Pathology – Research and Practice 210 (2014) 147–154
Contents lists available at ScienceDirect
Pathology – Research and Practice journal homepage: www.elsevier.com/locate/prp
Original article
Digital image analysis of inflammation markers in colorectal mucosa by using a spatial visualization method Elzbieta Kaczmarek a,∗ , Tomasz Banasiewicz b , Agnieszka Seraszek-Jaros a , Piotr Krokowicz c , Marcin Grochowalski c , Przemysław Majewski d , Jakub Zurawski e , Jacek Paszkowski b , Michał Drews b a
Department of Bioinformatics and Computational Biology, Pozna´ n University of Medical Sciences, Poland Department of General Surgery, Oncologic Gastroenterologic Surgery and Plastic Surgery, Pozna´ n University of Medical Sciences, Poland c Department of General and Colorectal Surgery, Pozna´ n University of Medical Sciences, Poland d Department of Clinical Pathomorphology, Pozna´ n University of Medical Sciences, Poland e Department of Biology and Environmental Protection, Pozna´ n University of Medical Sciences, Poland b
a r t i c l e
i n f o
Article history: Received 6 July 2013 Received in revised form 29 October 2013 Accepted 14 November 2013 Keywords: Ulcerative colitis IL-1 IL-6 TNF-␣ Digital image analysis
a b s t r a c t The aim of this study was to apply the spatial visualization method of digital images to quantitative analysis of pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel excised because of colitis ulcerosa in relation with selected clinical symptoms. Our preliminary study included 17 patients having undergone restorative proctocolectomy. Immunohistochemistry was performed for IL-1, IL-6 and TNF-␣. The area fraction and intensity fraction of the cytokines studied were determined by digital image analysis. The results were then categorized using Alfred Immunohistochemistry Score. The expression of IL-1, IL-6 and TNF-␣ was significantly higher in the rectum than in colonic segments (p < 0.01), and was associated with the patients’ clinical condition. The method of quantitative immunohistochemistry presented here allows for searching associations between the expression of biomarkers and clinical symptoms. Evaluation of inflammatory cytokines could be recommended in the active stage of the disease with present symptoms of bloody and mucus stools. A higher expression of IL-1, IL-6 and TNF in samples beyond large intestine correlates with an intensified clinical course of the disease. In patients without bleeding and mucus symptoms present in stools, no significant correlations were found. Therefore, the assessment of cytokines during remission or clinically silent stage might not be useful. © 2013 Published by Elsevier GmbH.
Introduction Research interest in a possible connection between inflammation and cancer has been growing since 1863, when Rudolf Virchow noted the leucocytes in neoplastic tissues and indicated that chronic inflammation supports cancerogenesis [3]. Mediators of the inflammatory response, e.g., cytokines and chemokines, induce the accumulation of immune cells and their activation within the inflamed tissue. This process may lead to the release of reactive oxygen and nitrogen species that can induce genetic and epigenetic changes including point mutations in tumor
∗ Corresponding author at: Department of Bioinformatics and Computational Biology, Poznan´ University of Medical Sciences, Dabrowski Street, 79, 60-529 Poznan, Poland. Tel.: +48 61 854 6909. E-mail address: [email protected] (E. Kaczmarek). 0344-0338/$ – see front matter © 2013 Published by Elsevier GmbH. http://dx.doi.org/10.1016/j.prp.2013.11.007
suppressor genes, DNA methylation and post-translational modifications. This sequence of changes in critical pathways responsible for maintaining normal cellular homeostasis can lead to the development and progression of cancer. Several chronic inflammatory diseases contribute to an increased risk of cancer. Observations that many malignancies are associated with chronic infection and inflammation support this hypothesis, for instance, inflammatory bowel disease is associated with colon cancer. Inflammatory bowel diseases, i.e. Crohn’s disease and colitis ulcerosa, are associated with increased rates of colon adenocarcinoma [30]. Tumor necrosis factor can initiate the inflammatory reactions of the immune system, and induces the production of other cytokines (e.g. IL-1, IL-6 and IL-8) and cytotoxic factors (e.g. nitric oxide, reactive oxygen species) by macrophages, which can mediate tumor suppression. Members of the IL-1 family played significant roles in many aspects of the cancerous process as key regulators of the balance between inflammation and immunity in the tumor microenvironment [11,32].
148
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
The importance of IL-1, IL-6 and TNF in the severity of inflammation in ulcerative colitis has been confirmed in clinical and experimental studies [1,12,13]. Increased levels of those proinflammatory cytokines are detected in active inflammatory bowel disease and correlate with the severity of inflammation, indicating that these cytokines may play a key role in the development of inflammatory bowel disease [24]. Analysis of the inflammed mucosa from patients with Crohn’s disease and ulcerative colitis have shown increased expression of certain proinflammatory cytokines such as IL-1, interleukin IL6 and TNF-␣ [19]. Early production of tumor necrosis factor is prominent in the initiation of a biologically highly complex system involving chemokines, cytokines and endothelial adhesions that recruits and activates neutrophils, macrophages and lymphocytes at the sites of infections. IL-1 and TNF-␣ increase the expression of high-affinity adhesion molecules on endothelial cells, stromal cells and leukocytes and, by this, promote infiltration of inflammatory cells from the blood into tissues. Dysregulation of tumor necrosis factor has also been implicated in a wide variety of autoimmune diseases, including Crohn’s disease; however, how tumor necrosis factor mediates disease-causing effects is not completely explained. The induction of proinflammatory genes by tumor necrosis factor has been linked to most diseases [5]. TNF␣, together with IL-1 and IL-6, is responsible for the development of clinical symptoms and stimulation of acute-phase proteins in ulcerative colitis [8]. Evaluation of severity of inflammatory lesions is an essential component in clinical staging of colitis ulcerosa. Immunohistochemistry (IHC) is an important technique for biomarker validation. It allows direct visualization of biomarker expression in histologically relevant regions of the examined tissue. Our search continues for objective markers enabling unambiguous possible assessment of severity of inflammatory lesions in the course of ulcerative colitis, potential risk of dysplasia and in consequence – of malignant transformation. Traditionally, pathologists examine visually and assign scores for IHC data in a semi-quantitative fashion incorporating the intensity and the distribution of specific staining. Due to staining heterogeneity of the tumor cells, the pathologist’s visual scoring is fraught with problems due to subjectivity in interpretation. Automated IHC measurements promise to overcome these limitations. Therefore, we decided to apply a spatial visualization method for quantitative evaluation of IHC data. Thus, the aim of this study is to present an application of the spatial visualization method of digital images for quantitative analysis of pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel excised because of colitis ulcerosa in relation with selected clinical symptoms.
Materials and methods Our preliminary study included 17 patients having undergone restorative proctocolectomy with J-pouch and temporary loop ileostomy or Hartmann’s colectomy with end ileostomy and preservation of a short rectal stump. The patients were operated on at the Department of General, Gastroenterological and Endocrinological Surgery and at the Department of General and Colorectal Surgery of the University of Medical Sciences in Poznan. Indications for surgical treatment included: exacerbating clinical symptoms (diarrhea, lower digestive tract bleeding, progressive loss of body weight, malnutrition, megacolon, fulminating colitis). Patients’ age in the study was on average 43 ± 14 years (range: 21–69 years), and the males to females ratio was 9:8. The mean duration of symptoms was 58 ± 31 months (range: 12–120
months). All patients were treated without the use of steroids agents and biological therapy, as well as immunosuppression in the last 2 months before surgery. The study excluded patients with other systemic diseases such as diabetes or rheumatoidal diseases. Inclusion criteria reduced the number of patients, but resulted in greater homogeneity of the group and reduced the potential impact of other factors on the level of investigated cytokines. Rectal mucosa was much more altered on macroscopic inspection, showing visible thinning of bowel wall, numerous ulcers and indicators of active bleeding. Mucosal samples of rectum and colon were collected from surgical specimens directly after resection, up to a maximum of 20 min after artery ligations. Bowel was cut in longitudinal line. The size of specimens was standardized (5 mm × 5 mm), specimens were taken from the transversal colon and from the middle part of the rectum, 5–6 cm below the dental line. Then, specimens were fixed in buffered 10% formalin (pH 7–7.8) for 48 h at stable temperature (22 ◦ C) (air-conditioned room) and processed in a standard fashion. Paraffin sections were stained with hematoxylin and eosin (H + E), and only blocks with the most pronounced lesions were selected for further immunohistochemical studies. The specimens were then histologically evaluated for severity of inflammation and dysplasia. Inflammation was scored on a scale of 0–2 [14], whereby: • 0 – lack of active inflammation (no indices of cryptitis); • 1 – mild signs of acute inflammation (cryptitis present in less than 50% of crypts); • 2 – moderate signs of acute inflammation (cryptitis present in more than 50% of crypts); • 3 – severe inflammation (ulcers and defects of mucosal epithelium). Microscopic study of tissue samples revealed moderate severity of inflammatory process: mean score 2.1 (range: 0–3).
