• View in gallery

    Photomicrographs of a rhodanine-stained histologic section of a scanned liver biopsy specimen from a dog with a hepatic copper concentration of 2,378 μg/g and scanned concentration of 1,992 μg/g before analytic markup (upper panel) and after analytic markup (lower panel). Notice marked distinction of copper granules and exclusion of lipofuscin and iron aggregates in macrophages (arrowheads [hue = 0.04; hue width = 0.1; color saturation = 0.1]).

  • View in gallery

    Linear regression analysis of measured hepatic copper concentration against a positive pixel algorithm for 139 canine biopsy specimens, determined from scanned images of rhodanine-stained liver sections.

  • View in gallery

    Linear regression analysis of measured canine hepatic copper concentration (n = 43 specimens; range, 93 to 498 μg/g) plotted against a positive pixel algorithm determined from scanned images of rhodanine-stained liver sections.

  • View in gallery

    Bland-Altman plot of the difference between paired measurements for canine liver biopsy specimens plotted against the mean copper concentration of paired specimens.

  • View in gallery

    Calculated liver copper concentrations for 17 entire canine liver biopsy specimens (whole biopsy specimen examined), 17 specimens with regions of lost (extinct) parenchyma and fibrosis, and 17 specimens with regenerative nodules. Concentrations were significantly higher in entire biopsy specimens than in the other 2 categories.

  • 1. Baggenstoss AH, Foulk WT, Butt HR, et al. The pathology of primary biliary cirrhosis with emphasis on histogenesis. Am J Clin Pathol 1964; 42:259276.

    • Search Google Scholar
    • Export Citation
  • 2. Johnston AN, Center SA, McDonough SP, et al. Influence of biopsy specimen size, tissue fixation, and assay variation on copper, iron, and zinc concentrations in canine livers. Am J Vet Res 2009; 70:15021511.

    • Search Google Scholar
    • Export Citation
  • 3. Lee Y-M, Kaplan MM. Primary sclerosing cholangitis. N Engl J Med 1995; 332:924933.

  • 4. Schlichting P, Holund B, Poulsen H. Liver biopsy in chronic aggressive hepatitis: diagnostic reproducibility in relation to size of specimen. Scand J Gastroenterol 1983; 18:2732.

    • Search Google Scholar
    • Export Citation
  • 5. Twedt DC, Sternlieib I, Gilbertsosn SR, et al. Clinical, morphologic, and chemical studies on copper toxicosis of Bedlington Terriers. J Am Vet Med Assoc 1979; 175:269275.

    • Search Google Scholar
    • Export Citation
  • 6. Sheehan D, Hrapchak B. Rhodanine stain. In: Theory and practice of histotechnology. 2nd ed. St Louis: Mosby, 1980;230.

  • 7. Aperio Image Analysis user's guide. Vista, Calif: Aperio Technologies Inc, 2008.

  • 8. Faa G, Nurchi V, Demelia L, et al. Uneven hepatic copper distribution in Wilson's disease. J Hepatol 1995; 22:303308.

  • 9. Crawford AR, Lin X-Z, Crawford JM. The normal adult human liver biopsy: a quantitative reference standard. Hepatology 1998; 28:323331.

  • 10. Guido M, Mangia A, Faa G. Chronic viral hepatitis: the histology report Z. Dig Liver Dis 2011; 43(suppl 4):S331S343.

  • 11. Cole TL, Center SA, Flood SN, et al. Diagnostic comparison of needle and wedge biopsy specimens of the liver in dogs and cats. J Am Vet Med Assoc 2002; 220:14831490.

    • Search Google Scholar
    • Export Citation
  • 12. Johnson GF, Gilbertson SR, Goldfischer S, et al. Cytochemical detection of inherited copper toxicosis of Bedlington Terriers. Vet Patho1 1984; 21:5760.

    • Search Google Scholar
    • Export Citation
  • 13. Thornburg LP, Beissenherz M, Dolan M, et al. Histochemical demonstration of copper and copper-associated protein in the canine liver. Vet Pathol 1985; 22:327332.

    • Search Google Scholar
    • Export Citation
  • 14. Ludwig J, Owen CA Jr, Barham SS, et al. The liver in the inherited copper disease of Bedlington Terriers. Lab Invest 1980; 43:8287.

Advertisement

Digital image analysis of rhodanine-stained liver biopsy specimens for calculation of hepatic copper concentrations in dogs

View More View Less
  • 1 Department of Clinical Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, 14853.
  • | 2 Department of Biomedical Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, 14853.
  • | 3 Department of Biomedical Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, 14853.

