The main cause of death in breast cancer patients is distant metastasis.1 It has been reported that 16–20% of newly diagnosed breast cancer patients have advanced disease associated with distant metastasis.2 Axillary lymph nodes (LNs) are commonly the first line of metastasis of breast tumour cells.1 3 In operable breast cancer, the three strongest prognostic determinants used in routine clinical practice are LN stage, primary tumour size and histological tumour grade. Recently, sentinel LN (SLN) dissection has become a treatment standard in the management of early-stage breast cancer, thereby illustrating the importance of accurate assessment of LN status in breast cancer staging.

There is, however, currently no internationally accepted standardised protocol for the histological examination of LNs and false-negative diagnoses can occur. The National Health Service Breast Screening Programme publishes guidelines for pathological reporting in breast cancer screening and acknowledges that standardisation of LN processing does not exist between, and perhaps within, UK laboratories.4 Current accepted laboratory practices vary, but generally include slicing nodes and reviewing single sections from each cut surface. This results in less than 1% of the total area of each node being assessed.5 Saphir and Amromin (1948) first suggested that evaluation of single H&E-stained sections of LNs was inadequate for consistent detection of axillary metastases.6 Their work initiated subsequent studies aimed at improving detection rates of metastases in LNs. A number of new techniques have been developed to aid the frequency of detection of small areas of metastases within LNs; these techniques include multiple levels/serial sectioning, immunohistochemistry (IHC), molecular methods, and even image analysis systems.7 Despite numerous reports advocating such techniques as means of improving detection rates,8^–11 their inherent labour intensive and costly nature impacts on feasibility of implementation into the routine clinical setting.

The advent of SLN technology offers opportunity for detailed analyses to be focussed upon this node, since it is the first to receive lymphatic drainage and consequently the most likely to develop metastases. A number of studies have since been carried out attempting to develop an optimum method of histological examination of SLNs to allow detection of small metastases in LNs.9 12 13 None of the studies has provided detailed information about the spatial distribution of metastatic deposits within the SLN however and, as yet, they do not appear to have been widely adopted into routine clinical practice. An alternative strategy might be to focus detailed laboratory analyses upon those areas of the SLN most likely to contain metastatic deposits. Some authors have suggested that such areas include the region of the inflow junction of the afferent lymphatic vessel.14 15

Clearly an incentive has been provided for a single standardised protocol for the histological examination of SLNs to be devised, implemented and recommended for standard clinical and laboratory practice. In order for this to be achieved, the most accurate and yet cost-effective means of analysing SLNs needs to be identified. A better understanding of the metastatic process and spatial arrangement of tumour deposits within SLNs is of paramount importance.

Three-dimensional (3D) reconstruction (3DR) is not a new concept; however the complexities of visualisation methods and data collection suitable for reconstruction of histological samples have been elusive. Confocal laser scanning microscopy became popular for a variety of 3D tissue reconstructions16 17; however, the processing and storage of data is expensive, and the equipment is not routinely available.18 19 Recently, serial section re-alignment techniques utilising digital images captured by flatbed scanners20 21 and digital cameras22 23 have been increasingly reported as they are financially more accessible. A number of studies have been performed on 3DRs of breast tissues20 23^–25; however, there appears to be a paucity of literature on 3D studies of LNs. Serial section re-alignment of breast cancer metastases in LNs was first described by Maddison et al in 2005.26 However, that study used partial nodes only, and the sophisticated camera scanning system employed to create the digital images for re-alignment was expensive and therefore not readily available.

This study aims to determine whether 3D voxel reconstructions of metastases within entire axillary SLNs from breast cancer patients could be created using more economical equipment such as a high-resolution flatbed scanner and to utilise these 3DR techniques to evaluate the early metastatic disease process in SLNs in an attempt to understand the nature of this process in breast cancer.

