Construction of tissue microarrays from core needle biopsies – a systematic literature review

In some clinical circumstances, core needle biopsy (CNB) may be the only source of material from cancer tissue for diagnostic use. The volume of tissue available in a CNB is low, and opportunities for research use can therefore be limited. The tissue microarray (TMA) principle, if applied to the use of CNBs, could facilitate research studies in circumstances where CNB specimens are available. However, various challenges are expected in applying such a technique in CNBs, which has limited their use in research. We therefore conducted a systematic review of the literature on this subject. A systematic search was carried out with CINAHL, EMBASE, the Cochrane library, and MEDLINE, to identify studies that have primarily developed methods for constructing TMAs from CNBs. Eight studies were found to meet the inclusion criteria; six of these employed the vertical rearrangement technique, and two used multiple layers of biopsy tissue. Representation of the CNB was significantly influenced by the quantity of tumour cells present in the original biopsy and the degree of heterogeneity of biomarker expression. This review shows that technologies have been developed to enable construction of TMAs from CNBs. However, challenges remain to improve amplification and representation.


Introduction
In some clinical circumstances, surgery may not be the preferred form of initial or primary therapy for breast cancer, and a core needle biopsy (CNB) will be the only available material from cancer tissue for diagnostic and research purposes. The volume of tissue and tumour in CNBs is far lower than that in tumour samples obtained from surgically resected specimens, which significantly reduces the opportunities for tissue-based biological research studies.
Methods to support biological studies in these clinical circumstances are clearly needed, e.g. to identify and validate the role of improved predictive biomarkers of therapy efficacy. Non-operative therapy, typically systemic therapy, is used in the metastatic setting, as neoadjuvant therapy to downstage the disease prior to surgery, and as an alternative therapy to surgery, particularly in patients with comorbidities. For instance,~40% of older women (>70 years) with early-stage breast cancer receive non-operative therapy, in the form of primary endocrine therapy, [1][2][3] because they refuse or are unfit/frail, owing to concomitant comorbidities.
Numerous published studies have indicated high levels of concordance in biological characteristics between CNBs and subsequent surgical excision or excision biopsies. [4][5][6][7][8] A meta-analysis carried out by Li et al. indicated high correlations for such key biomarkers of breast cancer biology, including oestrogen receptor, progesterone receptor, and HER2. 6 Similar results were obtained in the estimation of Gleason score in prostate cancer patients. 9,10 Given such evidence for high concordance between CNB and surgical specimens, analysis of detailed biomarker expression in tissue obtained by CNB would have great research potential. Although CNB provides enough tissue material for initial diagnosis, detailed profiling of the tumour requires a larger volume of tissue, which usually necessitates the use of tissue derived from surgical specimens. Therefore, if CNB is to be used for achieving this purpose, a technique that can maximize the utility of available tissue is required. Tissue microarray (TMA) techniques have been developed for the purpose of high-throughput immunohistochemical profiling of large cohorts of cases with limited amounts of tumour tissue. 11 In addition, they allow the analysis and evaluation of tissue-based assays in an efficient, cost-effective and uniform manner. 12,13 The length and thickness of the tissues obtained by CNB make the conventional TMA method developed by Kononen et al. 11 inappropriate for achieving the same objective with CNB. The small diameter of the biopsy material makes horizontal embedding less effective in amplifying the number of possible sections. Moreover, the biopsy diameter is often eroded after sectioning for initial diagnosis, so only a few sections are obtainable.
The current use of CNB is usually restricted to diagnostic purposes. Therefore, optimizing a method for constructing TMAs from CNBs presents an opportunity for researchers to profile tumours by using a scarce amount of tissue. On the other hand, given the potential technical challenges in such construction, it would be necessary to review the literature to identify and critically appraise different methods employed for that purpose.
This study aimed to systematically review the literature for studies that have explored the technology of constructing TMAs from paraffin-embedded CNBs, and to critically analyse their results.

Methods and search strategy
The literature review was carried out and presented according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guideline. 14 Electronic searches were used to review the literature for related studies. The studies were screened for their titles. The abstracts were then evaluated to determine the studies that would fit the selection criteria, and, finally, full articles of the eligible studies were reviewed.
A comprehensive search was carried out with CINAHL, EMBASE, the Cochrane library, and MED-LINE. The references of the studies were also screened for any other relevant studies.
The keywords used for the electronic search were identified by the use of Medical Subject Headings (MeSH), including; 'tissue microarrays AND core needle biopsy', 'construction AND tissue microarrays AND core needle biopsy', 'tissue microarrays AND needle biopsy', and 'tissue micro arrays AND core needle biopsy'.

