OUP user menu

PAX5 Expression in Nonhematopoietic Tissues
Reappraisal of Previous Studies

Daniel A. Morgenstern PhD, Fyeza Hasan MRCPCH, Sian Gibson MSc, Paul Winyard PhD, Neil J. Sebire FRCPath, John Anderson PhD
DOI: http://dx.doi.org/10.1309/AJCPZPQN0LUGKMME 407-415 First published online: 1 March 2010

Abstract

The Pax gene family encodes transcription factors with similar structures but distinctive roles in development and with limited expression in adult tissues. Reexpression of PAX proteins is frequently observed in human cancers, reflecting recapitulation of embryologic or developmental function. Defining expression of PAX family members is important in the immunohistochemical differential diagnosis of cancer, understanding oncogenesis, and defining targets for therapy. Immunostaining for PAX5 has become a commonly used technique in differential diagnosis of B-lineage hematologic malignancies. In seeking to define the range and degree of expression of PAX5 in nonhematologic pediatric cancers by immunohistochemical analysis with the anti-PAX5 monoclonal antibody routinely used in research and diagnosis, we observed strong immunostaining in a number of malignant tissues, including Wilms tumor. The pattern of expression of PAX5 in Wilms tumor was found to be identical to that of PAX2, raising the possibility of antibody cross-reactivity. This was subsequently confirmed by Western blotting and immunostaining of transfected cells and quantitative reverse transcriptase–polymerase chain reaction. Since the same PAX5 monoclonal antibody has been used consistently in the literature, these findings indicate a need for reappraisal of published PAX5 immunostaining results.

Key Words:
  • PAX5
  • PAX2
  • Wilms tumor
  • Immunohistochemistry
  • Antibody
  • Monoclonal

Pax genes constitute a family of 9 transcription factors that have important roles in regulating cell proliferation, migration, and differentiation during embryonic development and organogenesis.1 In recent years, there has been a growing interest in the role these genes may have in the development of human cancers, with dysregulated expression of various PAX proteins demonstrated in a range of tumor types.2 In alveolar rhabdomyosarcoma, for example, PAX3 and PAX7 are frequently expressed as chimeric proteins with the forkhead transcription factor (FKHR/FOXO1A), as a result of translocations between chromosome 2 (Pax3) or 1 (Pax7) and chromosome 13.3 The resulting fusion protein contributes to malignant transformation and may have a role in tumor immune escape.4

PAX5 (also known as B cell–specific activator protein) is a transcription factor essential for B-lymphocyte development. Within the hematopoietic system, expression is limited to cells of the B-lymphocyte lineage, and PAX5 expression has, therefore, found diagnostic use in defining B-cell lineage in a variety of hematologic malignancies.57 In recent years, there have been a number of studies examining the expression of PAX5 in nonhematologic malignancies. Our group recently identified PAX5 as a target antigen for tumor immunotherapy8 and, therefore, set out to investigate PAX5 expression in a variety of pediatric extracranial solid tumors. Subsequently, concerns about the possibility of cross-reactivity with other PAX proteins, particularly PAX2, led us to investigate the specificity of the most commonly used anti-PAX5 monoclonal antibody.

Materials and Methods

Tissue Arrays

Tissue samples from a range of pediatric extracranial solid tumors were obtained from hospital histology archived material: peripheral neuroectodermal tumor (n = 17), undifferentiated sarcoma (n = 6), neuroblastoma (n = 24), malignant rhabdoid tumor (n = 29), cystic mesoblastic nephroma (n = 13), rhabdomyosarcoma (n = 24), and Wilms tumor (n = 36). Following examination of an H&E-stained section, representative cores were removed from paraffin blocks with a skin biopsy punch and reembedded to produce tissue arrays for each tumor type. Osteosarcoma tissue arrays (99 tumors) were kindly provided by Adrienne Flanagan (Royal National Orthopaedic Hospital, Stanmore, England).