METHODS

SLNs were collected prospectively from 111 patients undergoing routine primary breast cancer surgery at Nottingham University Hospital (NUH). After routine histopathological reporting, the cases reported as containing macrometastases were recruited into the 3DR study. Those cases reported either as negative (no metastatic tumour cells present), or as containing micrometastases only, were deemed unsuitable for 3DR. A total of 19 SLNs were recruited into the study.

Routine SLN technology at NUH utilises peritumoral injection of ^99MTechnetium radio colloid (in the form of Nanocoll) and the dye patent blue V; together, these enable the SLN to be easily localised. This enabled the afferent lymphatic of the SLN to be easily identified at surgery and tagged with a suture upon removal. Subsequent laboratory handling and processing comprised formalin fixation for 24 h, then removal of the suture and inking of the afferent lymphatic pole. Each node was then sliced at 3 mm intervals and each slice processed in a separate cassette, with the cut surface placed downwards, and the inked pole toward the top of the cassette. This ensured that each slice could be accurately identified in relation to the original entire node. The SLN slices were then histologically processed and embedded in paraffin wax, ensuring that the orientation of each slice was maintained.

Each SLN slice was serially sectioned at 4 μm onto charged slides (X-tra Adhesive micro slides; Surgipath Europe, Peterborough, UK). Continued orientation was ensured by mounting each section with the afferent pole at the labelled end of the slides. Each node yielded, on average, approximately 100 sections for 3DR. All sections were dried overnight in a 37°C incubator. Sections were selected at 20 μm intervals and stained immunohistochemically in order to visualise the metastatic tumour cells within the SLN tissue. A standard avidin–biotin complex IHC technique was employed as previously described.23 The primary antibody used was the broad-spectrum cytokeratin mouse monoclonal antibody AE1/AE3 (Dako, Glostrup, Denmark) optimally diluted at 1/100, and sections were lightly counterstained with Mayer’s haematoxylin for 1 min.

All immunohistochemically stained sections were initially visualised using brightfield microscopy, where the location(s) of the metastatic tumour cells in relation to the afferent lymphatic pole of the node, and the pattern(s) by which they were growing within the node, were recorded. Sections were then digitally scanned at 1200 dpi using an Epson Perfection 4990 Photo flatbed scanner. Pilot experiments by our group confirmed that 1200 dpi resolution was optimal for 3DR purposes of areas of metastases within LNs. The dataset was processed through a series of software packages: Image Chopper (Maddisys, Crowborough, East Sussex, UK), 3DFi, FiAlign, and FiRender (Source BioScience plc, Nottingham, UK), as previously described.21 The Image Chopper software was used to create 10 separate images from the flatbed scan where 10 slides were scanned at a time. The 3DFi software was then used to convert the resulting colour jpegs into 256 greyscale eight-bit tagged image format (TIFF) files and to crop each of these files to exactly the same size (both are pre-requisites for the FiAlign software). FiAlign was used to align each section with the next by applying transformations and rotations. This was performed manually, working on the first (reference) and second (active) sections. The active slice was moved either by rotating (left mouse button) or by XY movement (right mouse button) in relation to the reference slice until the highest cross-correlation coefficient, as shown on the FiAlign user interface screen, was achieved for each pair of sections. The active slice was then used as the reference slice, and the next section became the active slice. This process was repeated until all sections were aligned. Finally the aligned greyscale dataset was imported into the FiRender software, in order to produce (render) the final 3D reconstruction. FiRender uses back-to-front rendering in order to assign an opacity to each voxel (3D pixel). A colour palette was then applied with a blue colour assigned to normal lymph node tissue and red assigned to the metastatic breast tumour cells.

The dataset was first viewed in the Volume Slice Viewer (VSV) interface. This enabled the stack of re-aligned sections to be viewed in the X, Y and Z axes separately. The volume was then rendered to create a 3DR, which could be orientated and viewed from any angle using azimuth, elevation and roll controls. The VSV and 3DR for each of the 19 nodes were viewed, and the results were compared with those of the original H&E-stained and IHC-stained sections, in order to confirm that each reconstruction was a true representation of the original node, and metastasis within. An animation of each 3DR was then created, and the resulting videoclips were saved.