I N C L U S I O N C R I T E R I A
Primary studies that developed methods for constructing TMAs from paraffin-embedded CNBs were included. The search was restricted to work published in the English language between 1998 and January 2015. E X C L U S I O N C R I T E R I A Studies utilizing a conventional approach for the construction of TMA blocks, as described by Kononen et al., 11 were excluded.

T Y P E O F T I S S U E
Only CNBs obtained from human tissues were considered.

Primary outcome
The primary objective was to evaluate methods for the construction of TMAs from CNBs.

Secondary outcome
The review also: (i) evaluated the representativeness (homogeneity, antigenicity, and morphology) of tissue obtained from TMA blocks based on CNBs; (ii) investigated the number of obtainable sections; and (iii) investigated the density of cores per block.

Results
The search databases and the number of studies associated with identified keywords are summarized in Table 1. Figure 1 shows a flow chart illustrating different phases of the systematic review. Finally, a summary of the outcome measures is given in Table 2. Eight studies were found to meet the inclusion criteria ( Table 2). Jhavar et al. 15 and Datta et al. 16 were the earliest to have published methods for construction of TMAs from CNBs. Their methods used the most representative area of the biopsy, and rearranged it vertically in the recipient TMA block. The

M E T H O D F O R C O N S T R U C T I N G T M A S F R O M C N B S
The methods developed for constructing TMAs from CNBs consist of two steps: extracting the biopsy from the donor block, and integrating the extracted biopsy within the recipient TMA block ( Figure 2). However, for the purpose of ensuring accurate positioning of biopsy segments within the recipient block and increasing the number of cores included in the TMA block, some techniques included an intermediate step that involved vertical embedding of biopsies into an intermediate block.
Jhavar et al. 15 started by cutting paraffin blocks to obtain a segment of cubic wax containing the biopsy, which was arranged in a steel mould holding the biopsy in a vertical position. McCarthy et al. 17 modified the method of Jhavar et al. 15 in two respects; first, by using predesigned knives for extraction of the biopsies of consistent length from the donor block; and second, by utilizing a rubber template to design the recipient TMA block. These modifications allowed the construction of a TMA block with an extra capacity for cores, and allowed easy extraction of tissue from the donor blocks. A further modification of the method used by Jhavar et al. 15  Tissue micro array AND core needle biopsy 0 the method of Jhavar et al. 15 used originally for constructing the TMA block. Datta et al. 16 used unfragmented biopsy segments. After extraction of the biopsy from the donor block with a scalpel, it was embedded in a predesigned template in order to reform wax around the biopsy in a cylindrical shape. Biopsies were then transferred into the TMA blocks, which were designed with a manual arrayer. As an alternative to extracting the biopsy from the donor block by scalpel, Vogel and Bultmann 19 used a skin biopsy punch to obtain tissues enriched with tumour cells from the donor block. Their method included melting wax around extracted tissue on a hot plate at 65°C prior to transferring it to the TMA block. Studies utilizing vertical rearrangement techniques were found to integrate tissue cores within the TMA block either by annealing the TMA or by melting the TMA block and subsequently cooling it.
Fridman et al. 20 utilized a method in which case selection was limited by the presence of a minimum of 30% of tumour cells in the biopsy. Their method included melting the original wax block, and then staining the tissue cores with eosin for better visualization of the extracted tissue segment. Following this, each biopsy was fragmented into equal parts and re-embedded vertically in an intermediate block. A manual arrayer was then used to transfer cores from the intermediate block to the TMA block.
Kishen et al. 21 applied a multiple-layer approach for making a biopsy TMA block that included embedding two segments of a biopsy (2 mm each) on top of each other. The tissue used was microscopically confirmed to be rich in tumour cells prior to the TMA block being made.
Komiya et al. 22 started by cutting a thick section (30 lm) from the donor block; then, segments rich in tumour with a size of 3 mm were dissected from the section and arranged in line on a paraffin sheet of thickness 100 lm. Both tumour segments and the paraffin sheet were rolled up into cylindrical reels.  The cylindrical reels were then divided into sections and transferred into a predesigned recipient block.