Abstract

Objective—To evaluate the accuracy of digitally scanned rhodanine-stained liver biopsy specimens for determination of hepatic copper concentration and compare results with qualitatively assigned histologic copper scores in dogs.

Sample—353 liver biopsy specimens from dogs.

Procedures—Specimens (n = 139) with quantified copper concentration ranging from 93 to 6,900 μg/g were allocated to group 1 (< 400 μg/g [37]), group 2 (401 to 1,000 μg/g [27]), group 3 (1,001 to 2,000 μg/g [34]), and group 4 (> 2,001 μg/g [41]); stained with rhodanine; and digitally scanned and analyzed with a proprietary positive pixel algorithm. Measured versus calculated copper concentrations were compared, and limits of agreement determined. Influence of nodular remodeling, fibrosis, or parenchymal loss on copper concentration was determined by digitally analyzing selected regions in 17 specimens. After method validation, 214 additional liver specimens underwent digital scanning for copper concentration determination. All sections (n = 353) were then independently scored by 2 naive evaluators with a qualitative scoring schema. Agreement between assigned scores and between assigned scores and tissue copper concentrations was determined.

Results—Linear regression was used to develop a formula for calculating hepatic copper concentration ≥ 400 μg/g from scanned sections. Copper concentrations in unremodeled specimens were significantly higher than in remodeled specimens. Qualitative scores widely overlapped among quantitative copper concentration groups.

Conclusions and Clinical Relevance—Calculated copper concentrations determined by means of digital scanning of rhodanine-stained liver sections were highly correlated with measured values and more accurate than qualitative copper scores, which should improve diagnostic usefulness of hepatic copper concentrations and assessments in sequential biopsy specimens.

Abstract

Objective—To evaluate the accuracy of digitally scanned rhodanine-stained liver biopsy specimens for determination of hepatic copper concentration and compare results with qualitatively assigned histologic copper scores in dogs.

Sample—353 liver biopsy specimens from dogs.

Procedures—Specimens (n = 139) with quantified copper concentration ranging from 93 to 6,900 μg/g were allocated to group 1 (< 400 μg/g [37]), group 2 (401 to 1,000 μg/g [27]), group 3 (1,001 to 2,000 μg/g [34]), and group 4 (> 2,001 μg/g [41]); stained with rhodanine; and digitally scanned and analyzed with a proprietary positive pixel algorithm. Measured versus calculated copper concentrations were compared, and limits of agreement determined. Influence of nodular remodeling, fibrosis, or parenchymal loss on copper concentration was determined by digitally analyzing selected regions in 17 specimens. After method validation, 214 additional liver specimens underwent digital scanning for copper concentration determination. All sections (n = 353) were then independently scored by 2 naive evaluators with a qualitative scoring schema. Agreement between assigned scores and between assigned scores and tissue copper concentrations was determined.

Results—Linear regression was used to develop a formula for calculating hepatic copper concentration ≥ 400 μg/g from scanned sections. Copper concentrations in unremodeled specimens were significantly higher than in remodeled specimens. Qualitative scores widely overlapped among quantitative copper concentration groups.

Conclusions and Clinical Relevance—Calculated copper concentrations determined by means of digital scanning of rhodanine-stained liver sections were highly correlated with measured values and more accurate than qualitative copper scores, which should improve diagnostic usefulness of hepatic copper concentrations and assessments in sequential biopsy specimens.

Increasing prevalence of copper-associated hepatopathy has been suspected in Labrador Retrievers and other purebred and mixed-breed dogs since 1997 on the basis of findings in biopsy specimens submitted to the diagnostic pathology service at the New York State College of Veterinary Medicine, Cornell University. Investigation of this phenomenon among other large- and small-breed dogs requires objective quantification of liver copper concentrations. At present, this is achieved with fresh or fresh frozen tissue, tissue fixed in neutral-buffered 10% formalin, or formalin-fixed tissues embedded in paraffin (which require paraffin extraction). However, variation in copper distribution in a biopsy specimen associated with architectural remodeling can lead to an erroneous assessment of the severity of hepatic copper accumulation.1–4 On the basis of microscopic assessment of rhodanine-stained liver biopsy specimens in dogs with copper-associated hepatopathy, this appears to reflect areas of parenchymal loss, extensive fibrosis, or regenerative nodules, which seemingly have lower copper concentrations than unremodeled tissue. Consequently, because quantification of liver copper is conventionally done with a small specimen of biopsied tissue, measurements may not be consistent with the copper concentration in most of the liver. This phenomenon hinders accurate assessment of hepatic copper status made on the basis of sequential liver biopsy specimens. We reasoned that developing a method for liver copper quantification that used the same biopsy sections examined microscopically for histologic assessment might offer an alternative and more reliable assessment strategy.