Immunohistochemical studies Tissue samples were processed using the streptavidin-biotinperoxidase technique (LSAB kit, DAKO, K0675). Antigens were exposed in a water bath at a temperature of 96 ◦ C in a citrate buffer (pH 6.0) during 90 min. Activity of endogenous peroxidase was inhibited using 3% H2 O2 . Subsequently, preparations were incubated overnight at room temperature with antibodies anti-IL-1 (MAB263), anti-IL-6 (MAB261) and anti-TNF-␣ (MAB226). During subsequent 30 min incubation, biotinylated antibody was used, which was then incubated with peroxidase–streptavidin complex. Between consecutive incubations, preparations were flushed with TBS buffer (pH 7.6) using DAB-3.3 chromogen (SIGMA-ALDRICH, D5637), and the antigen was located. Preparations were stained with hematoxylin and after dehydration cover glass was applied. Reaction devoid of primary antibody was used as negative control for immunocytochemical studies. Histological slides were examined with Olympus DP-12 microscope coupled to a digital camera. Color microscopic images of size 2048 × 1536 pixels were acquired and archived using 40× objective (at least 10 images in every slide with an immunopositive reaction). Using the software DP12-BSW, images were downloaded directly from the camera control unit to the PC. To assess the expression of markers studied within inflammatory infiltrates, a quantitative analysis was used. The area fraction and markers’ intensity fraction were determined by using a spatial visualization technique [15]. The received results were then categorized by using Alfred Immunohistochemistry Score [2] in the following way:
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
149
• the proportion score (PS) was assigned on the basis of the approximate percentage of positively stained cells into four categories (0: no positive cells, 1: >0–1%, 2: >1–10%, 3: >10–33%, 4: >33–66% and 5: >66% positive cells), • the intensity score (IS) of the staining product was graded into four categories: no staining (0), weakly positive staining (1), i.e. the computed intensity was less than 33%, intermediate staining intensity (2), i.e. the computed intensity >33–66%, strongly positive staining (3), i.e. the intensity was greater than 66%.
Table 1 Clinical condition of the study group.
The total score (TS) is TS = PS + IS, TS range 0, 2–8. In addition, the expression of markers within inflammatory infiltrates was evaluated by pathologists with the use of the points scoring system proposed by Bernstein et al. [4], whereby:
WBC – leukocytes, Hgb – hemoglobin, RBC – erythrocytes, ESR – erythrocyte sedimentation rate.
• 0 – denotes lack of expression in infiltrating cells; • 1 – less than half of 10 field examined presents infiltrates with reaction products, weak expression; • 2 – the expression is present in 1–10 cells in each visual field (moderate expression); • 3 – the expression is present in 11–30 cells in each visual field; • 4 – the expression is visible in over 30 cells in each visual field. Microscopic assessment was performed using a 40-fold magnification. Quantitative analysis of micrographs The IHC were quantitatively evaluated using image-processing methods, including a spatial visualization of the markers’ expression [15,16]. First, raw color images were converted to a HSI color space (hue, saturation, intensity). Then, intervals for the colors’ hue, saturation and intensity representing the specific IHC reactions were derived from the histogram of color distribution. Pixels representing the specific markers were converted into voxels by introducing brightness as their third dimension. The expression of the specific marker was revealed on a spatial view by reducing the scenery behind to a background using brightness thresholding. The background pixels were filled with white color, and this resulted in a flat, horizontal surface, while colors representing the specific marker remained unchanged (Figs. 1–3). Objects representing specific markers were then extracted by thresholding in HIS color space. Their surface was computed as the total number of voxels, which belonged to the extracted objects. To compute the area fraction of the extracted objects, the spatial objects were then orthogonally projected onto a plane. The number of color pixels representing the specific marker was divided by the size of the view field (i.e. the size of the image) and represented in percent. The intensity of the positive reaction was computed as the average of the brightest and darkest color values in the selected object, i.e. (max(H,S,I) + min(H,S,I))/2 and also expressed as the percent of the highest possible intensity value (i.e. the darkest color of the reaction). Then, the area fraction and intensity results were categorized using the Alfred Immunohistochemistry Score [2] in the following way: • the proportion score (PS) was assigned on the basis of the approximate percentage of positively stained cells into four categories (0: no positive cells, 1: >0–1%, 2: >1–10%, 3: >10–33%, 4: >33–66% and 5: >66% positive cells), • the intensity score (IS) of the staining product was graded into four categories (0: no staining, 1: weakly positive, i.e. the computed intensity was less than 33%, 2: intermediate, i.e. the computed intensity >33–66%, 3: strongly positive, i.e. the intensity was greater than 66%).
Clinical condition of patients
Mean ± SD
Duration of symptoms (months) Number of bowel movements WBC (103 /l) Hgb (g/100 ml) RBC (106 /ml) ESR Total protein (g/dl) Platelets (×109 /l)
58 12 8.1 12.8 4.3 24 6.7 353
± ± ± ± ± ± ± ±
31 4 3.9 1.9 0.6 14 0.9 149
Minimum
Maximum
12 5 2.2 8.2 3.1 3 4.4 108
120 20 17 15.9 5.8 48 8.5 600
The total score (TS) is TS = PS + IS, TS range 0, 2–8. Statistical analysis Two observers first analyzed the expression of studied cytokines in a pilot sample of 50 images. To evaluate inter- and intra-observer variability of the measurements, the mean difference in the measurements of the two observers was first calculated, as well as their respective average, and 95% confidence intervals. The 95% agreement limits were evaluated according to the method proposed by Bland and Altman [6]. The internal consistency of the results was evaluated on the basis of Cronbach’s alpha coefficient, with values >0.8 being considered indicative of good reliability [7,9]. Descriptive statistics of all results included mean values and standard deviations. The Mann–Whitney test was used to compare the results between independent groups. A correlation between expression of the studied cytokines and progression of symptoms was assessed by Spearman rank correlation coefficient. A p value of less than 0.05 was considered to be statistically significant. The power of the tests used to detect hypothetical differences between colon and rectum was also estimated. The data were analyzed using Statistica v.10.1 (Statsoft, Inc., http://www.statsoft.com). Intra-observer variability was assessed by using Prism v.5 (GraphPad Software Inc., http://www.graphpad.com). Results Comparisons between the measurements of area fractions of studied cytokines obtained by the two evaluators (inter-observer variability by Bland and Altman method) were found to be dispersed around the mean in 95% agreement limits (from −1.173 to 1.232) with bias equal 0.029. The internal consistency (Cronbach’s alpha) was significantly high: 0.947 with 95% lower confidence limit 0.746. The clinical condition of the patients is presented in Table 1. Microscopic evaluation of tissue samples revealed a moderate severity of the inflammatory process: mean score 2.1 (range: 0–3). The mean score reflecting severity of proctitis was 2.7 (range: 1.53), while that of colitis – 1.7 (range: 0–2.7). A quantitative evaluation of the area fraction and the intensity fraction by using the spatial visualization technique is presented in Table 2. Alfred IHC proportion scores, intensity scores and total scores derived from the computed area and intensity fractions are also presented in Table 2. The intensity of the IHC reactions was comparable for all studied cytokines. The expression of IL-1, IL-6 and TNF-␣, measured on the basis of area fraction of positive reaction, was significantly lower in the rectum than in colonic segments (Table 3). However, Alfred IHC total scores assigned to raw quantitative results of the cytokine expression were not significantly different for IL-6 (Table 3). The expression of the studied cytokines assessed by the spatial visualization technique
150
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
Fig. 1. Expression of IL-1 in colon (upper row) and rectum (lower row): raw image (left), 3D view of Il-1 (middle) and extracted reaction (right).
Fig. 2. Expression of IL-6 in colon (upper row) and rectum (lower row): raw image (left), 3D view of Il-6 (middle) and extracted reaction (right).
Fig. 3. Expression of TNF-␣ in colon (upper row) and rectum (lower row): raw image (left), 3D view of TNF-␣ (middle) and extracted reaction (right).
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
151
Table 2 Summary statistics of studied cytokines in rectum and colon. Area fraction (Mean ± SD) IL-1 rectum IL-1 colon IL-6 rectum IL-6 colon TNF-␣ rectum TNF-␣ colon
0.8 1.8 1.0 1.5 0.9 2.6
± ± ± ± ± ±
0.7 1.1 0.8 0.9 0.5 1.5
Intensity fraction (Mean ± SD) 53 53 53 53 52 54
± ± ± ± ± ±
Alfred IHC proportion score (PS) (Mean ± SD)
7 6 7 6 8 5
1.2 1.7 1.4 1.7 1.3 1.9
± ± ± ± ± ±
0.2 0.5 0.5 0.5 0.5 0.2
Alfred IHC intensity score (IS) (Mean ± SD) 2 2 2 2 2 2
± ± ± ± ± ±
0 0 0 0 0 0
Alfred IHC total score (PS) (Mean ± SD) 3.2 3.7 3.4 3.5 3.3 3.9
± ± ± ± ± ±
0.4 0.5 0.5 0.6 0.5 0.2
Table 3 Differences between the area fraction and Alfred IHC total score of studied cytokines in rectum and colon. Cytokines
Area fraction (Mean ± SD)
IL-1 rectum IL-1 colon IL-6 rectum IL-6 colon TNF-␣ rectum TNF-␣ colon
0.8 1.8 1.0 1.5 0.9 2.6
± ± ± ± ± ±
0.7 1.1 0.8 0.9 0.5 1.5
Test M–W (p-value)
Test power (%)
0.003
76
0.047
47
0.001
91
Alfred IHC total score (TS) (Mean ± SD) 3.2 3.7 3.4 3.5 3.3 3.9
± ± ± ± ± ±
0.4 0.5 0.5 0.6 0.5 0.2
Test M–W (p-value)
Test power (%)
0.003
83
0.601
50
0.001
94
Fig. 4. A correlation between the expression of IL-1 in rectum of male patients vs. content of hemoglobin (r = −0.6833, p < 0.05), RBC (r = −0.6666, p < 0.05) and total protein (r = −0.7615, p < 0.05).