Immunohistochemical Analysis

Immunohistochemical analysis for PAX5 was performed on 4-μm-thick sections. The sections were deparaffinized with xylene and rehydrated in graded alcohol solutions before blocking of endogenous peroxidase activity with 3% hydrogen peroxide in phosphate-buffered saline (PBS). Antigen retrieval was performed in citrate buffer, pH 6.2, using a pressure-cooking device, before incubation with PAX5 monoclonal antibody (clone 24, BD Biosciences, Franklin Lakes, NJ) at a dilution of 1:100 or polyclonal anti-PAX2 (Zymed, San Francisco, CA) at 1:200. Visualization of antibody staining was with the DAKO EnVision system (DAKO, Carpinteria, CA) or Super Sensitive Polymer-HRP system (BioGenex, San Ramon, CA), and slides were counterstained with hematoxylin. For PAX5 staining, an internal positive control sample with a section of tonsil was included, and primary antibody was omitted in negative control samples. Only nuclear staining was considered positive, and assessment of positive staining excluded infiltrating or tumor-associated PAX5+ lymphocytes. We used the following scoring system: 0, no reactivity; 1+, rare positive tumor cells; 2+, 1% to 10% positive cells; 3+, 10% to 50% positive cells; and 4+, more than 50% positive cells.

Cell Transfection

To assess antibody specificity, lipofectamine 2000 (Invitrogen, Carlsbad, CA) was used for transient transfection of 293T cells with a mammalian expression vector, pcDNA3.1/Hygro(+) (Invitrogen), containing the full-length Pax5 coding sequence (kindly provided by Andreas Himmelmann, Zurich, Switzerland) or with a pCMV-HA vector containing the Pax2a coding sequence (kind gift of Gregory Dressler, University of Michigan, Ann Arbor). In control samples, empty pcDNA3.1 vector was used. To confirm efficient transfection, an enhanced green fluorescent protein (eGFP)-expressing plasmid was added at a 1:10 ratio of eGFP to Pax, and transfection was confirmed by flow cytometric analysis of an aliquot of cells.

Western Blotting

Equal numbers of transfected 293T cells were harvested and lysed in RIPA buffer (Pierce Biotechnology, Rockford, IL) containing protease inhibitor cocktail (Pierce). Equal volumes of cell lysates were separated on 10% sodium dodecyl sulfate–polyacrylamide gels and transferred onto nitrocellulose membranes (Hybond, Amersham Biosciences Europe, Freiburg, Germany). After blocking for 1 hour in 5% milk powder in tris(hydroxymethyl)aminomethane-buffered saline, the membrane was washed and probed overnight at 4°C with anti-PAX5 (BD Biosciences) at a dilution of 1:500, anti-PAX2 (Zymed) at a dilution of 1:1000, or rat anti–α-tubulin (Serotec) diluted 1:500 in PBS. Following further washing steps, membranes were incubated with horseradish peroxidase–conjugated antimouse (DAKO), antirabbit (GE Healthcare), or antirat IgG (Abcam, Cambridge, England) at 1:2,000 dilution. Bands were visualized using the ECL Plus chemiluminescence kit (Amersham) and exposed onto x-ray film (Sigma, St Louis, MO).

Immunocytochemistry

Transfected 293T cells were harvested by scraping in cold PBS and then fixed for 1 hour in 10% formalin. Formalin-resuspended cells were embedded in 5% agarose in PBS in a 24-well plate in a water bath at 65°C. After cooling at 4°C for 30 minutes to allow the agarose to set, pellets were processed by dehydration and embedding in wax. Subsequent immunostaining was performed as per tissue immunohistochemical analysis.

Quantitative Reverse Transcriptase–Polymerase Chain Reaction

Total RNA was extracted from normal control tissues (liver, skin, and lymph node) and from 8 Wilms tumor specimens (paraffin-embedded archival tissue blocks) using Trizol (Invitrogen). The Wilms tumor specimens included were selected from the samples previously immunostained with the PAX5 monoclonal antibody. Samples with strong, widespread staining were selected for one subgroup, and those with weak or only scattered positive staining were included in a second subgroup. Following elimination of any contaminating DNA using DNase I (Invitrogen), complementary DNA was synthesized using the SuperScript II Reverse Transcriptase Kit (Invitrogen) according to the manufacturer’s instructions. Gene-specific primer and probes for Pax5, Pax2, and 18S RNA were obtained from Applied Biosystems (Foster City, CA), and quantitative polymerase chain reaction (qPCR) was performed in triplicate or quadruplicate on an ABI 7000 Real Time PCR system (AME Bioscience, Toroed, Norway). Expression of Pax5 and Pax2 in different tissues was normalized to 18S, and quantification was relative to expression in liver. Relative quantitation was determined by the 2−δδCT method. RNA from paraffin-embedded archival material may be of poor quality, and samples in which the cycle number to achieve threshold value for the 18S housekeeper gene was more than 25 were excluded. The final quantitative analysis, therefore, included the 3 normal tissue control samples and 4 Wilms tumor specimens.