Qualitative histologic grading of the severity of hepatocellular copper accumulation (qualitative scores range from 0 to 5) is typically considered with the amount of measured copper to guide treatment recommendations. However, variability in histologic scores may occur because of differences in interpretation of biopsy specimens among pathologists. Most dogs with mild accumulation of copper (copper granules in occasional hepatocytes [< 2 hepatocytes/10× {objective lens} field; grade 1]) or obvious small to moderate numbers of copper granules (< 50% of zone 3 hepatocytes [< 4 clusters of cells containing copper granules/10× field; grade 2]) usually have no clinical signs of hepatic injury, may or may not have a mild to modest increase in alanine aminotransferase activity, and may or may not have histologic evidence of hepatocyte injury. Continued monitoring of liver enzyme activities for dogs with grade 1 copper accumulation is recommended, whereas treatment of dogs with group 2 copper accumulation includes administration of antioxidants, restriction of copper intake in food and water, and 2 to 4 months of chelation treatment. Dogs in which all hepatic lobules contain groups of hepatocytes (> 50% but < 75% of zone 3 hepatocytes with moderate to large numbers of copper granules; grade 3), dogs in which most (> 75%) of zone 3 hepatocytes contain moderate to large numbers of copper positive granules (grade 4), and dogs with panlobular copper accumulation (grade 5) have vague and cyclic clinical signs (eg, inappetance, weight loss, vomiting, or lethargy) associated with progressive or sustained increases in alanine aminotransferase activity associated with zone 3 histologic hepatocellular injury. However, for some of these dogs, lesions and copper accumulations are serendipitously discovered when liver biopsy specimens are collected during other procedures (eg, exploratory laparotomy for enteric foreign body removal or splenectomy for bleeding mass lesions). Dogs with severe copper-associated liver injury develop jaundice, ascites, and hepatic failure and may be in critical condition when first evaluated.5 Dogs categorized as having grade 3 to grade 5 copper-associated liver disease are treated with a minimum of 6 months of chelation along with antioxidants and restricted copper intake, and lifelong strategies to avoid reaccumulation of hepatic copper (ie, every-other-day administration of a chelator or long-term oral administration of zinc, and restricted copper intake). Chelation therapy is expensive, may be associated with drug toxicosis, and is often prescribed long term in dogs with grade 3 to 5 copper accumulation. Thus, accurate assessment of tissue copper concentrations has relevance in both diagnosing and managing copper-associated hepatopathy. Furthermore, sequential assessment of hepatic copper concentration may also be used to refine long-term treatment.

The purpose of the study reported here was to evaluate the accuracy of digitally scanned rhodanine-stained liver biopsy specimens for determination of hepatic copper concentration, evaluate variation in tissue copper concentrations in unremodeled and remodeled liver (regenerative nodules and areas with parenchymal loss and fibrosis), and compare results with qualitatively assigned histologic copper scores in dogs. We hypothesized that copper concentrations determined from a digitally scanned rhodanine-stained liver sample could replace analysis of a separate portion of liver, thereby reducing discordance between measured and qualitatively estimated copper accumulation.

Materials and Methods

Liver sections used in this project were obtained either antemortem as liver biopsy specimens (n = 213) during general anesthesia for diagnostic purposes or after euthanasia (140). Care provided for each animal involved in this study was in compliance with the recommendations of the Institutional Animal Use and Care policies of Cornell University.

Biopsy specimens from 139 dogs (89 Labrador Retrievers, 18 Doberman Pinschers, 9 mixed-breed dogs, 4 Cocker Spaniels, 3 West Highland White Terriers, 2 Pembroke Welsh Corgis, 2 Bedlington Terriers, and 1 each of 12 additional breeds) that had liver copper concentration quantified by atomic absorption spectroscopya from 93 to 6,900 μg/g (dry-matter basis) were selected for method validation from the liver biopsy database at the College of Veterinary Medicine, Cornell University. Tissue specimens were large wedge biopsy specimens (n = 102; 85% with ≥ 2 tissue sections) or 14-gauge-needle biopsy specimens (with at least three 1.5-cm-long cores). Sections of liver were stained for copper with the rhodanine method and standard histologic techniques6 with a minor variation; after thorough mixing of the rhodanine-saturated stock solution, 4 mL was diluted in 50 mL of deionized water, and the working solution was preheated to 58°C for 30 minutes, immediately prior to use. A qualitative histologic copper score was assigned on the basis of H&E- and rhodanine-stained sections, according to a scheme (Appendix).