Fig. 5. A correlation between the expression of IL-1 in rectum of patients with bleeding symptoms vs. content of hemoglobin (r = −0.4519, p < 0.05), and total protein (r = −0.7586, p < 0.05)
was associated with the patients’ clinical condition and clinical symptoms. In particular, an increasing expression of IL-1 in rectum of male patients was correlated with decreasing number of red blood cells (r = −0.6666, p < 0.05) and decreasing content of hemoglobin (r = −0.6833, p < 0.05) and total protein (r = −0.7615, p < 0.05) (Fig. 4). In females, these correlations were not statistically significant. In patients with bleeding symptoms (independently on gender), a higher expression of IL-1 in rectum was correlated with decreasing content of hemoglobin (r = −0.6648, p < 0.05) and total protein IL-1 (r = −0.7586, p < 0.05) (Fig. 5). In patients without bleeding symptoms, these correlations were not statistically significant. In patients with presence of mucus, an increasing expression of TNF-␣ in colonic segments was significantly correlated with decreasing number of RBC (r = −0.4519, p < 0.05) and total protein content (r = −0.6336, p < 0.05) (Fig. 6). In patients without mucus, these correlations were not significant. In patients with extraintestinal symptoms, an increasing expression of IL-6 in colonic segments was significantly correlated with lower number of white blood cells (r = −0.7352, p < 0.05) and content hemoglobin (r = −0.9062, p < 0.05) (Fig. 7). Moreover, in cases with extraintestinal symptoms, the RBC count significantly decreased when
Fig. 6. A correlation between the expression of TNF-␣ in colon of patients with mucus vs. RBC (r = −0.6648, p < 0.05), and total protein (r = −0.6336, p < 0.05).
152
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
Fig. 7. A correlation between the expression of IL-6 in colon of patients with extraintestinal symptoms vs. WBC (r = −0.7352, p < 0.05), and content of hemoglobin (r = −0.9062, p < 0.05).
IL-6+ increased in colonic segments (r = 0.9062, p < 0.05). In patients without extraintestinal symptoms, these correlations were not statistically significant. Discussion A usefulness of pro-inflammatory cytokines was evaluated in several studies. Matsuda et al. [23] investigated the expressions of tumor necrosis factor TNF-␣, interleukin IL-6, IL-8 and IL-10 mRNAs in the colonic mucosa of patients with ulcerative colitis. Ljung et al. [20] reported that rectal nitric oxide levels parallel down-regulation of inducible nitric oxide synthase, tumor necrosis factor TNF-␣, interleukin-1 and interferon-␥, and may serve as a quantitative biomarker of intestinal inflammation. The central role of TNF-␣ in inflammatory response in the gut could be understood not only by its deleterious effect on tissue, but also on its up-stream function in initiating the inflammatory response. Olsen et al. [25–27] published reports of mucosal cytokine mRNA expression levels (tumor necrosis factor alpha TNF-␣, interferon gamma IFN-␥, interleukin-IL-18, IL-4, and IL-10) using real-time PCR in patients with UC. The levels of TNF-␣ mRNA were clearly increased and correlated with the grade of inflammation. Their research results underscore the importance of TNF-␣ in the pathogenesis of UC. Moreover, the authors’ experience routine measurements of mucosal TNF-␣ gave valuable clinical information regarding the treatment of UC patients. In our study, the expression of cytokines IL-1; IL-6, TNF was correlated with a progression of clinical symptoms, in particular bleeding and mucus admixture during an active stage of the disease. The inflammatory activity within colorectal neoplastic was assessed by immunohistochemistry in paired colonic adenoma and adjacent normal colonic mucosa samples, and adenomas exhibiting increasing degrees of epithelial cell dysplasia [21]. A targeted array of inflammatory cytokine and receptor genes, validated by RT-PCR, was applied to assess inflammatory gene expression. This study indicated several inflammatory cytokine genes, which are dysregulated in adenomas (CXL1, CXL2, CXL3, CCL20, Il8, CCL23, CCL19, CCL21, CCL5). Recent studies suggest that bacterial endotoxins may be associated with various chronic diseases, including colorectal adenomas and cancer. Concentrations of inflammatory cytokines: IL-4, IL-6, IL-8, IL-10, IL-12, tumor necrosis factor-alpha (TNF-␣), and interferon-␥ (IFN-␥) in plasma were quantified by ELISA and mRNA expression levels in rectal mucosal biopsies by quantitative RT-PCR [17].
These findings suggest that interactions between elevated plasma endotoxin concentrations and inflammatory cytokines may be relevant to the development of colorectal adenomas. Cases showed a trend of having higher plasma TNF-␣ levels than controls, but none of the other plasma or rectal mucosal cytokine levels differed between cases and controls. Elevated mucosal IL-12 levels were associated with having multiple adenomas. Higher concentrations of plasma endotoxin predicted increased plasma IL-12 levels and rectal mucosal IL-12 and IL-17 gene expression. These findings confirm a usefulness of studies of inflammatory cytokines (including IL-1, IL-6) and tumor necrosis factor (TNF-␣) in colorectal diseases. To assess a usefulness of biomarkers, it is important to use an objective technique of their expression [31]. Prasad and Prabhu [29] indicate that there is a great necessity for development of techniques and software for staining intensity quantification, which a medical researcher could easily use without requiring high level computer skills. A number of studies have developed their own custom written programs aiming at obtaining faster, simpler and cheaper solutions. Currently available image analysis programs, for instance Aperio, Lucia, Metaview, Metamorph, Adobe Photoshop, Image Pro Plus, ImageJ, Scion and Cell Profiler, are also used for the evaluation of expressions using immunohistochemical staining. However, high throughput imaging poses challenges of converting the raw images to interpretable information. The image analysis tools available are either general purpose tools requiring some image analysis skills and user interaction or commercial packages which are expensive. Therefore, for medical applications, in particular an evaluation of IHC images, tools for counting and quantification of stained biomarkers in the specimen should provide facility for full automation and will not require prior training. Aperio technologies have developed an open source software for IHC image analysis [28] which determines dubbed positive pixel count, calculates the area and the intensity of staining and assigns slides to four categories, negative, weak, medium, and strong. The HSI color model is used to divide the color space into two classes, positive and negative. The positive color class is divided by thresholding into three intensity ranges, which results in four total categories of staining. The total pixel count for each category is calculated. For manual thresholding, a maximum intensity and a minimum intensity are selected to limit the range of valid intensities, while for automatic thresholding, amplitude and edge statistics are used to determine the range of intensity values that belong to the stained nuclei. Our experience with applications of the spatial visualization method in the evaluation of thyroid and parathyroid lesions [15,16] encouraged us to use this method for digital image analysis of microscopic views of immunohistochemically stained specimens for pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel. In our study, objects representing specific markers were extracted by color thresholding. Their surface was computed as the total number of voxels belonging to the extracted objects. To compute the area fraction of the extracted objects, the spatial objects were then orthogonally projected onto a plane. The number of color pixels representing the specific marker was divided by the size of the view field (i.e., the size of the image) and represented in percentages. The main advantage of our method is faster image analysis, up to 50 images per hour, including visual control of each image. The analysis can be more objective by using the same filters of colors, brightness, and saturation for a specific sequence of images. Similarly, as in our study, the extent of staining is calculated as the total number of DAB-positive pixels divided by the union of the total number of H-positive pixels and the total number of DAB-positive pixels. The staining intensity is calculated from the DAB positive area, as a mean pixel value of original DAB image. The mean intensity value is scaled to range from 0 to
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