R E P R E S E N T A T I O N O F T I S S U E O B T A I N E D F R O M T H E B I O P S Y T M A B L O C K
Here, representation refers to three aspects, namely, homogeneity between cores obtained from different biopsies of each patient, preservation of antigenicity, and morphology of the original tissue. Jhavar et al. 15 found that antigenicity was fully preserved in tissue obtained from prostate CNBs; their investigation included staining tissue obtained from 123 individuals known to have prostate cancer. Antigenicity was investigated by evaluating three markers used for routine diagnosis, namely, low molecular weight keratin (CAM5.2), high molecular weight keratin (LP34), and prostate-specific antigen (PSA). Similar findings were reported by Fridman et al., 20 who stained the produced sections with high molecular weight cytokeratin, PIN cocktail (p63 + p504S), and PSA. Their work demonstrated that the morphological and immunohistochemical characteristics of the original tumour CNB were maintained in the TMA format.
Datta et al. 16 used three cores rich in prostatic carcinoma cells from each individual CNB, in order to provide a better estimation of protein expression. Seventeen sets of CNBs and the corresponding radical prostatectomy specimens were stained for Ki67 antigen to investigate the concordance between the two types of specimen, and an estimated correlation coefficient (r) of 0.4994 (P = 0.04) was found.
A tumour detection rate of 66-79% was achieved from TMA blocks constructed from CNBs obtained from two series of prostate cancer patients (n = 303), with the method described by McCarthy et al. 17 Vogel and Bultmann 19 reported high concordance between sections obtained from biopsy TMAs, excision biopsies, and mastectomy specimens, in terms of HER2 status in breast cancer. The investigations included immunohistochemistry (IHC), fluorescence in-situ hybridization (FISH), and automated brightfield double in-situ hybridization (BDISH). The IHC results showed minor variations between different tissue types, whereas FISH and BDISH produced identical estimates of ErbB-2 gene amplification.
Komiya et al. 22 demonstrated the representativeness of CNBs in TMAs by showing the prognostic value of Ki67, p53 and bcl-2 in prostate cancer, based on the correlation between the IHC results of these biomarkers and survival data of 58 patients.

N U M B E R O F C O R E S A N D O B T A I N A B L E S E C T I O N S
The capacity of a TMA block to include biopsy samples depends on the diameter of the cores and the distance between the cores. For instance, Jhavar et al., 15 who included the integration of biopsy and surrounding wax with the TMA block, produced a smaller number of cores that could be inoculated into the final TMA block. On the other hand, Vogel and Bultmann 19 managed to include a large number of samples by removing the paraffin surrounding the biopsy segments. In addition, methods such as that of McCarthy et al., 18 who used a manual arrayer, were found to create a TMA block with higher capacity. This review indicates that the number of TMA cores ranged from 20 to 187 per TMA block. The methods used for constructing TMAs from CNBs were found to have a standardized CNB length of 3-4 mm. The number of obtainable representative sections was influenced directly by the number of tumour cells present in the original biopsy. The thickness of sections used for IHC staining ranged between 3 lm and 5 lm. [15][16][17][18][19][20][21]