Tissues used for method validation included at least 27 specimens each with hepatic copper concentration (on a dry-matter basis) in 4 targeted ranges for linear regression analysis: group 1, ≤ 400 μg/g, representing the reference range for hepatic copper concentration; group 2, 401 to 1,000 μg/g; group 3, 1,001 to 2,000 μg/g; and group 4, ≥ 2,000 μg/g. Groups were designated with consideration that liver copper concentrations categorized clinically as within the reference limit (< 400 μg/g), mildly increased (401 to 1,000 μg/g), moderately increased (1,001 to 2,000 μg/g), or severely increased (> 2,000 μg/g) influence clinical treatment recommendations.

All rhodanine-stained sections were microscopically inspected for staining adequacy before inclusion in the study and performance of digitized imaging procedures. Staining quality also was confirmed by examination of positive and negative copper control tissue sections stained concurrently. A 20× magnification digital image was created from a 4-μm-thick section of liver tissue stained with rhodanine.b Each study section was digitally scanned and analyzed without knowledge of the quantified tissue copper concentration. The entire biopsy specimen (operator selected the entire biopsy area by digitally marking the tissue perimeter) was scanned with numerous manually applied focus points placed in the section perimeter and interior. After scanning was complete, the automated focus report was evaluated, and if the report was less than optimal, the section was rescanned with additional focus points applied. The copper pigment in the digital images was quantified by use of a positive pixel algorithmc following the manufacturer's recommendations.7 After the hue value was set (0.04, orange), the input parameters for hue range (0.08 to 0.2) and color saturation threshold (0.08 to 0.3) were fine-tuned empirically for each slide by an operator who was not aware of the atomic absorption spectroscopy results. Software default intensity thresholds (unitless measures for the brightness of the pixels) for weak signals (220 to 175), medium signals (175 to 100), and strong signals (100 to 0) were maintained. Fine-tuning was accomplished by use of a markup image and visual inspection that confirmed inclusion of all copper-positive granules and exclusion of lipofuscin (yellow-brown hue, 0.12), hemosiderin (brown hue, 0.1), and RBCs (hue, 0). The total intensity and number of positive, strong-positive, weak-positive, and negative signals (pixels) and total area examined (mm2) were determined.

After validating the accuracy of calculating hepatic copper concentrations from digitally scanned rhodanine-stained liver specimens, an additional 214 liver biopsy specimens from Labrador Retrievers were scanned and tissue copper concentrations were calculated. Qualitative tissue copper scores were assigned independently by 2 investigators (SAC and SPM) naive to tissue copper concentrations, for each rhodanine-stained liver biopsy specimen (n = 353). Agreement between assigned scores was determined. Hepatic copper concentrations were then compared with assigned copper scores to determine overlap of qualitative scoring among the quantitative tissue copper categories.

Seventeen biopsy specimens with marked regenerative nodules, 11 of which included regions of parenchymal loss and fibrosis, were used to investigate the influence of tissue remodeling on hepatic copper concentrations. In these specimens, hepatic copper concentrations were calculated for the entire biopsy specimen and then independently for remodeled regions that were at least 3 × 7 mm in dimension, consistent with a small-needle biopsy.

Statistical analysis—Measured and calculated hepatic copper concentrations in each group were non-Gaussian in distribution as determined by box-and-whisker plots and the Kolmogorov-Smirnov test. Copper data accordingly are reported as median and range values. Linear regression was used to calculate tissue copper concentrations with a positive pixel algorithm according to the following relationship: (total positive plus total strong positive pixels)/(total pixels [positive and negative] minus weak positive pixels); the F statistic and R2 were calculated, and the 95% confidence and 95% prediction intervals were determined. Because the linear regression formula based on analysis of all specimens determined a formula (346.6 + 64,102 × positive pixel algorithm calculation) that limited accuracy for specimens in the reference range, a linear regression analysis was also performed for specimens with a concentration < 500 μg/g, which was thereafter used to determine copper concentrations when the primary linear regression calculated a copper value ≤ 400 μg/g. A Bland-Altman plot of the difference of the mean copper concentration of paired specimens (measured and calculated copper concentrations on entire biopsy sections) was used to examine the difference in quantified and calculated tissue copper concentrations. A Wilcoxon signed rank test was used to assess differences between measured and calculated tissue copper concentrations in each group and to investigate differences in calculated copper concentrations between an entire biopsy specimen and remodeled regions (ie, nodular regeneration and fibrosis with parenchymal loss). A 2-sided P value of < 0.05 identified significant differences. The number of specimens in each group (groups 1 through 5) with a calculated copper score designating a different group assignment (that might influence treatment recommendations) also was enumerated. A Spearman rank correlation was used to assess the association between qualitative copper scores and tissue copper concentrations in 353 specimens. The overlap of qualitatively assigned copper scores in each group category was detailed in table format. Statistical analyses were performed with commercial software.d,e Values of P < 0.05 were considered significant.