100% to compensate for the effect of different DAB thresholds in subsequent routine use [18]. The automated tumor segmentation method seems to be affected more by the increase of the image compression level than the automated IHC quantification method. The storage space needed for digital whole-slide tissue images can be reduced significantly with image compression and scaling, and studied automated image analysis algorithms perform adequately with resulting images. Matkowskyj et al. [22] presented another approach in a study focused on determining the absolute amount of chromogen by computing the cumulative signal strength of IHC images. The energy of images captured in Photoshop was calculated by using a tool originally designed in Matlab. The Authors applied this method to analyze the gastrin-releasing peptide receptor. Human colon cancers variably express this receptor as a function of tumor differentiation, so this tissue type represents a good model for demonstrating the enhanced power of digital quantitative IHC. In Matkowskyj’s study, the area of interest was selected for the control and the experimental image. In general, algorithms based on pixel counting cannot determine the absolute amount of chromogen, while algorithms based on color thresholding followed by a computation of the number of pixels present within a color/brightness range presuppose that the only information worth evaluating exists within a predefined spectral range. According to this algorithm, the cumulative energy of the control image was calculated and subtracted from that of the experimental image. Another major issue related to the automated quantification approaches generally proposed by literature is the significance of the measures of stain intensity [10]. In fact, despite all the guidelines proposing the intensity of the stain in terms of “weak”, “medium” or “strong” staining, the translation of this qualitative concept into a quantitative measure extracted from digitalized image is far from being straightforward. A variation in staining intensity is not included in the scoring system proposed by Bernstein. This stimulated our interest in introducing the Alfred IHC scoring system on the basis of objectively derived measures of immunohistochemical reaction and its intensity. The measurements of staining intensity are generally based on the strong assumption that “n-times darker means m-times more expression”. However, the darkness of the stain is determined not only by the antibody–antigen reaction per se, but also by a number of other factors that are related to the multiple steps of amplification inherent to immunostaining. Moreover, temperature, time of incubation, fixation, concentration of Antibodies, batch of Antibodies, crossreactivity and background staining are important factors in determining the relation between antigen activation and immunostaining. This seriously questions the significance of the traditional scoring procedures as proposed by the guidelines, invalidating the practice of measuring stain intensity from the digitalized images as a direct and absolute indicator of protein expression. The opinions about the feasibility of a full standardization of IHC are still controversial because of the many variables that need to be controlled. However, in the last few years, there have been extensive efforts towards the solution of the problem. Fortunately, the IHC community seems to be now fully aware of the dimension of this problem, and the required actions are being taken. In conclusion, the method of quantitative immunohistochemistry presented in this paper allows for searching associations between the expression of biomarkers and selected clinical symptoms. Evaluation of the expression of inflammatory cytokines could be recommended in active stage of the disease with present symptoms of bloody and mucus stools. A higher expression of IL-1, IL-6 and TNF in samples beyond large intestine correlates with intensified clinical course of the disease. Therefore, it can be considered as a certain decision-prognosis factor in patients with active form of
153
the disease. In specimens derived from patients without bleeding and mucus symptoms present in stools no significant correlations were found, therefore the assessment of cytokines during remission or clinically silent stage might not be useful. References [1] I. Abdallah Hajj Hussein, J.N. Freund, J.M. Reimund, A. Shams, M. Yamine, A. Leone, A.R. Jurjus, Enteropathogenic E. coli sustains iodoacetamide-induced ulcerative colitis-like colitis in rats: modulation of IL-1, IL-6, TNF-␣, COX-2, and apoptosis, J. Biol. Regul. Homeost. Agents 26 (2012) 515–526. [2] D.C. Allred, J.M. Harvey, M. Berardo, G.M. Clark, Prognostic and predictive factors in breast cancer by immunohistochemical analysis, Mod. Pathol. 11 (1998) 155–168. [3] F. Balkwill, A. Mantovani, Inflammation and cancer: back to Virchow? Lancet 357 (2001) 539–545. [4] C.N. Bernstein, M. Sargent, W.M. Gallatin, Beta2 integrin/ICAM expression in Crohn’s disease, Clin. Immunol. Immunopathol. 86 (1998) 147–160. [5] L. Bertazza, S. Mocellinl, M. DeWitte, D. Nitti, TNF and cancer: friends and/or foes? in: A. Ben-Baruch (Ed.), The Inflammatory Milieu of Tumors: Cytokines and Chemokines that Affect Tumor Growth and Metastasis, Bentham e Books, 2012, pp. 29–56, eISBN: 978-1-60805-256-1. [6] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurements Lancet 12 (1986) 307–310. [7] J.M. Bland, D. Altman, Statistics notes: Cronbach’s alpha, BMJ 314 (1997) 275. [8] M. Bosani, S. Ardizzone, G.B. Porro, Biologic targeting in the treatment of inflammatory bowel diseases, Biologics 3 (2009) 77–97. [9] L.J. Cronbach, Coefficient alpha and the internal structure of tests, Psychometrika 16 (1951) 297–334. [10] S. Di Cataldo, E. Ficarra, E. Macii, Computer-aided techniques for chromogenic immunohistochemistry: status and directions, Comput. Biol. Med. 42 (2012) 29–56, http://www.benthamscience.com/ebooks/9781608052561/ [11] S.E. Erdman, V.P. Rao, T. Poutahidis, A.B. Rogers, C.L. Taylor, E.A. Jackson, Z. Ge, C.W. Lee, D.B. Schauer, G.N. Wogan, S.R. Tannenbaum, J.G. Fox, Nitric oxide and TNF-alpha trigger colonic inflammation and carcinogenesis in Helicobacter hepaticus-infected, Rag2-deficient mice, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 1027–1032. [12] E.O. Farombi, I.A. Adedara, B.O. Ajayi, O.R. Ayepola, E.E. Egbeme, Kolaviron, a natural antioxidant and anti-inflammatory phytochemical prevents dextran sulphate sodium-induced colitis in rats, Basic Clin. Pharmacol. Toxicol. 113 (2013) 49–55. [13] W.W. Häuser, C. Schmidt, A. Stallmach, Depression and mucosal proinflammatory cytokines are associated in patients with ulcerative colitis and pouchitis – a pilot study, J. Crohns Colitis 5 (2011) 350–353. [14] M.M. Hefti, D.B. Chessin, N. Harpaz, R.M. Steinhagen, T.A. Ullman, Severity of inflammation as a predictor of colectomy in patients with chronic ulcerative colitis, Dis. Colon Rectum 52 (2009) 193–197. [15] E. Kaczmarek, K. Lacka, P. Majewski, P. Trzeciak, A. Gorna, D. JarmolowskaJurczyszyn, A. Kluk, Selected markers of proliferation and apoptosis in the parathyroid lesions: a spatial visualization and qualification, J. Mol. Histol. 39 (2008) 509–517. ˛ [16] E. Kaczmarek, K. Lacka, D. Jarmolowska-Jurczyszyn, P. ASidor, Majewski, Changes of B and T lymphocytes and selected apopotosis markers in Hashimoto’s thyroiditis, J. Clin. Pathol. 64 (2011) 626–630. [17] M. Kang, P. Edmundson, F. Araujo-Perez, A.N. McCoy, J. Galanko, T.O. Keku, Association of plasma endotoxin, inflammatory cytokines and risk of colorectal adenomas, Cancer 13 (2013) 91–98. [18] J. Konsti, M. Lundin, N. Linder, C. Haglund, C. Blomqvist, H. Nevanlinna, K. Aaltonen, S. Nordling, J. Lundin, Effect of image compression and scaling on automated scoring of immunohistochemical stainings and segmentation of tumor epithelium, Diagn. Pathol. 7 (2012) 7–29. [19] J. Kurtovic, I. Segal, I. Recent advances in biological therapy for inflammatory bowel disease, Trop. Gastroenterol. 25 (2004) 9–14. [20] T. Ljung, L.G. Axellson, M. Herulf, J.O. Lundberg, P.M. Hellstrom, Early changes in rectal nitric oxide and mucosal inflammatory mediators in Crohn’s colitis in response to infliximab treatment, Aliment. Pharmacol. Ther. 25 (2007) 925–932. [21] M.H. McLean, G.I. Murray, K.N. Stewart, G. Norrie, C. Mayer, G.L. Hold, J. Thompson, N. Fyfe, M. Hope, N.A. Mowat, J.E. Drew, E.M. El-Omar, The inflammatory microenvironment in colorectal neoplasia, PLoS ONE 7 (2011) e15366. [22] K.A. Matkowskyj, R. Cox, R.T. Jensen, R.V. Benya, Quantitative immunohistochemistry by measuring cumulative signal strength accurately measures receptor number, J. Histochem. Cytochem. 51 (2003) 205–214. [23] R. Matsuda, T. Koide, C. Tokoro, T. Yamamoto, T. Godai, T. Morohashi, Y. Fujita, D. Takahashi, I. Kawana, S. Suzuki, S. Umemura, Quantitative cytokine mRNA expression profiles in the colonic mucosa of patients with steroid naive ulcerative colitis during active and quiescent disease, Inflamm. Bowel Dis. 15 (2009) 328–334. [24] D.C. Montrose, N.A. Horelik, J.P. Madigan, G.D. Stoner, L.S. Wang, R.S. Bruno, H.J. Park, C. Giardina, D.W. Rosenberg, Anti-inflammatory effects of freezedried black raspberry powder in ulcerative colitis, Carcinogenesis 32 (2011) 343–350. [25] T. Olsen, R. Goll, G. Cui, A. Husebekk, B. Vonen, G.S. Birketvedt, J. Florholmen, Tissue levels of tumor necrosis factor-alpha correlates with grade of
154
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
inflammation in untreated ulcerative colitis, Scand. J. Gastroenterol. 42 (2007) 1312–1320. [26] T. Olsen, R. Goll, G. Cui, A. Husebekk, J. Florholmen, Infliximab therapy decreases the levels of TNF-alpha and IFN-gamma mRNA in colonic mucosa of ulcerative colitis, Scand. J. Gastroenterol. 44 (2009) 727–735. [27] T. Olsen, R. Goll, G. Cui, I. Christiansen, J. Florholmen, TNF-alpha gene expression in colorectal mucosa as a predictor of remission after induction therapy with infliximab in ulcerative colitis, Cytokine 46 (2009) 222–227. [28] A.H. Olson, Image Analysis Using the Aperio ScanScope. Technical Manual, Aperio Technologies Inc., 2006, http://www.quorumtechnologies.com/ pdf files/AnalysisWhitePaper.pdf
[29] K. Prasad, G.K. Prabhu, Image analysis tools for evaluation of microscopic views of immunohistochemically stained specimen in medical research – a review, J. Med. Syst. 36 (2012) 2621–2631. [30] D. Raman, P.J. Baugher, Y.M. Thu, A. Richmond, Role of chemokines in tumor growth, Cancer Lett. 256 (2007) 137–165. [31] Z. Theodosiou, I.N. Kasampalidis, G. Livanos, M. Zervakis, I. Pitas, K. Lyroudia, Automated analysis of FISH and immunohistochemistry images: a review, Cytometry 71A (2007) 439–450. [32] N.L. Vujanovic, Role of TNF family ligands in antitumor activity of natural killer cells, Int. Rev. Immunol. 20 (2001) 415–437.
