC or supernatant DC, defining an immature phenotype. These results suggested that bladder cancer DC might be able to migrate from the tumour, and raised the possibility that they migrate into the urine of patients with cancer, but without maturing. Urine DC were present in all 18 patients with bladder cancer, as 2.1 (2.9)% of the mononuclear cells. Figure 2 compares DC in a representative sample of a patient with bladder cancer to one with no cancer; although cells fell within the DC gate in the latter, there were fewer total events.
The myeloid marker CD11c was expressed on 42–99% of urine DC (15 samples). There was minimal expression of the co-stimulatory molecules CD86, and the maturation marker CD83, defining an immature DC phenotype consistent with that of bladder cancer tissue DC and migratory DC (Fig. 3). CD123 was not significantly expressed in five urine specimens tested, indicating an absence of CD123+ plasmacytoid DC in the urine. Blood from 24 patients with bladder cancer had a lower mean (sd) percentage of DC than urine, at 0.74 (0.27)% vs 2.1 (2.9)%, and all samples contained a subset of DC that were CD123+/CD11c− cells, indicating the presence of plasmacytoid DC which were not present in the urine (data not shown).
Electron micrographs confirmed the presence of DC in the urine by showing their characteristic veils and a cytoplasm that contained few granules and vacuoles, and were phenotypically immature type 2 DC with a phenotype similar to that of skin Langerhans cells (Fig. 4) . The urine DC comprised 4–6% of the separated urine cells on electron microscopy (two samples). The urine also contained cells that labelled with antibodies to HLA-DR, Langerin, DC-SIGN and DC-LAMP (Fig. 4). DC-SIGN is a DC-specific adhesion molecule used to bind to T cells, and DC-LAMP is active in the processing of exogenous antigen. Langerin is expressed on immature epithelial DC.
In the 12 patients treated, there was a significant increase in the percentage of CD45+ cells in the urine at the end of treatment (P = 0.03). Of the 12 patients who had urine samples taken before and during treatment, six developed local recurrence of their bladder cancer. Five of these patients had a lower percentage of DC in the urine before the sixth instillation of BCG (P = 0.14; Fig. 5A). In the group who did not develop recurrence, four patients had a greater percentage of DC in their urine samples (P = 0.14; Fig. 5B).
All urine samples collected from patients with bladder cancer contained DC. The cells were identified and shown to have an immature phenotype using flow cytometry. This identification was validated by electron microscopy and immunohistochemistry. These urine DC were similar to those obtained from cancer tissue. We confirmed the presence of infiltrating DC in bladder cancer tissue and showed that DC may migrate from bladder cancer tissue ex vivo and remain phenotypically immature.
There has been a marked increase in the use of DC-based immunotherapy in patients with cancer, but the fundamental biology and function of DC within and around human tumours is poorly understood. DC cancer-vaccine trials in humans have been limited to patients with advanced cancers who have had little in the way of clinically measurable responses . The poor clinical outcome may be a result of large tumour burdens that are immunologically insurmountable. It would therefore seem logical to study a form of immunotherapy that has therapeutic benefit in patients with a low tumour burden. BCG therapy for superficial bladder cancer fits this model. However, before we assess what effect BCG might have on DC in bladder cancer, we need to confirm that there are tumour-infiltrating DC. Immunohistology has previously been used to identify DC in bladder cancer tissue and found them to be minimally activated . The present study supports these findings, with the identification of DC in bladder cancer using tissue digestion and flow cytometry. Furthermore, we showed the continued migratory capacity of the DC, which was not accompanied by an up-regulation in co-stimulatory or maturation markers. By contrast, skin DC and colonic DC up-regulate maturation and co-stimulatory markers when left to migrate from tissue samples [21,22]. This discrepancy may be explained by studies showing that tumour cells can inhibit the up-regulation of maturation and co-stimulatory molecules on DC . These immature DC phenotypes may lead to a tolerogenic response to tumour antigen by inducing T-cell anergy or regulatory T cells .
DC in the urine were difficult to detect because there were so few. To identify this population, a novel combination of techniques using flow cytometry was developed. The advantages of this method are that changes in urine DC could be measured during immunotherapy with BCG, with no need to disrupt the urothelium by taking a biopsy. Contaminating circulating blood DC could not be excluded, as microscopic and even macroscopic traces of blood are commonly found in the urine of patients with bladder cancer. This blood is likely to arise from the site of the tumour epithelium and may reflect the local tumour environment. Despite possible blood contamination the urine, DC composition differs from that of blood. By showing a higher mean percentage of DC, and no CD123+/CD11c− plasmacytoid subset of DC, in urine than in blood, there is little doubt that DC from the bladder tumour environment constitutes a proportion of the DC identified in urine.
The precise role of plasmacytoid DC in tumour immunology is yet to be fully elucidated. They have been shown to infiltrate solid tumours , express TLR9 and produce interferon-α when stimulated by CpG oligonucleotides . Polarization of a favourable immune response may be possible by stimulating this subset of DC . The absence of plasmacytoid DC in the urine may relate to a functional down-regulation of TLR9, as seen in head and neck cancer , but this has not been described in bladder cancer.
As a preliminary trial of this technique in clinical practice we analysed the urine DC in 12 patients having intravesical treatment with BCG for superficial bladder cancer. Patients who subsequently developed bladder cancer recurrence had a lower percentage of DC in the urine at the end of therapy (Fig. 5A). Whether or not this proportional change of DC represented a real decrease in numbers was not established. Many patient and environmental factors confound the determination of precise DC numbers, including patient renal function, hydration status, urine volume, tumour size and haematuria, and therefore percentages of DC as a proportion of the urine mononuclear cell population were analysed. While this trend was not significant, the data raise the possibility that BCG might stimulate DC migration in patients who respond to therapy. If this migration to the urine was mirrored by the migration of DC to lymph nodes via the afferent lymph, a favourable anti-tumour response in draining lymph nodes might follow.
Other studies have identified the T-helper 1 cytokine interleukin-2 as a predictor of BCG response [27–29]. Increased levels of interleukin-2 in the urine and plasma have been correlated with the success of BCG. Interleukin-2 is a T-cell growth factor produced by T cells, and may act by augmenting a favourable T-helper 1 response to tumour antigen. This represents the product of a process initiated by BCG; we propose that DC may be the initiators/inhibitors of this response and therefore fundamentally more important. Currently interleukin-2 can be measured with commercially available kits, while identifying DC is more complicated.
In conclusion, DC identified in the urine were phenotypically immature and similar to the DC that had migrated from cancer tissue ex vivo. The immaturity of urinary DC was consistent with their lack of ability to stimulate allogeneic T cells to proliferate, recorded in two experiments (data not shown). The allogeneic stimulation, although interesting, was difficult; conditions in the urine and cell separation may have adversely affected DC, and because mononuclear cells were used as the stimulant, of which only a small percentage were DC. Prospective randomized studies are now required to evaluate urine DC in monitoring and managing patients, and as a basis on which to adapt BCG installation schedules to individual patients.
J.D. Beatty is supported by a Royal College of Surgeons (England)/British Urological Foundation fellowship and a grant from the Ralph Shackman Trust. Prof S.C. Knight is funded by a MRC grant.
CONFLICT OF INTEREST
None declared. Source of funding: BUF/RCSE Grant, MRC Grant and Ralph Shackman Fellowship.