Results

Photomicrographs of digitally scanned rhodanine-stained liver sections were obtained (Figure 1). Linear regression analysis revealed a significant linear fit (R2 = 0.97; P < 0.001) between measured and calculated hepatic copper concentrations determined with the positive pixel algorithm (Figure 2) in liver specimens from 139 dogs. A second linear regression analysis (y = 194.7 + 105,694x; R2 = 0.65; P < 0.001), developed to more accurately predict tissue copper concentrations < 400 μg/g, was determined with data from 43 biopsy specimens (measured copper concentration range, 93 to 498 μg/g [Figure 3]).

Figure 1—
Figure 1—

Photomicrographs of a rhodanine-stained histologic section of a scanned liver biopsy specimen from a dog with a hepatic copper concentration of 2,378 μg/g and scanned concentration of 1,992 μg/g before analytic markup (upper panel) and after analytic markup (lower panel). Notice marked distinction of copper granules and exclusion of lipofuscin and iron aggregates in macrophages (arrowheads [hue = 0.04; hue width = 0.1; color saturation = 0.1]).

Citation: American Journal of Veterinary Research 74, 12; 10.2460/ajvr.74.12.1474

Figure 2—
Figure 2—

Linear regression analysis of measured hepatic copper concentration against a positive pixel algorithm for 139 canine biopsy specimens, determined from scanned images of rhodanine-stained liver sections.

Citation: American Journal of Veterinary Research 74, 12; 10.2460/ajvr.74.12.1474

Figure 3—
Figure 3—

Linear regression analysis of measured canine hepatic copper concentration (n = 43 specimens; range, 93 to 498 μg/g) plotted against a positive pixel algorithm determined from scanned images of rhodanine-stained liver sections.

Citation: American Journal of Veterinary Research 74, 12; 10.2460/ajvr.74.12.1474

The median calculated and measured copper concentrations were significantly different only for group 3 (P = 0.003 [Table 1]), although median values were in the designated range. The primary regression formula segregated specimens with reference range copper concentrations from those with increased hepatic copper concentrations, accurately estimating hepatic copper concentrations > 347 μg/g. Six of 139 specimens (measured copper concentrations of 1,260, 2,130, 2,900, 3,040, 3,140, and 4,400 μg/g) had calculated copper concentrations outside the 95% prediction interval of the linear regression. However, calculated copper concentrations for these specimens did not change group categorization for any individual.

Table 1—

Measured and calculated copper concentrations (μg/g of dry weight of liver) in 139 canine liver biopsy specimens in 4 groups determined by quantitative copper concentration.

GroupCopper concentration (μg/g)No. of specimensMeasured copper (μg/g)Calculated copper (μg/g)
Group 1< 40037219 (93–398)215 (195–345)
Group 2400–1,00027613 (427–981)532 (393–1284)
Group 31,001–2,000341,251 (1,015–1,915)1,172 (705–1,825)
Group 4≥ 2,001413,160 (2,030–6,390)3,154 (1,700–6,267)

Calculated hepatic copper concentrations changed group designation assigned by quantified copper concentrations in 0%, 22.2%, 11.8%, and 14.6% of dogs in groups 1 to 4, respectively. In each instance in which group designation would be altered, the magnitude of disagreement was relative to the measured group copper concentration (ie, larger differences occurred in group 4, which had the highest tissue copper concentration). Differences between quantified and calculated hepatic copper concentrations in specimens with discordant group designations were 46, 67, 126, 140, 142, and 332 μg/g in group 2; 183, 235, 373, and 555 μg/g in group 3; and 305, 330, 332, 470, 497, and 501 μg/g in group 4. Discordant group classification involved specimens with measured copper concentrations that were generally close to the calculated values (Table 1; differences were < 133 μg/g for group 2, < 261 μg/g for group 3, and < 371 μg/g for group 4). In 11 of 16 specimens with changed group categorization, calculated copper concentration moved classification to the immediately lower group. Only a single specimen with a quantified copper concentration of 553 μg/g in group 2 was downgraded to group 1 (reference range). Thus, in only a single specimen would redesignation have possibly changed treatment considerations.