Contents lists available at ScienceDirect
Pathology – Research and Practice journal homepage: www.elsevier.com/locate/prp
Original article
Digital image analysis of inflammation markers in colorectal mucosa by using a spatial visualization method Elzbieta Kaczmarek a,∗ , Tomasz Banasiewicz b , Agnieszka Seraszek-Jaros a , Piotr Krokowicz c , Marcin Grochowalski c , Przemysław Majewski d , Jakub Zurawski e , Jacek Paszkowski b , Michał Drews b a
Department of Bioinformatics and Computational Biology, Pozna´ n University of Medical Sciences, Poland Department of General Surgery, Oncologic Gastroenterologic Surgery and Plastic Surgery, Pozna´ n University of Medical Sciences, Poland c Department of General and Colorectal Surgery, Pozna´ n University of Medical Sciences, Poland d Department of Clinical Pathomorphology, Pozna´ n University of Medical Sciences, Poland e Department of Biology and Environmental Protection, Pozna´ n University of Medical Sciences, Poland b
a r t i c l e
i n f o
Article history: Received 6 July 2013 Received in revised form 29 October 2013 Accepted 14 November 2013 Keywords: Ulcerative colitis IL-1 IL-6 TNF-␣ Digital image analysis
a b s t r a c t The aim of this study was to apply the spatial visualization method of digital images to quantitative analysis of pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel excised because of colitis ulcerosa in relation with selected clinical symptoms. Our preliminary study included 17 patients having undergone restorative proctocolectomy. Immunohistochemistry was performed for IL-1, IL-6 and TNF-␣. The area fraction and intensity fraction of the cytokines studied were determined by digital image analysis. The results were then categorized using Alfred Immunohistochemistry Score. The expression of IL-1, IL-6 and TNF-␣ was significantly higher in the rectum than in colonic segments (p < 0.01), and was associated with the patients’ clinical condition. The method of quantitative immunohistochemistry presented here allows for searching associations between the expression of biomarkers and clinical symptoms. Evaluation of inflammatory cytokines could be recommended in the active stage of the disease with present symptoms of bloody and mucus stools. A higher expression of IL-1, IL-6 and TNF in samples beyond large intestine correlates with an intensified clinical course of the disease. In patients without bleeding and mucus symptoms present in stools, no significant correlations were found. Therefore, the assessment of cytokines during remission or clinically silent stage might not be useful. © 2013 Published by Elsevier GmbH.
Introduction Research interest in a possible connection between inflammation and cancer has been growing since 1863, when Rudolf Virchow noted the leucocytes in neoplastic tissues and indicated that chronic inflammation supports cancerogenesis [3]. Mediators of the inflammatory response, e.g., cytokines and chemokines, induce the accumulation of immune cells and their activation within the inflamed tissue. This process may lead to the release of reactive oxygen and nitrogen species that can induce genetic and epigenetic changes including point mutations in tumor
∗ Corresponding author at: Department of Bioinformatics and Computational Biology, Poznan´ University of Medical Sciences, Dabrowski Street, 79, 60-529 Poznan, Poland. Tel.: +48 61 854 6909. E-mail address: [email protected] (E. Kaczmarek). 0344-0338/$ – see front matter © 2013 Published by Elsevier GmbH. http://dx.doi.org/10.1016/j.prp.2013.11.007
suppressor genes, DNA methylation and post-translational modifications. This sequence of changes in critical pathways responsible for maintaining normal cellular homeostasis can lead to the development and progression of cancer. Several chronic inflammatory diseases contribute to an increased risk of cancer. Observations that many malignancies are associated with chronic infection and inflammation support this hypothesis, for instance, inflammatory bowel disease is associated with colon cancer. Inflammatory bowel diseases, i.e. Crohn’s disease and colitis ulcerosa, are associated with increased rates of colon adenocarcinoma [30]. Tumor necrosis factor can initiate the inflammatory reactions of the immune system, and induces the production of other cytokines (e.g. IL-1, IL-6 and IL-8) and cytotoxic factors (e.g. nitric oxide, reactive oxygen species) by macrophages, which can mediate tumor suppression. Members of the IL-1 family played significant roles in many aspects of the cancerous process as key regulators of the balance between inflammation and immunity in the tumor microenvironment [11,32].
148
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
The importance of IL-1, IL-6 and TNF in the severity of inflammation in ulcerative colitis has been confirmed in clinical and experimental studies [1,12,13]. Increased levels of those proinflammatory cytokines are detected in active inflammatory bowel disease and correlate with the severity of inflammation, indicating that these cytokines may play a key role in the development of inflammatory bowel disease [24]. Analysis of the inflammed mucosa from patients with Crohn’s disease and ulcerative colitis have shown increased expression of certain proinflammatory cytokines such as IL-1, interleukin IL6 and TNF-␣ [19]. Early production of tumor necrosis factor is prominent in the initiation of a biologically highly complex system involving chemokines, cytokines and endothelial adhesions that recruits and activates neutrophils, macrophages and lymphocytes at the sites of infections. IL-1 and TNF-␣ increase the expression of high-affinity adhesion molecules on endothelial cells, stromal cells and leukocytes and, by this, promote infiltration of inflammatory cells from the blood into tissues. Dysregulation of tumor necrosis factor has also been implicated in a wide variety of autoimmune diseases, including Crohn’s disease; however, how tumor necrosis factor mediates disease-causing effects is not completely explained. The induction of proinflammatory genes by tumor necrosis factor has been linked to most diseases [5]. TNF␣, together with IL-1 and IL-6, is responsible for the development of clinical symptoms and stimulation of acute-phase proteins in ulcerative colitis [8]. Evaluation of severity of inflammatory lesions is an essential component in clinical staging of colitis ulcerosa. Immunohistochemistry (IHC) is an important technique for biomarker validation. It allows direct visualization of biomarker expression in histologically relevant regions of the examined tissue. Our search continues for objective markers enabling unambiguous possible assessment of severity of inflammatory lesions in the course of ulcerative colitis, potential risk of dysplasia and in consequence – of malignant transformation. Traditionally, pathologists examine visually and assign scores for IHC data in a semi-quantitative fashion incorporating the intensity and the distribution of specific staining. Due to staining heterogeneity of the tumor cells, the pathologist’s visual scoring is fraught with problems due to subjectivity in interpretation. Automated IHC measurements promise to overcome these limitations. Therefore, we decided to apply a spatial visualization method for quantitative evaluation of IHC data. Thus, the aim of this study is to present an application of the spatial visualization method of digital images for quantitative analysis of pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel excised because of colitis ulcerosa in relation with selected clinical symptoms.
Materials and methods Our preliminary study included 17 patients having undergone restorative proctocolectomy with J-pouch and temporary loop ileostomy or Hartmann’s colectomy with end ileostomy and preservation of a short rectal stump. The patients were operated on at the Department of General, Gastroenterological and Endocrinological Surgery and at the Department of General and Colorectal Surgery of the University of Medical Sciences in Poznan. Indications for surgical treatment included: exacerbating clinical symptoms (diarrhea, lower digestive tract bleeding, progressive loss of body weight, malnutrition, megacolon, fulminating colitis). Patients’ age in the study was on average 43 ± 14 years (range: 21–69 years), and the males to females ratio was 9:8. The mean duration of symptoms was 58 ± 31 months (range: 12–120
months). All patients were treated without the use of steroids agents and biological therapy, as well as immunosuppression in the last 2 months before surgery. The study excluded patients with other systemic diseases such as diabetes or rheumatoidal diseases. Inclusion criteria reduced the number of patients, but resulted in greater homogeneity of the group and reduced the potential impact of other factors on the level of investigated cytokines. Rectal mucosa was much more altered on macroscopic inspection, showing visible thinning of bowel wall, numerous ulcers and indicators of active bleeding. Mucosal samples of rectum and colon were collected from surgical specimens directly after resection, up to a maximum of 20 min after artery ligations. Bowel was cut in longitudinal line. The size of specimens was standardized (5 mm × 5 mm), specimens were taken from the transversal colon and from the middle part of the rectum, 5–6 cm below the dental line. Then, specimens were fixed in buffered 10% formalin (pH 7–7.8) for 48 h at stable temperature (22 ◦ C) (air-conditioned room) and processed in a standard fashion. Paraffin sections were stained with hematoxylin and eosin (H + E), and only blocks with the most pronounced lesions were selected for further immunohistochemical studies. The specimens were then histologically evaluated for severity of inflammation and dysplasia. Inflammation was scored on a scale of 0–2 [14], whereby: • 0 – lack of active inflammation (no indices of cryptitis); • 1 – mild signs of acute inflammation (cryptitis present in less than 50% of crypts); • 2 – moderate signs of acute inflammation (cryptitis present in more than 50% of crypts); • 3 – severe inflammation (ulcers and defects of mucosal epithelium). Microscopic study of tissue samples revealed moderate severity of inflammatory process: mean score 2.1 (range: 0–3).