Results

PAX Immunostaining in Pediatric Tumors

We examined PAX5 expression in a variety of pediatric extracranial tumors by immunohistochemical analysis with the anti-PAX5 antibody most commonly cited in the literature: a monoclonal antibody against PAX5 (amino acids 151–306), clone 24 supplied by BD Biosciences. No expression, in the presence of appropriate positive control samples, was seen in the majority of tumor types examined (data not shown), including osteosarcoma (n = 99), peripheral neuroectodermal tumor (n = 17), undifferentiated sarcoma (n = 6), neuroblastoma (n = 24), malignant rhabdoid tumor (n = 29), or cystic mesoblastic nephroma (n = 13). As previously reported, embryonal rhabdomyosarcoma (n = 9) was also negative for PAX5 expression, whereas alveolar rhabdomyosarcoma (n = 15) showed positive staining in 9 cases (60%).9 Most striking was the positive immunostaining seen in Wilms tumor, with 30 (83%) of 36 tumors showing positive staining Image 1. Nuclear immunoreactivity was seen most strongly within the epithelial component, with variable expression in blastema and no staining in stroma. The degree of staining varied among samples depending on the relative contribution of epithelial, blastemal, and stromal elements (score 0, 6 samples; 1+, 6 samples; 2+, 8 samples; 3+, 6 samples; and 4+, 10 samples). Next, we examined fetal kidney and surgical resections of cystic dysplastic kidney and showed strong nuclear staining with the PAX5 antibody within developing epithelial structures in fetal kidney and in the epithelial layer surrounding cystic structures in dysplastic kidney (Image 1).

It is important to note that the pattern of anti-PAX5 antibody immunoreactivity in Wilms tumor, fetal kidney, and cystic dysplastic kidney appeared strikingly similar to the patterns of expression reported for PAX2,1012 raising the possibility that the antibody might cross-react with other members of the Pax gene family, particularly PAX2. We performed PAX5 and PAX2 immunostaining in serial sections and demonstrated very similar patterns of expression (Image 1); both antibodies stained condensing epithelial structures in fetal kidney and the epithelial component in Wilms tumor. In cystic dysplastic kidney, although both antibodies appeared to stain the cystic epithelium, this was stronger with PAX2 compared with PAX5 antibody. The PAX5 antibody, in contrast with anti-PAX2, clearly labels B-lineage lymphocytes in lymph node (Image 1), but the staining pattern similar to that of PAX2 in other structures raised the strong possibility of cross-reactivity. We were surprised that no previously reported study using the clone 24 PAX5 monoclonal antibody had excluded such a possibility, and the manufacturer was unable to provide data in this regard.

Confirmation of Antibody Specificity Using Transfected Cells

To confirm or refute the specificity of the clone 24 PAX5 monoclonal antibody, we transfected 293T cells (which do not endogenously express PAX5 or PAX2) with mammalian expression vectors encoding human Pax5 and Pax2. Western blotting of cell lysates clearly demonstrated recognition of the PAX5 protein by the clone 24 anti-PAX5 antibody, but it is important to note that this antibody also recognized a similarly sized band in cells transfected with Pax2 Image 2. Similarly, although the anti-PAX2 antibody recognized PAX2 protein following transfection, there was also cross-reactivity with transfected PAX5. These results were confirmed by immunohistochemical analysis using cell pellets that were fixed in formalin, embedded in agarose gel, and processed in a manner identical to that used for the tumor tissue samples in the original analysis. Although cells transfected with empty vector showed no positive staining with either antibody, cells transfected with either the Pax5 or Pax2 plasmids were both recognized by the PAX5 antibody and by the PAX2 antibody Image 3. Therefore, the clone 24 PAX5 monoclonal antibody, while recognizing PAX5 itself, also cross-reacts with PAX2, and positive immunostaining with this antibody cannot be taken as confirmation of PAX5 expression.