A Bland-Altman plot (Figure 4) of the difference between paired (measured and calculated) copper concentrations for a single sample plotted against the mean copper concentration of the paired specimens identified 5 specimens outside the 95% confidence interval for the limits of agreement. Nevertheless, calculated copper concentrations for these specimens did not change group designation and thus would not have influenced clinical case management. Each of these specimens had measured copper concentrations exceeding 2,000 μg/g.

Figure 4—
Figure 4—

Bland-Altman plot of the difference between paired measurements for canine liver biopsy specimens plotted against the mean copper concentration of paired specimens.

Citation: American Journal of Veterinary Research 74, 12; 10.2460/ajvr.74.12.1474

As expected, calculated copper concentrations were significantly greater for an entire biopsy specimen (n = 17; median, 2,205 μg/g [range, 1,108 to 9,506 μg/g]) compared with regions of regenerative nodules (17; median, 1,042 μg/g [range, 299 to 5,567 μg/g]; P = 0.005) or parenchymal loss and fibrosis (11; median, 1,052 μg/g [range, 290 to 3263 μg/g]; P = 0.004; Figure 5). No significant difference (P = 0.2) was found in copper concentration in regenerative nodules, compared with regions of fibrosis and parenchymal loss.

Figure 5—
Figure 5—

Calculated liver copper concentrations for 17 entire canine liver biopsy specimens (whole biopsy specimen examined), 17 specimens with regions of lost (extinct) parenchyma and fibrosis, and 17 specimens with regenerative nodules. Concentrations were significantly higher in entire biopsy specimens than in the other 2 categories.

Citation: American Journal of Veterinary Research 74, 12; 10.2460/ajvr.74.12.1474

Evaluation of agreement between assigned copper scores by the 2 investigators yielded a κ statistic of 0.84 (P < 0.001). Although qualitative copper scores were significantly correlated with measured and calculated copper concentrations (R = 0.84; P < 0.001), there was broad overlap in qualitative copper scores that were assigned to biopsy specimens with a wide range of copper concentrations (Table 2). Although many dogs (104/164 [63%]) with liver copper concentration in the reference range (≤ 400 μg/g) were assigned a qualitative copper score of 0, the remainder had microscopically visible copper granules, with 9 (5.5%) dogs assigned qualitative copper scores ≥ 2. Additionally, of 88 dogs in group 2, some (10/88 [11.4%]) lacked visible copper granules on microscopic examination. The remaining dogs in this group were assigned qualitative copper scores that ranged from 1 through 4. Of dogs in group 3, 2 lacked visible copper granules, whereas the remainder had qualitative scores that ranged from 2 through 5. All dogs in group 4 were assigned qualitative copper scores ≥ 3. Because some dogs in each group were assigned a qualitative score of 3 and dogs in groups 2 through 4 were assigned qualitative copper scores of 4, there was broad variability in visual scoring that precluded accurate group designation.

Table 2—

Qualitative copper scores in 353 canine liver biopsy specimens in 4 groups determined by quantitative copper concentration (μg/g of dry weight of liver).

   Qualitative copper score
GroupCopper concentration (μg/g)No. of specimens012345
Group 1< 400164104519000
Group 2400–1,000881020243040
Group 31,001–2,0005721232164
Group 4> 2,00144000101618

Discussion

The apparent increase in hepatic copper concentrations and associated liver injury seen in dogs during the last 15 years at the authors' institution has necessitated determination of liver copper concentrations in an expanding subset of liver biopsy specimens. Copper-associated liver injury is not always anticipated by clinicians, which necessitates retrospective retrieval of sampled tissue for copper analysis that delays case assessment and treatment recommendations. Because only a portion of a liver biopsy specimen or a separate tissue sample is submitted for quantitative copper analysis, such assessments should always be reconciled with qualitative assessments to guard against erroneous conclusions. Findings suggested that each method of copper assessment (measured single sample vs qualitative scoring) is subject to errors. Subjectivity of qualitative assessments is reduced by adherence to a schema that includes consideration of zonal involvement and the number and percentage of hepatocytes containing copper granules. However, findings clearly indicated that assigned qualitative copper scores were unable to clearly differentiate between quantitative copper categories because broadly overlapping scores for tissues with a wide range of copper concentrations were evident.