Immunohistochemical studies Tissue samples were processed using the streptavidin-biotinperoxidase technique (LSAB kit, DAKO, K0675). Antigens were exposed in a water bath at a temperature of 96 ◦ C in a citrate buffer (pH 6.0) during 90 min. Activity of endogenous peroxidase was inhibited using 3% H2 O2 . Subsequently, preparations were incubated overnight at room temperature with antibodies anti-IL-1 (MAB263), anti-IL-6 (MAB261) and anti-TNF-␣ (MAB226). During subsequent 30 min incubation, biotinylated antibody was used, which was then incubated with peroxidase–streptavidin complex. Between consecutive incubations, preparations were flushed with TBS buffer (pH 7.6) using DAB-3.3 chromogen (SIGMA-ALDRICH, D5637), and the antigen was located. Preparations were stained with hematoxylin and after dehydration cover glass was applied. Reaction devoid of primary antibody was used as negative control for immunocytochemical studies. Histological slides were examined with Olympus DP-12 microscope coupled to a digital camera. Color microscopic images of size 2048 × 1536 pixels were acquired and archived using 40× objective (at least 10 images in every slide with an immunopositive reaction). Using the software DP12-BSW, images were downloaded directly from the camera control unit to the PC. To assess the expression of markers studied within inflammatory infiltrates, a quantitative analysis was used. The area fraction and markers’ intensity fraction were determined by using a spatial visualization technique [15]. The received results were then categorized by using Alfred Immunohistochemistry Score [2] in the following way:
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
149
• the proportion score (PS) was assigned on the basis of the approximate percentage of positively stained cells into four categories (0: no positive cells, 1: >0–1%, 2: >1–10%, 3: >10–33%, 4: >33–66% and 5: >66% positive cells), • the intensity score (IS) of the staining product was graded into four categories: no staining (0), weakly positive staining (1), i.e. the computed intensity was less than 33%, intermediate staining intensity (2), i.e. the computed intensity >33–66%, strongly positive staining (3), i.e. the intensity was greater than 66%.
Table 1 Clinical condition of the study group.
The total score (TS) is TS = PS + IS, TS range 0, 2–8. In addition, the expression of markers within inflammatory infiltrates was evaluated by pathologists with the use of the points scoring system proposed by Bernstein et al. [4], whereby:
WBC – leukocytes, Hgb – hemoglobin, RBC – erythrocytes, ESR – erythrocyte sedimentation rate.
• 0 – denotes lack of expression in infiltrating cells; • 1 – less than half of 10 field examined presents infiltrates with reaction products, weak expression; • 2 – the expression is present in 1–10 cells in each visual field (moderate expression); • 3 – the expression is present in 11–30 cells in each visual field; • 4 – the expression is visible in over 30 cells in each visual field. Microscopic assessment was performed using a 40-fold magnification. Quantitative analysis of micrographs The IHC were quantitatively evaluated using image-processing methods, including a spatial visualization of the markers’ expression [15,16]. First, raw color images were converted to a HSI color space (hue, saturation, intensity). Then, intervals for the colors’ hue, saturation and intensity representing the specific IHC reactions were derived from the histogram of color distribution. Pixels representing the specific markers were converted into voxels by introducing brightness as their third dimension. The expression of the specific marker was revealed on a spatial view by reducing the scenery behind to a background using brightness thresholding. The background pixels were filled with white color, and this resulted in a flat, horizontal surface, while colors representing the specific marker remained unchanged (Figs. 1–3). Objects representing specific markers were then extracted by thresholding in HIS color space. Their surface was computed as the total number of voxels, which belonged to the extracted objects. To compute the area fraction of the extracted objects, the spatial objects were then orthogonally projected onto a plane. The number of color pixels representing the specific marker was divided by the size of the view field (i.e. the size of the image) and represented in percent. The intensity of the positive reaction was computed as the average of the brightest and darkest color values in the selected object, i.e. (max(H,S,I) + min(H,S,I))/2 and also expressed as the percent of the highest possible intensity value (i.e. the darkest color of the reaction). Then, the area fraction and intensity results were categorized using the Alfred Immunohistochemistry Score [2] in the following way: • the proportion score (PS) was assigned on the basis of the approximate percentage of positively stained cells into four categories (0: no positive cells, 1: >0–1%, 2: >1–10%, 3: >10–33%, 4: >33–66% and 5: >66% positive cells), • the intensity score (IS) of the staining product was graded into four categories (0: no staining, 1: weakly positive, i.e. the computed intensity was less than 33%, 2: intermediate, i.e. the computed intensity >33–66%, 3: strongly positive, i.e. the intensity was greater than 66%).
Clinical condition of patients
Mean ± SD
Duration of symptoms (months) Number of bowel movements WBC (103 /l) Hgb (g/100 ml) RBC (106 /ml) ESR Total protein (g/dl) Platelets (×109 /l)
58 12 8.1 12.8 4.3 24 6.7 353
± ± ± ± ± ± ± ±
31 4 3.9 1.9 0.6 14 0.9 149
Minimum
Maximum
12 5 2.2 8.2 3.1 3 4.4 108
120 20 17 15.9 5.8 48 8.5 600
The total score (TS) is TS = PS + IS, TS range 0, 2–8. Statistical analysis Two observers first analyzed the expression of studied cytokines in a pilot sample of 50 images. To evaluate inter- and intra-observer variability of the measurements, the mean difference in the measurements of the two observers was first calculated, as well as their respective average, and 95% confidence intervals. The 95% agreement limits were evaluated according to the method proposed by Bland and Altman [6]. The internal consistency of the results was evaluated on the basis of Cronbach’s alpha coefficient, with values >0.8 being considered indicative of good reliability [7,9]. Descriptive statistics of all results included mean values and standard deviations. The Mann–Whitney test was used to compare the results between independent groups. A correlation between expression of the studied cytokines and progression of symptoms was assessed by Spearman rank correlation coefficient. A p value of less than 0.05 was considered to be statistically significant. The power of the tests used to detect hypothetical differences between colon and rectum was also estimated. The data were analyzed using Statistica v.10.1 (Statsoft, Inc., http://www.statsoft.com). Intra-observer variability was assessed by using Prism v.5 (GraphPad Software Inc., http://www.graphpad.com). Results Comparisons between the measurements of area fractions of studied cytokines obtained by the two evaluators (inter-observer variability by Bland and Altman method) were found to be dispersed around the mean in 95% agreement limits (from −1.173 to 1.232) with bias equal 0.029. The internal consistency (Cronbach’s alpha) was significantly high: 0.947 with 95% lower confidence limit 0.746. The clinical condition of the patients is presented in Table 1. Microscopic evaluation of tissue samples revealed a moderate severity of the inflammatory process: mean score 2.1 (range: 0–3). The mean score reflecting severity of proctitis was 2.7 (range: 1.53), while that of colitis – 1.7 (range: 0–2.7). A quantitative evaluation of the area fraction and the intensity fraction by using the spatial visualization technique is presented in Table 2. Alfred IHC proportion scores, intensity scores and total scores derived from the computed area and intensity fractions are also presented in Table 2. The intensity of the IHC reactions was comparable for all studied cytokines. The expression of IL-1, IL-6 and TNF-␣, measured on the basis of area fraction of positive reaction, was significantly lower in the rectum than in colonic segments (Table 3). However, Alfred IHC total scores assigned to raw quantitative results of the cytokine expression were not significantly different for IL-6 (Table 3). The expression of the studied cytokines assessed by the spatial visualization technique
150
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
Fig. 1. Expression of IL-1 in colon (upper row) and rectum (lower row): raw image (left), 3D view of Il-1 (middle) and extracted reaction (right).
Fig. 2. Expression of IL-6 in colon (upper row) and rectum (lower row): raw image (left), 3D view of Il-6 (middle) and extracted reaction (right).
Fig. 3. Expression of TNF-␣ in colon (upper row) and rectum (lower row): raw image (left), 3D view of TNF-␣ (middle) and extracted reaction (right).
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
151
Table 2 Summary statistics of studied cytokines in rectum and colon. Area fraction (Mean ± SD) IL-1 rectum IL-1 colon IL-6 rectum IL-6 colon TNF-␣ rectum TNF-␣ colon
0.8 1.8 1.0 1.5 0.9 2.6
± ± ± ± ± ±
0.7 1.1 0.8 0.9 0.5 1.5
Intensity fraction (Mean ± SD) 53 53 53 53 52 54
± ± ± ± ± ±
Alfred IHC proportion score (PS) (Mean ± SD)
7 6 7 6 8 5
1.2 1.7 1.4 1.7 1.3 1.9
± ± ± ± ± ±
0.2 0.5 0.5 0.5 0.5 0.2
Alfred IHC intensity score (IS) (Mean ± SD) 2 2 2 2 2 2
± ± ± ± ± ±
0 0 0 0 0 0
Alfred IHC total score (PS) (Mean ± SD) 3.2 3.7 3.4 3.5 3.3 3.9
± ± ± ± ± ±
0.4 0.5 0.5 0.6 0.5 0.2
Table 3 Differences between the area fraction and Alfred IHC total score of studied cytokines in rectum and colon. Cytokines
Area fraction (Mean ± SD)
IL-1 rectum IL-1 colon IL-6 rectum IL-6 colon TNF-␣ rectum TNF-␣ colon
0.8 1.8 1.0 1.5 0.9 2.6
± ± ± ± ± ±
0.7 1.1 0.8 0.9 0.5 1.5
Test M–W (p-value)
Test power (%)
0.003
76
0.047
47
0.001
91
Alfred IHC total score (TS) (Mean ± SD) 3.2 3.7 3.4 3.5 3.3 3.9
± ± ± ± ± ±
0.4 0.5 0.5 0.6 0.5 0.2
Test M–W (p-value)
Test power (%)
0.003
83
0.601
50
0.001
94
Fig. 4. A correlation between the expression of IL-1 in rectum of male patients vs. content of hemoglobin (r = −0.6833, p < 0.05), RBC (r = −0.6666, p < 0.05) and total protein (r = −0.7615, p < 0.05).