PAX2 Expression, but Not PAX5 Expression, in Wilms Tumor by Quantitative Reverse Transcriptase–PCR

The finding of cross-reactivity of the PAX5 antibody with PAX2 raised the strong possibility that the positive immunostaining seen in Wilms tumor samples actually represented PAX2 expression and not that of PAX5. To confirm this, we examined transcription of both Pax genes in Wilms tumor specimens and several normal tissue samples by quantitative reverse transcriptase (qRT)-PCR. All 4 Wilms tumor specimens for which there were good quality qPCR data showed low levels of Pax5 expression compared with lymph node positive control samples Table 1. In contrast, Pax2 expression was seen in all 4 tumors and correlated with the degree of immunoreactivity using the anti-PAX5 monoclonal antibody. There was no correlation between immunostaining with the PAX5 antibody and Pax5 messenger RNA (mRNA) expression by qRT-PCR. Thus, it can be concluded that PAX5 is not expressed to a significant degree in Wilms tumor and the positive immunostaining seen with clone 24 monoclonal antibody is entirely a result of cross-reactivity with other PAX proteins, most probably PAX2.

Image 1

Immunostaining of adjacent 4-μm sections of human fetal kidney (approximately 20 weeks’ gestation), Wilms tumor, cystic dysplastic kidney, and lymph node with anti-PAX5 (A, C, E, and G) and anti-PAX2 (B, D, F, and H) antibodies. A and B, In the developing fetal kidney, immunostaining with both antibodies is seen in condensing mesenchyme and developing tubular epithelial structures, whereas more mature glomeruli are negative. C and D, In the Wilms tumor specimen in which epithelial structures predominate, there is widespread, identically patterned staining with both antibodies. Positive staining of structures in Wilms tumor and developing kidney with the PAX5 antibody likely represents cross-reactivity with PAX2 rather than true expression (see the text). E and F, In cystic dysplastic kidney, there is strong PAX2 staining of dysplastic epithelium as previously reported. Although these structures are also stained with the PAX5 antibody (arrows), this is weaker and likely represents cross-reactivity of the antibody with PAX2. In contrast, the PAX5 antibody strongly stains B-lineage lymphocytes within the cystic dysplastic kidney specimen (arrowheads) and within the lymph node–positive control sample (G), H shows negative staining of lymph node by anti-PAX2 (AH, ×100).

Discussion

The data presented herein lead to 2 important conclusions. First, the specificity of the most frequently cited anti-PAX5 antibody (clone 24) must now be cast into considerable doubt, requiring a reappraisal of published literature using this antibody. Second, although positive immunostaining with this antibody cannot be used to prove PAX5 expression (owing to cross-reactivity), the absence of nuclear staining (in the presence of appropriate positive control samples) does identify tumors (or tissues) in which there is not significant expression of PAX5 protein. Thus, in view of recent interest in PAX5 as a therapeutic target,8 these data are important for demonstrating that PAX5 is probably not expressed significantly in pediatric nonhematologic tumors, with the possible exception of alveolar rhabdomyosarcoma,9,13 and indicating that other studies identifying PAX5 expression in nonhematologic tumors and tissues may need to be reappraised.

Image 2

Western blot analysis of cell lysates. 293T cells were transiently transfected with mammalian expression vectors expressing PAX5 or PAX2 (2 separate replicates shown) or empty vector. Blots were reprobed with anti–α-tubulin as a control for equal loading. Although the PAX5 antibody recognizes an appropriately sized band (expected size, 52 kDa) in the lysate from PAX5-transfected cells, the antibody also recognizes a similarly sized band in the lysates from PAX2-transfected cells, ie, the antibody cross-reacts with PAX2 (expected size, 46 kDa). Labeling with the PAX2 antibody similarly reveals cross-reactivity of this polyclonal antibody with PAX5.