Significant variation in tissue copper distribution was associated with architectural remodeling (lower copper concentrations in regions of nodular regeneration, parenchymal loss, or fibrosis). This was confirmed for the first time in canine biopsy specimens by use of the digital scanning methodology. Variation in distribution of tissue copper is recognized to influence assessment of hepatic copper in humans with Wilson's disease, where it compromises interpretation of quantitative metal analyses and prospective qualitative assessments.8 Unequal distribution of copper in tissue sections likely contributed to discordance in some specimens in the present study, particularly those identified by the Bland-Altman plot.

Findings in this study caution against the use of qualitative or quantitative assessments as stand-alone methods for deducing tissue copper concentration used for recommending treatment interventions or for sequentially assessing response to chelation treatments. Any method of hepatic copper assessment is further compromised by evaluation of tiny tissue specimens. Qualitative sample assessment requires availability of enough acinar units to allow determination of zonal histologic lesions and copper distribution.9–11 Quantitative sample evaluation requires a minimum of 15 mg of wet weight tissue.2 These criteria are consistent with those recommended for best assessment of hepatic copper status in humans, in which nonuniform lesions, nonuniform copper distribution, and small sample size influence qualitative copper assessments and histologic staging.1,3,4,9,10 Small specimens retrieved by needle core biopsy devices are especially problematic in a fibrotic liver because such specimens often fragment or are incompletely retrieved.11 Because these circumstances thwart application of the digitized image method validated in the present study, small tissue specimens were intentionally excluded from the study.

Findings of this study substantiated good clinical application for digital scanning of an entire liver biopsy specimen stained with rhodanine for tissue copper determination in canine liver. Only 5 of 139 (3.6%) specimens had calculated copper concentrations outside the 95% prediction interval of the primary linear regression formula. In each of these 5 instances, there would have been no clinically relevant consequence because each dog had a markedly increased tissue copper concentration (> 2,000 μg/g); there was no change in copper group categorization. Discordance in these instances likely reflected architectural remodeling and tissue copper distribution that influenced measured copper concentrations. It has been suggested that rhodanine staining of canine liver tissue limits visual detection of copper granules to specimens with a concentration ≥ 400 μg/g.12–14 Yet findings of the present study confirmed that stainable hepatic copper is seen and can be detected by digitized scanning at concentrations as low as 200 μg/g. Although tissue copper concentrations in the reference range were best estimated from digitally scanned images by use of a second linear regression formula, this procedure would be unwarranted in diagnostic specimens assigned a qualitative copper score of 0 and with a calculated copper concentration of ≤ 400 μg/g.

Even though the calculated copper concentration changed quantitative group classification in 22.2%, 11.8%, and 14.6% of dogs in groups 2, 3, and 4, respectively, the differences were small and would not have changed treatment strategies. In only a single group 2 dog with a slightly increased measured hepatic copper concentration (533 μg/g) did calculated copper concentration downgrade group classification to within the reference range (calculated copper concentration was 393 μg/g). However, this discrepancy would have been evident upon microscopic evaluation of the rhodanine-stained biopsy specimen. It also is important to acknowledge that the apparent variation between methods of copper determination also may reflect the measured copper values. We previously reported an interassay coefficient of variation ranging from 36% to 50% for copper concentrations in liver tissue containing low and high copper concentrations, respectively.2 This is proposed to reflect differences in tissue copper distribution in liver biopsy specimens because analytic standards used for method validation have a coefficient of variation < 4%.

Digital scanning of rhodanine-stained liver sections and determination of liver copper concentration with a positive pixel algorithm offered a distinct advantage over qualitative copper assessments that may be influenced by inter- and intraobserver differences and biases, and also offered an advantage over tissue copper quantification because the entire microscopic biopsy specimen, rather than a separate specimen, was used to determine the tissue copper calculation. This obviates disparity caused by distributional differences in copper concentration between specimens designated for microscopic and quantitative evaluations. These findings suggested that digital scanning of rhodanine-stained liver sections will improve sequential biopsy specimen assessments used to judge efficacy of treatment for excessive hepatic copper retention. However, this method requires operator expertise in pathology of the liver; this expertise is needed for fine-tuning the hue range and color saturation thresholds to exclude interfering substances (ie, hemosiderin, lipofuscin, and erythrocyte congestion from the orange-red rhodanine signal) and to optimize signal detection (hue intensity).

a.