Fig. 5. A correlation between the expression of IL-1 in rectum of patients with bleeding symptoms vs. content of hemoglobin (r = −0.4519, p < 0.05), and total protein (r = −0.7586, p < 0.05)
was associated with the patients’ clinical condition and clinical symptoms. In particular, an increasing expression of IL-1 in rectum of male patients was correlated with decreasing number of red blood cells (r = −0.6666, p < 0.05) and decreasing content of hemoglobin (r = −0.6833, p < 0.05) and total protein (r = −0.7615, p < 0.05) (Fig. 4). In females, these correlations were not statistically significant. In patients with bleeding symptoms (independently on gender), a higher expression of IL-1 in rectum was correlated with decreasing content of hemoglobin (r = −0.6648, p < 0.05) and total protein IL-1 (r = −0.7586, p < 0.05) (Fig. 5). In patients without bleeding symptoms, these correlations were not statistically significant. In patients with presence of mucus, an increasing expression of TNF-␣ in colonic segments was significantly correlated with decreasing number of RBC (r = −0.4519, p < 0.05) and total protein content (r = −0.6336, p < 0.05) (Fig. 6). In patients without mucus, these correlations were not significant. In patients with extraintestinal symptoms, an increasing expression of IL-6 in colonic segments was significantly correlated with lower number of white blood cells (r = −0.7352, p < 0.05) and content hemoglobin (r = −0.9062, p < 0.05) (Fig. 7). Moreover, in cases with extraintestinal symptoms, the RBC count significantly decreased when
Fig. 6. A correlation between the expression of TNF-␣ in colon of patients with mucus vs. RBC (r = −0.6648, p < 0.05), and total protein (r = −0.6336, p < 0.05).
152
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
Fig. 7. A correlation between the expression of IL-6 in colon of patients with extraintestinal symptoms vs. WBC (r = −0.7352, p < 0.05), and content of hemoglobin (r = −0.9062, p < 0.05).
IL-6+ increased in colonic segments (r = 0.9062, p < 0.05). In patients without extraintestinal symptoms, these correlations were not statistically significant. Discussion A usefulness of pro-inflammatory cytokines was evaluated in several studies. Matsuda et al. [23] investigated the expressions of tumor necrosis factor TNF-␣, interleukin IL-6, IL-8 and IL-10 mRNAs in the colonic mucosa of patients with ulcerative colitis. Ljung et al. [20] reported that rectal nitric oxide levels parallel down-regulation of inducible nitric oxide synthase, tumor necrosis factor TNF-␣, interleukin-1 and interferon-␥, and may serve as a quantitative biomarker of intestinal inflammation. The central role of TNF-␣ in inflammatory response in the gut could be understood not only by its deleterious effect on tissue, but also on its up-stream function in initiating the inflammatory response. Olsen et al. [25–27] published reports of mucosal cytokine mRNA expression levels (tumor necrosis factor alpha TNF-␣, interferon gamma IFN-␥, interleukin-IL-18, IL-4, and IL-10) using real-time PCR in patients with UC. The levels of TNF-␣ mRNA were clearly increased and correlated with the grade of inflammation. Their research results underscore the importance of TNF-␣ in the pathogenesis of UC. Moreover, the authors’ experience routine measurements of mucosal TNF-␣ gave valuable clinical information regarding the treatment of UC patients. In our study, the expression of cytokines IL-1; IL-6, TNF was correlated with a progression of clinical symptoms, in particular bleeding and mucus admixture during an active stage of the disease. The inflammatory activity within colorectal neoplastic was assessed by immunohistochemistry in paired colonic adenoma and adjacent normal colonic mucosa samples, and adenomas exhibiting increasing degrees of epithelial cell dysplasia [21]. A targeted array of inflammatory cytokine and receptor genes, validated by RT-PCR, was applied to assess inflammatory gene expression. This study indicated several inflammatory cytokine genes, which are dysregulated in adenomas (CXL1, CXL2, CXL3, CCL20, Il8, CCL23, CCL19, CCL21, CCL5). Recent studies suggest that bacterial endotoxins may be associated with various chronic diseases, including colorectal adenomas and cancer. Concentrations of inflammatory cytokines: IL-4, IL-6, IL-8, IL-10, IL-12, tumor necrosis factor-alpha (TNF-␣), and interferon-␥ (IFN-␥) in plasma were quantified by ELISA and mRNA expression levels in rectal mucosal biopsies by quantitative RT-PCR [17].
These findings suggest that interactions between elevated plasma endotoxin concentrations and inflammatory cytokines may be relevant to the development of colorectal adenomas. Cases showed a trend of having higher plasma TNF-␣ levels than controls, but none of the other plasma or rectal mucosal cytokine levels differed between cases and controls. Elevated mucosal IL-12 levels were associated with having multiple adenomas. Higher concentrations of plasma endotoxin predicted increased plasma IL-12 levels and rectal mucosal IL-12 and IL-17 gene expression. These findings confirm a usefulness of studies of inflammatory cytokines (including IL-1, IL-6) and tumor necrosis factor (TNF-␣) in colorectal diseases. To assess a usefulness of biomarkers, it is important to use an objective technique of their expression [31]. Prasad and Prabhu [29] indicate that there is a great necessity for development of techniques and software for staining intensity quantification, which a medical researcher could easily use without requiring high level computer skills. A number of studies have developed their own custom written programs aiming at obtaining faster, simpler and cheaper solutions. Currently available image analysis programs, for instance Aperio, Lucia, Metaview, Metamorph, Adobe Photoshop, Image Pro Plus, ImageJ, Scion and Cell Profiler, are also used for the evaluation of expressions using immunohistochemical staining. However, high throughput imaging poses challenges of converting the raw images to interpretable information. The image analysis tools available are either general purpose tools requiring some image analysis skills and user interaction or commercial packages which are expensive. Therefore, for medical applications, in particular an evaluation of IHC images, tools for counting and quantification of stained biomarkers in the specimen should provide facility for full automation and will not require prior training. Aperio technologies have developed an open source software for IHC image analysis [28] which determines dubbed positive pixel count, calculates the area and the intensity of staining and assigns slides to four categories, negative, weak, medium, and strong. The HSI color model is used to divide the color space into two classes, positive and negative. The positive color class is divided by thresholding into three intensity ranges, which results in four total categories of staining. The total pixel count for each category is calculated. For manual thresholding, a maximum intensity and a minimum intensity are selected to limit the range of valid intensities, while for automatic thresholding, amplitude and edge statistics are used to determine the range of intensity values that belong to the stained nuclei. Our experience with applications of the spatial visualization method in the evaluation of thyroid and parathyroid lesions [15,16] encouraged us to use this method for digital image analysis of microscopic views of immunohistochemically stained specimens for pro-inflammatory cytokines IL-1, IL-6 and TNF-␣ in various segments of large bowel. In our study, objects representing specific markers were extracted by color thresholding. Their surface was computed as the total number of voxels belonging to the extracted objects. To compute the area fraction of the extracted objects, the spatial objects were then orthogonally projected onto a plane. The number of color pixels representing the specific marker was divided by the size of the view field (i.e., the size of the image) and represented in percentages. The main advantage of our method is faster image analysis, up to 50 images per hour, including visual control of each image. The analysis can be more objective by using the same filters of colors, brightness, and saturation for a specific sequence of images. Similarly, as in our study, the extent of staining is calculated as the total number of DAB-positive pixels divided by the union of the total number of H-positive pixels and the total number of DAB-positive pixels. The staining intensity is calculated from the DAB positive area, as a mean pixel value of original DAB image. The mean intensity value is scaled to range from 0 to
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