Discussion
The length and thickness of donor CNBs were found to influence successful tissue extraction from the donor block and the integration of the cores within a recipient TMA block. The conventional method for constructing TMAs developed by Kononen et al. 11 has been limited by the need for relatively large amounts of tissue. The literature review indicates two approaches for the use of CNBs in TMA construction, namely, vertical rearrangement, and a multiple-layer approach for 'amplification' of tissue.
Embedding a fragmented biopsy with surrounding wax would preserve the original structure of the biopsy. In addition, transferring biopsies with their surrounding wax would make it easier and safer to maintain tissue structure than handling the biopsy itself. Moreover, an advantage of this method is that there would be better integration with the TMA block, as wax surrounding the biopsy would easily anneal to the TMA block. Nevertheless, obtaining a biopsy from the donor block with surrounding paraffin would result in a lower number of cores being  Figure 2. Approaches for constructing tissue microarrays from core needle biopsies.
included in a final recipient block, owing to the wider area required. However, the use of a predesigned tool to extract the biopsy in order to reduce the amount of paraffin surrounding the biopsy might allow the TMA block to include more cores. A method for supporting a higher-density TMA block includes the use of a manual or automated arrayer for taking cores from the CNB. However, there is a limitation of this practice with non-straight-embedded cores or a biopsy that deviates slightly from the vertical position; punching biopsies from donor blocks with such defects can result in less tissue being transferred into the recipient block.
Extraction procedures involving the melting of the whole donor block help to achieve a straighter biopsy, although the accuracy in reorienting the original embedded biopsy might be lost, leading to the possible misplacement of the preselected area rich in tumour cells. Instead of melting the whole block, extracting and melting the segment rich in tumour cells would avoid destroying the original material, and would help to re-form and integrate the biopsy within the TMA block.
Biopsies can sometimes contain a small amount of tumour (<4 mm) that is not sufficient to be embedded vertically on the TMA block. Alternatively, in such circumstances, multiple layers can be extracted from CNB biopsies and embedded one over the other. For this method, TMA cores with the required length are therefore needed to produce a reasonable number of sections. However, stratified cores might contain intralayer gaps, resulting in sections without representative tissue. Embedding multiple layers in parallel or in-line on the surface of the recipient blocks would not help to amplify material from these blocks, as the depth of tissue on sections will be similar to the thickness of original biopsies.
Integration of the biopsy with its surrounding wax in the recipient block is a critical step in constructing the CNB TMA, as it is associated directly with biopsy orientation and the quality of the section. The articles reviewed herein applied two methods for integrating biopsies with surrounding wax, namely: (i) softening the recipient block; and/or (ii) melting the whole recipient block and re-solidifying it. Although the softening method clearly helps to even the surface of the recipient block, this method will not ensure that cores transferred into the recipient block are completely annealed with the surrounding wax. One of the advantages of melting the donor block as compared with the softening method is obtaining tissue that is perfectly integrated with the surrounding wax. However, the orientation of the tissues within the recipient block is still an issue, and further work should focus on preserving tissue during the integration step.
It is important for sections obtained from biopsy TMAs to represent the tumour characteristics of the whole tumour. Issues with the standard method of Kononen et al., 11 such as intra-tumour heterogeneity, might affect the assessment of tumour biology. For instance, expression of the proliferation marker Ki67 has shown considerable heterogeneity between different tumour areas. 23,24 Researchers developing TMAs have considered two technical strategies to avoid misrepresentation of the original tumour resulting from small TMA cores, namely, ensuring a proper core diameter, and increasing the number of cores from each tumour case.
The diameter of tissue taken by CNB usually ranges between 1.2 mm and 2.5 mm. Moreover, most studies that have constructed biopsy TMA blocks have used complete cross-sections of the biopsy. A biopsy that has been extensively sectioned for initial tumour diagnosis might have smaller tissue spots. Singh et al. 25 suggested that the lower number of tumour cells available in TMA sections leads to reduced representativeness of the section and adversely affects the reliability of biopsy TMAs. Intra-tumour variations in biomarker expression among different tumour cells could result in reduced concordance between tissues from CNB and surgical specimens in terms of protein expression. For conventional TMAs, the sampling strategy proves that three cores or more would properly represent the whole tumour tissue and reduce the error associated with tumour heterogeneity. [26][27][28] Datta et al. 16 investigated the optimal number of cores from each individual biopsy needed for better representation of the whole tumour in patients with prostate cancer, and demonstrated that tumour biology could be accurately estimated with three core biopsies from each patient. However, the number of cores per block depends on the degree of biomarker expression heterogeneity. It is also important to mention that there are situations in which CNBs should not be subjected to harvesting for TMAs. Small amounts of invasive tumour tissue within the core or extensive sectioning of CNBs for initial tumour diagnosis, with the potential for tumour depletion, can be considered as contraindications to the use of TMAs. This is to ensure preservation of diagnostic core material for subsequent review whenever needed, particularly in relatively recent cases. Another contraindication to CNB TMA is the lack of representation of the index invasive tumour, such as for CNBs from invasive carcinoma that contain pure ductal carcinoma in situ (DCIS), DCIS associated with microinva-sion, or tiny amounts of invasive tumour tissue that are unlikely to be represented in TMA cores.

Conclusion
Constructing TMAs from CNBs requires modification of the conventional methodology. Further developments are required to improve amplification of tissue from the biopsy. Owing to the issue of tumour heterogeneity, as identified in the studies reviewed, further work should also focus on enhancing the representativeness of tissues obtained from biopsy TMAs.