Varian AA-1275, Varian Inc, Palo Alto, Calif.

b.

Aperio ScanScope CS, Aperio Technologies Inc, Vista, Calif.

c.

Aperio Image Analysis, version 8.1, Aperio Technologies Inc, Vista, Calif.

d.

Statistix 9, Analytical Software, Tallahassee, Fla.

e.

Analyse-it, version 2.26 for Excel 12+, Analyse-it Software Ltd, Leeds, West Yorkshire, England.

References

  • 1. Baggenstoss AH, Foulk WT, Butt HR, et al. The pathology of primary biliary cirrhosis with emphasis on histogenesis. Am J Clin Pathol 1964; 42:259276.

    • Search Google Scholar
    • Export Citation
  • 2. Johnston AN, Center SA, McDonough SP, et al. Influence of biopsy specimen size, tissue fixation, and assay variation on copper, iron, and zinc concentrations in canine livers. Am J Vet Res 2009; 70:15021511.

    • Search Google Scholar
    • Export Citation
  • 3. Lee Y-M, Kaplan MM. Primary sclerosing cholangitis. N Engl J Med 1995; 332:924933.

  • 4. Schlichting P, Holund B, Poulsen H. Liver biopsy in chronic aggressive hepatitis: diagnostic reproducibility in relation to size of specimen. Scand J Gastroenterol 1983; 18:2732.

    • Search Google Scholar
    • Export Citation
  • 5. Twedt DC, Sternlieib I, Gilbertsosn SR, et al. Clinical, morphologic, and chemical studies on copper toxicosis of Bedlington Terriers. J Am Vet Med Assoc 1979; 175:269275.

    • Search Google Scholar
    • Export Citation
  • 6. Sheehan D, Hrapchak B. Rhodanine stain. In: Theory and practice of histotechnology. 2nd ed. St Louis: Mosby, 1980;230.

  • 7. Aperio Image Analysis user's guide. Vista, Calif: Aperio Technologies Inc, 2008.

  • 8. Faa G, Nurchi V, Demelia L, et al. Uneven hepatic copper distribution in Wilson's disease. J Hepatol 1995; 22:303308.

  • 9. Crawford AR, Lin X-Z, Crawford JM. The normal adult human liver biopsy: a quantitative reference standard. Hepatology 1998; 28:323331.

  • 10. Guido M, Mangia A, Faa G. Chronic viral hepatitis: the histology report Z. Dig Liver Dis 2011; 43(suppl 4):S331S343.

  • 11. Cole TL, Center SA, Flood SN, et al. Diagnostic comparison of needle and wedge biopsy specimens of the liver in dogs and cats. J Am Vet Med Assoc 2002; 220:14831490.

    • Search Google Scholar
    • Export Citation
  • 12. Johnson GF, Gilbertson SR, Goldfischer S, et al. Cytochemical detection of inherited copper toxicosis of Bedlington Terriers. Vet Patho1 1984; 21:5760.

    • Search Google Scholar
    • Export Citation
  • 13. Thornburg LP, Beissenherz M, Dolan M, et al. Histochemical demonstration of copper and copper-associated protein in the canine liver. Vet Pathol 1985; 22:327332.

    • Search Google Scholar
    • Export Citation
  • 14. Ludwig J, Owen CA Jr, Barham SS, et al. The liver in the inherited copper disease of Bedlington Terriers. Lab Invest 1980; 43:8287.

Appendix

Scheme for classification of copper content in canine liver biopsy specimens.

Grade 0: Copper not detected, care taken to avoid confusing lipofuscin or hemosiderin with copper granules.
Grade 1: Variable copper granules in an occasional hepatocyte (< 2 hepatocytes/10× field).
Grade 2: Obvious small to moderate numbers of copper granules in < 50% of zone 3 hepatocytes (or < 4 clusters/10× field).
Grade 3: All lobules contain groups of hepatocytes (> 50% but < 75%) predominantly in zone 3 with moderate to large numbers of copper granules. Macrophages with copper granules may also be present
Grade 4: Most (> 75%) zone 3 hepatocytes contain moderate to large numbers of copper granules. Macrophages with copper granules may also be present.
Grade 5: Panlobular presence of large numbers of copper granules in hepatocytes, usually associated with copper-containing macrophages. Tissue remodeling may obscure lobular zones.

Contributor Notes

Address correspondence to Dr. Center (sac6@cornell.edu).