100% to compensate for the effect of different DAB thresholds in subsequent routine use [18]. The automated tumor segmentation method seems to be affected more by the increase of the image compression level than the automated IHC quantification method. The storage space needed for digital whole-slide tissue images can be reduced significantly with image compression and scaling, and studied automated image analysis algorithms perform adequately with resulting images. Matkowskyj et al. [22] presented another approach in a study focused on determining the absolute amount of chromogen by computing the cumulative signal strength of IHC images. The energy of images captured in Photoshop was calculated by using a tool originally designed in Matlab. The Authors applied this method to analyze the gastrin-releasing peptide receptor. Human colon cancers variably express this receptor as a function of tumor differentiation, so this tissue type represents a good model for demonstrating the enhanced power of digital quantitative IHC. In Matkowskyj’s study, the area of interest was selected for the control and the experimental image. In general, algorithms based on pixel counting cannot determine the absolute amount of chromogen, while algorithms based on color thresholding followed by a computation of the number of pixels present within a color/brightness range presuppose that the only information worth evaluating exists within a predefined spectral range. According to this algorithm, the cumulative energy of the control image was calculated and subtracted from that of the experimental image. Another major issue related to the automated quantification approaches generally proposed by literature is the significance of the measures of stain intensity [10]. In fact, despite all the guidelines proposing the intensity of the stain in terms of “weak”, “medium” or “strong” staining, the translation of this qualitative concept into a quantitative measure extracted from digitalized image is far from being straightforward. A variation in staining intensity is not included in the scoring system proposed by Bernstein. This stimulated our interest in introducing the Alfred IHC scoring system on the basis of objectively derived measures of immunohistochemical reaction and its intensity. The measurements of staining intensity are generally based on the strong assumption that “n-times darker means m-times more expression”. However, the darkness of the stain is determined not only by the antibody–antigen reaction per se, but also by a number of other factors that are related to the multiple steps of amplification inherent to immunostaining. Moreover, temperature, time of incubation, fixation, concentration of Antibodies, batch of Antibodies, crossreactivity and background staining are important factors in determining the relation between antigen activation and immunostaining. This seriously questions the significance of the traditional scoring procedures as proposed by the guidelines, invalidating the practice of measuring stain intensity from the digitalized images as a direct and absolute indicator of protein expression. The opinions about the feasibility of a full standardization of IHC are still controversial because of the many variables that need to be controlled. However, in the last few years, there have been extensive efforts towards the solution of the problem. Fortunately, the IHC community seems to be now fully aware of the dimension of this problem, and the required actions are being taken. In conclusion, the method of quantitative immunohistochemistry presented in this paper allows for searching associations between the expression of biomarkers and selected clinical symptoms. Evaluation of the expression of inflammatory cytokines could be recommended in active stage of the disease with present symptoms of bloody and mucus stools. A higher expression of IL-1, IL-6 and TNF in samples beyond large intestine correlates with intensified clinical course of the disease. Therefore, it can be considered as a certain decision-prognosis factor in patients with active form of
153
the disease. In specimens derived from patients without bleeding and mucus symptoms present in stools no significant correlations were found, therefore the assessment of cytokines during remission or clinically silent stage might not be useful. References [1] I. Abdallah Hajj Hussein, J.N. Freund, J.M. Reimund, A. Shams, M. Yamine, A. Leone, A.R. Jurjus, Enteropathogenic E. coli sustains iodoacetamide-induced ulcerative colitis-like colitis in rats: modulation of IL-1, IL-6, TNF-␣, COX-2, and apoptosis, J. Biol. Regul. Homeost. Agents 26 (2012) 515–526. [2] D.C. Allred, J.M. Harvey, M. Berardo, G.M. Clark, Prognostic and predictive factors in breast cancer by immunohistochemical analysis, Mod. Pathol. 11 (1998) 155–168. [3] F. Balkwill, A. Mantovani, Inflammation and cancer: back to Virchow? Lancet 357 (2001) 539–545. [4] C.N. Bernstein, M. Sargent, W.M. Gallatin, Beta2 integrin/ICAM expression in Crohn’s disease, Clin. Immunol. Immunopathol. 86 (1998) 147–160. [5] L. Bertazza, S. Mocellinl, M. DeWitte, D. Nitti, TNF and cancer: friends and/or foes? in: A. Ben-Baruch (Ed.), The Inflammatory Milieu of Tumors: Cytokines and Chemokines that Affect Tumor Growth and Metastasis, Bentham e Books, 2012, pp. 29–56, eISBN: 978-1-60805-256-1. [6] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurements Lancet 12 (1986) 307–310. [7] J.M. Bland, D. Altman, Statistics notes: Cronbach’s alpha, BMJ 314 (1997) 275. [8] M. Bosani, S. Ardizzone, G.B. Porro, Biologic targeting in the treatment of inflammatory bowel diseases, Biologics 3 (2009) 77–97. [9] L.J. Cronbach, Coefficient alpha and the internal structure of tests, Psychometrika 16 (1951) 297–334. [10] S. Di Cataldo, E. Ficarra, E. Macii, Computer-aided techniques for chromogenic immunohistochemistry: status and directions, Comput. Biol. Med. 42 (2012) 29–56, http://www.benthamscience.com/ebooks/9781608052561/ [11] S.E. Erdman, V.P. Rao, T. Poutahidis, A.B. Rogers, C.L. Taylor, E.A. Jackson, Z. Ge, C.W. Lee, D.B. Schauer, G.N. Wogan, S.R. Tannenbaum, J.G. Fox, Nitric oxide and TNF-alpha trigger colonic inflammation and carcinogenesis in Helicobacter hepaticus-infected, Rag2-deficient mice, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 1027–1032. [12] E.O. Farombi, I.A. Adedara, B.O. Ajayi, O.R. Ayepola, E.E. Egbeme, Kolaviron, a natural antioxidant and anti-inflammatory phytochemical prevents dextran sulphate sodium-induced colitis in rats, Basic Clin. Pharmacol. Toxicol. 113 (2013) 49–55. [13] W.W. Häuser, C. Schmidt, A. Stallmach, Depression and mucosal proinflammatory cytokines are associated in patients with ulcerative colitis and pouchitis – a pilot study, J. Crohns Colitis 5 (2011) 350–353. [14] M.M. Hefti, D.B. Chessin, N. Harpaz, R.M. Steinhagen, T.A. Ullman, Severity of inflammation as a predictor of colectomy in patients with chronic ulcerative colitis, Dis. Colon Rectum 52 (2009) 193–197. [15] E. Kaczmarek, K. Lacka, P. Majewski, P. Trzeciak, A. Gorna, D. JarmolowskaJurczyszyn, A. Kluk, Selected markers of proliferation and apoptosis in the parathyroid lesions: a spatial visualization and qualification, J. Mol. Histol. 39 (2008) 509–517. ˛ [16] E. Kaczmarek, K. Lacka, D. Jarmolowska-Jurczyszyn, P. ASidor, Majewski, Changes of B and T lymphocytes and selected apopotosis markers in Hashimoto’s thyroiditis, J. Clin. Pathol. 64 (2011) 626–630. [17] M. Kang, P. Edmundson, F. Araujo-Perez, A.N. McCoy, J. Galanko, T.O. Keku, Association of plasma endotoxin, inflammatory cytokines and risk of colorectal adenomas, Cancer 13 (2013) 91–98. [18] J. Konsti, M. Lundin, N. Linder, C. Haglund, C. Blomqvist, H. Nevanlinna, K. Aaltonen, S. Nordling, J. Lundin, Effect of image compression and scaling on automated scoring of immunohistochemical stainings and segmentation of tumor epithelium, Diagn. Pathol. 7 (2012) 7–29. [19] J. Kurtovic, I. Segal, I. Recent advances in biological therapy for inflammatory bowel disease, Trop. Gastroenterol. 25 (2004) 9–14. [20] T. Ljung, L.G. Axellson, M. Herulf, J.O. Lundberg, P.M. Hellstrom, Early changes in rectal nitric oxide and mucosal inflammatory mediators in Crohn’s colitis in response to infliximab treatment, Aliment. Pharmacol. Ther. 25 (2007) 925–932. [21] M.H. McLean, G.I. Murray, K.N. Stewart, G. Norrie, C. Mayer, G.L. Hold, J. Thompson, N. Fyfe, M. Hope, N.A. Mowat, J.E. Drew, E.M. El-Omar, The inflammatory microenvironment in colorectal neoplasia, PLoS ONE 7 (2011) e15366. [22] K.A. Matkowskyj, R. Cox, R.T. Jensen, R.V. Benya, Quantitative immunohistochemistry by measuring cumulative signal strength accurately measures receptor number, J. Histochem. Cytochem. 51 (2003) 205–214. [23] R. Matsuda, T. Koide, C. Tokoro, T. Yamamoto, T. Godai, T. Morohashi, Y. Fujita, D. Takahashi, I. Kawana, S. Suzuki, S. Umemura, Quantitative cytokine mRNA expression profiles in the colonic mucosa of patients with steroid naive ulcerative colitis during active and quiescent disease, Inflamm. Bowel Dis. 15 (2009) 328–334. [24] D.C. Montrose, N.A. Horelik, J.P. Madigan, G.D. Stoner, L.S. Wang, R.S. Bruno, H.J. Park, C. Giardina, D.W. Rosenberg, Anti-inflammatory effects of freezedried black raspberry powder in ulcerative colitis, Carcinogenesis 32 (2011) 343–350. [25] T. Olsen, R. Goll, G. Cui, A. Husebekk, B. Vonen, G.S. Birketvedt, J. Florholmen, Tissue levels of tumor necrosis factor-alpha correlates with grade of
154
E. Kaczmarek et al. / Pathology – Research and Practice 210 (2014) 147–154
inflammation in untreated ulcerative colitis, Scand. J. Gastroenterol. 42 (2007) 1312–1320. [26] T. Olsen, R. Goll, G. Cui, A. Husebekk, J. Florholmen, Infliximab therapy decreases the levels of TNF-alpha and IFN-gamma mRNA in colonic mucosa of ulcerative colitis, Scand. J. Gastroenterol. 44 (2009) 727–735. [27] T. Olsen, R. Goll, G. Cui, I. Christiansen, J. Florholmen, TNF-alpha gene expression in colorectal mucosa as a predictor of remission after induction therapy with infliximab in ulcerative colitis, Cytokine 46 (2009) 222–227. [28] A.H. Olson, Image Analysis Using the Aperio ScanScope. Technical Manual, Aperio Technologies Inc., 2006, http://www.quorumtechnologies.com/ pdf files/AnalysisWhitePaper.pdf
[29] K. Prasad, G.K. Prabhu, Image analysis tools for evaluation of microscopic views of immunohistochemically stained specimen in medical research – a review, J. Med. Syst. 36 (2012) 2621–2631. [30] D. Raman, P.J. Baugher, Y.M. Thu, A. Richmond, Role of chemokines in tumor growth, Cancer Lett. 256 (2007) 137–165. [31] Z. Theodosiou, I.N. Kasampalidis, G. Livanos, M. Zervakis, I. Pitas, K. Lyroudia, Automated analysis of FISH and immunohistochemistry images: a review, Cytometry 71A (2007) 439–450. [32] N.L. Vujanovic, Role of TNF family ligands in antitumor activity of natural killer cells, Int. Rev. Immunol. 20 (2001) 415–437.