PAX5 controls the development of B lymphocytes,14 and its pattern of expression during hematopoiesis and in hematologic malignancies is well established. PAX5–/– mice have complete absence of functional B cells and abnormal midbrain development (correlating with expression of PAX5 in the developing central nervous system) but are otherwise phenotypically unremarkable,15 suggesting a limited role for PAX5 in the development of other tissues. Expression of PAX5 in normal adult tissues and in nonhematologic malignancies has been investigated in numerous studies during the last decade. The majority have used immunohistochemical analysis using anti-PAX5 antibody (clone 24), the specificity of which must now be cast into doubt as a result of the data presented herein. This monoclonal antibody was raised against amino acids 151 to 306 of the human PAX5 protein (BD Biosciences data sheet), although the exact epitope recognized is not known. The immunogen does not include the highly conserved N-terminal DNA-binding domain Figure 1. Nevertheless, the PAX gene subfamilies (eg, PAX2, PAX5, and PAX8) share considerable sequence homology outside this region, and antibody cross-reactivity with other members of the PAX gene family (particularly PAX2 and PAX8) is a strong possibility on the basis of the sequence comparisons alone. The data presented herein demonstrate that this cross-reactivity indeed occurs. Also of note, although the PAX2 antibody used in this study has been well characterized as recognizing PAX2,12,16 with epitopes between amino acids 270 and 338, cross-reactivity with PAX5 (as seen in the data presented herein) has also been previously reported.17

Image 3

Immunostaining of fixed and agarose-embedded 293T cells transiently transfected with empty vector or PAX5 or PAX2 expression vectors. Control cells (empty vector) showed no staining with PAX5 (A) or PAX2 (B) antibody. The PAX5 antibody stained a proportion of cells following transfection with PAX5 plasmid (C), confirming that the antibody recognizes PAX5. However, there was also positive staining of cells transfected with PAX2 plasmid (D), indicating cross-reactivity of the antibody with PAX2. Similar cross-reactivity was seen with the PAX2 polyclonal antibody (E and F) (AF, ×200).

Torlakovic et al18 examined expression in 123 normal tissue samples and 51 neuroendocrine tumors by immunohistochemical analysis using monoclonal anti-PAX5 antibody (clone 24). Western blotting was used to confirm that the antibody recognized PAX5 expressed in the pre-B lymphoblastic cell line, Daudi, but the possibility of cross-reactivity with other members of the Pax gene family was not explicitly addressed. Limited positive staining was seen in adult tissues, with the exception of scattered neurons within the medulla oblongata and periaqueductal gray matter, and “strong expression” in epididymis and in mesonephric rests and hyperplasia. PAX2 expression in (mouse) epididymis has been confirmed at the protein and RNA levels,19 raising the strong possibility of PAX2/PAX5 immunologic cross-reactivity in epididymis. In support of this claim, PAX2 has been shown to have a crucial role in urogenital development,20,21 with PAX2–/– mice lacking kidneys, ureters, and genital tracts (all mesoderm-derived), while having preserved urethra, bladder, and prostate, all endoderm-derived structures.22 In contrast, PAX5 expression within the developing urogenital tract has not been reported, and PAX5–/– mice appear to have no urogenital abnormalities.15 Therefore the apparent PAX5 expression in epididymis and mesonephric rests and hyperplasia reported18 probably represent cross-reactivity with PAX2.

View this table:
Table 1
Figure 1

Human PAX5/PAX2 protein sequence comparison. The region in red is the immunogen for the Pax5 monoclonal antibody (clone 24). The underlined region is the conserved DNA binding domain. Compared using the FASTA format. Human protein sequences from http://www.uniprot.org.

Mhawech-Fauceglia et al23 undertook a high-throughput immunohistochemical screen of more than 3,700 benign and malignant nonhematologic tumors using the clone 24 anti-PAX5 monoclonal antibody. They demonstrated positive staining in Merkel cell carcinoma (a neuroendocrine malignancy of the skin) and in small cell lung carcinoma and suggested that the antibody was a useful marker for neuroendocrine tumors. The expression of PAX5 in Merkel cell carcinoma and small cell carcinoma has been independently confirmed by immunoreactivity (again using the clone 24 antibody) and, it is important to note, by RT-PCR to correlate immunostaining with the presence of PAX5 mRNA.24 Recent reports have confirmed PAX5 as a marker of high-grade neuroendocrine lung carcinomas.25,26 Although PAX5 expression in a number of specific neuroendocrine tumors has been established, it is much less clear the degree to which PAX5 expression can be taken as a general marker for neural crest–derived cells with neuronal or neuroendocrine differentiation (including, eg, neuroblastoma) as has been proposed.26,27 Although PAX5 expression, at least at the mRNA level, has been demonstrated in neuroblastoma cell lines,27 there is no evidence for expression in patient-derived neuroblastoma samples. Indeed, the data presented herein and those published in abstract form by Myers et al28 explicitly demonstrated no immunostaining with monoclonal antibody clone 24 in a total of 46 neuroblastoma specimens, indicating no significant expression of PAX5 (or, indeed, PAX2).