Although the cohort of SLNs was too small to perform statistical analyses, some basic clinicopathological data for each case were recorded in order for preliminary analyses of trends between metastatic growth patterns and clinicopathological factors to be assessed.

This study was approved by the Nottingham Research Ethics Committee (LREC no. 05/Q2402/117).

RESULTS

Brightfield microscopical analysis

The immunohistochemically stained sections of each SLN were analysed with respect to the location of the metastasis in relation to the afferent pole and to the growth pattern(s) present, as summarised in table 1.

In the majority of cases (17/19; 89.5%) metastatic breast tumour cells were present in the afferent pole of the lymph node. Of these 17 cases, 7 (41%) had metastases in the afferent pole only (fig 1A), 5 (29%) had metastases in both poles but predominantly afferent (fig 1B), 2 (12%) had metastases in both poles but predominantly efferent (fig 1C) and 3 (18%) were found to have metastases spread equally over both poles (fig 1D); only 2 (10.5%) cases out of the total number studied were found to have metastases in the efferent pole only (fig 1E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Examples of sentinel lymph nodes with metastases, localised using AE1/AE3 cytokeratin antibody immunohistochemistry, located at the afferent pole only (A), both poles but predominantly afferent (B), both poles but predominantly efferent (C), mid afferent/efferent poles (D), and the efferent pole only (E).

Three different growth patterns of metastatic tumour deposits were noted within the SLNs: sinusoidal (fig 2), where the tumour cells were present within the subcapsular sinus of the lymph node; nodular (fig 3), where the tumour cells formed a clump, or nodule, of tightly packed cells in the cortex and and/or medulla of the node; and diffuse (fig 4), where the tumour cells were dispersed through the cortex and/or medulla of the node in a diffuse spreading pattern. In some instances only one growth pattern was seen (as in figs 2 and 3), whereas in others they were seen in combination (as in fig 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Corresponding (A) H&E-stained and (B) AE1/AE3-stained sections of a sentinel lymph node expressing a sinusoidal pattern of metastasis. (C) Sinusoidal case Volume Slice Viewer screenshot showing the volume from the (i) z axis, (ii) x axis and (iii) y axis. (D) Sinusoidal case three-dimensional reconstruction screenshot illustrating tumour cells at the periphery and on the surface of the node.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Corresponding (A) H&E-stained and (B) AE1/AE3-stained sections of a sentinel lymph node expressing a nodular pattern of metastasis. (C) Nodular case Volume Slice Viewer (VSV) screenshot. (D) Nodular case VSV screenshot showing second area of metastasis deeper within node. (E) Nodular case three-dimensional reconstruction screenshot.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Corresponding (A) H&E- and (B) AE1/AE3-stained sections of a sentinel lymph node expressing a diffuse pattern of metastasis. (C) Diffuse case Volume Slice Viewer screenshot. (D) Diffuse case three-dimensional reconstruction screenshot.

The sinusoidal pattern was never seen alone. The nodular pattern was seen alone in seven (36.8%) cases and the diffuse pattern was seen alone in one (5.3%) case. The sinusoidal pattern was seen in combination with the nodular pattern in nine (47.4%) cases, and, of these nine, the nodular pattern was predominant in seven (77.8%) cases, and the sinusoidal pattern was predominant in two (22.2%) cases. The nodular pattern was seen in combination with the diffuse pattern in two (10.5%) cases and in both instances the diffuse pattern was predominant. The sinusoidal pattern was not seen in combination with the diffuse pattern.

3DR analysis

The findings of the brightfield microscopy analyses were confirmed by viewing the data in the VSVs and 3DRs The VSVs and rendered 3DRs of each SLN were analysed and compared with the original IHC-stained sections. The red/blue colour palette was shown to easily distinguish tumour cells from normal nodal tissue. The red areas visible in the VSVs and the subsequent 3DRs were confirmed to be true representations of the metastatic tumour cells, and the blue areas were confirmed to be representative of the non-involved SLN tissue. By viewing the image in three dimensions, the tumour could actually be seen infiltrating the LN. This is not seen when viewing two-dimensional (2D) sections alone.