The possible role of PAX5 in other nonhematologic malignancies also remains unclear. PAX5 expression at the mRNA level has been demonstrated in bladder transitional cell carcinoma, but not in normal urothelium, and expression correlates with tumor grade.29 Increased levels of Pax5 mRNA expression are associated with a poorer clinical outcome.30 Nevertheless, both of these studies examined only mRNA expression of Pax5. A recently published preliminary study investigated PAX5 protein expression in bladder cancer and showed no correlation with grade, stage, or clinical outcome.31 Although strong immunostaining (again with the clone 24 monoclonal antibody) was seen in infiltrating lymphocytes, staining of urothelial tumor cells was seen in only 10% of cases and, even in these cases, was generally reported as “weak.” Given the data presented herein, it is possible that this immunostaining is due to cross-reactivity with PAX2 and does not necessarily represent true PAX5 protein expression. In the bladder cancer cell line EJ, for example, high levels of Pax2 (plus Pax6 and Pax8) mRNA expression have been detected, whereas Pax5 was not detectable.32 Jensen et al6 similarly observed weak PAX5 immunostaining in only 2 of 339 bladder carcinomas. This discrepancy between Pax5 expression at the mRNA and protein levels also arises in central nervous system malignancies. Although Pax5 mRNA expression has been demonstrated in astrocytoma33 and medulloblastoma,34 it is unclear whether this corresponds to PAX5 protein expression in these tumors. A preliminary analysis including 2 medulloblastomas and 7 astrocytomas showed no PAX5 immunostaining.6

Therefore, although positive immunostaining with the clone 24 antibody may indicate PAX5 expression, the possibility of cross-reactivity with other closely associated members of the Pax gene family (particularly PAX2 and PAX8) needs to be carefully considered. Although expression of PAX5 in B cell–derived hematologic malignancies is well established, immunohistochemical screening of other tissues and tumor types may produce false-positive results. Although PAX5 appears to be expressed in 2 specific types of neuroendocrine tumors (Merkel cell carcinoma and small cell lung carcinoma), it is probably not correct to envisage PAX5 as a general marker of neural crest–derived cells with neuronal or neuroendocrine differentiation—specifically pediatric neuroblastoma appears not to express PAX5. Although expression of PAX5 in normal epididymis, mesonephric tissues, and Wilms tumor has been reported (see, eg, Sica et al35), these data must come into doubt in the light of the results presented herein. In particular, it is unlikely that PAX5 is expressed to any significant extent in Wilms tumor; rather the positive immunostaining with the clone 24 antibody represents PAX2 expression. More generally, these data emphasize the crucial importance of ensuring antibody specificity. For all antibodies, particularly those directed against individual members of closely related families of proteins, the possibility of cross-reactivity must be considered and explicitly addressed.

Acknowledgments

We thank Vanessa Walf-Vorderwülbecke for helpful advice on qPCR.

Notes

Upon completion of this activity you will be able to:

  • discuss the use of the clone 24 anti-PAX5 monoclonal antibody.

  • list the tumor types in which PAX5 expression has been confirmed.

  • describe the major roles of PAX5 in development.

The ASCP is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The ASCP designates this educational activity for a maximum of 1 AMA PRA Category 1 Credit™ per article. This activity qualifies as an American Board of Pathology Maintenance of Certification Part II Self-Assessment Module.

The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose.

Questions appear on p 511. Exam is located at www.ascp.org/ajcpcme.

Footnotes

  • Supported by the Institute of Child Health Biomedical Research Centre, London; and the National Institute for Health Research (Dr Morgenstern), London.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
View Abstract