Three cases have been selected to illustrate each of the metastatic tumour patterns seen: sinusoidal, nodular and diffuse. For each of these three cases, a photomicrograph of a representative single section stained with H&E, a corresponding photomicrograph of the section immunohistochemically stained with AE1/AE3, a screenshot of the volume slice viewer and a screenshot of the 3D reconstruction are depicted (see figs 2–4). Furthermore, an animated videoclip of the 3DR of each of these cases can be viewed.27

The case in fig 2 shows a predominantly sinusoidal pattern of metastasis, which partly invades the node. This is not easily discriminated on the H&E-stained section (fig 2A) but is clearly observed on the AE1/AE3-stained section (Fig 2B), and the findings are confirmed in the subsequent VSV (Fig 2C) and 3DR (Fig 2D). The case shown in fig 3 is an example of a small but distinct nodular growth pattern. The nodule is present at the periphery of the node, but not in the perinodal sinus. This is seen in both the H&E-stained (fig 3A) and the AE1/AE3 stained sections (fig 3B), and is confirmed in the VSV (fig 3C, D) and 3DR (fig 3E). Interestingly, analysis of the entire volume of this node within the VSV indicates that in fact two separate nodules are present within this node. This is not seen when viewing 2D sections alone. The case shown in fig 4 is an example of a predominantly diffuse pattern of metastasis, along with a somewhat more compacted band of cells encompassing the edge of the diffuse area, more nodular in appearance. This is seen on both the H&E-stained (fig 4A) and AE1/AE3-stained sections (fig 4B), and confirmed in the subsequent VSV (fig 4C) and 3DR (fig 4D) as scattered red dots throughout a large proportion of the node (diffuse) and small areas of compacted red cells (nodules).

Since this study was based on SLNs that had been previously diagnosed as positive by histological examination, none of these cases contained isolated tumour cells or micrometastases that could not be seen on the initial histological examination.

Correlations between metastatic growth patterns and clinicopathological data

The results of the metastatic growth patterns observed in each SLN were analysed with respect to clinicopathological data, as summarised in table 2.

Both cases with a sinusoidal metastatic pattern were tumours of no special type (NST). Of the 14 cases expressing a nodular metastatic pattern, 10 (71.4%) were NST, 3 (21.4%) were tubular mixed, and 1 (7.2%) was pure lobular. All three cases with a diffuse pattern were of tubular mixed tumour type. High tumour grade seemed to be associated with the sinusoidal pattern, and low grade associated with the diffuse pattern where both cases with sinusoidal pattern were of tumour grade 3, while only 4/14 (28.6%) of the cases with a nodular pattern were grade 3. Of the 14 cases with a nodular pattern, 4 (28.6%) had vascular invasion (VI), and 10 (71.4%) did not. None of the cases with a diffuse pattern had VI, but both of the sinusoidal cases did.

DISCUSSION

The aims of this study were to determine whether 3D voxel reconstructions of metastases within entire axillary LNs from breast cancer patients could be created using readily available and economical equipment, and whether such 3DR techniques could be utilised to evaluate the early metastatic disease process in SLNs in an attempt to understand the nature of this process in breast cancer.

In order for successful 3DR of an entire LN to be accomplished, it is imperative that slices taken at cut-up are pieced back together accurately. Routine laboratory handling of LNs at NUH involves slicing at 2 mm intervals perpendicular to the long axis, and embedding the slices in one or multiple paraffin wax blocks. It is not possible, therefore, to retrospectively assign each slice accurately to its original position within the node. It is for this reason that retrospective cases from the routine archives could not be used, and that prospective sample collection was required. SLNs were chosen since this would enable early metastatic disease to be studied, and orientation of the nodes in relation to the afferent lymphatic to be assigned. The spatial distribution of the metastases could then be studied.

The results have demonstrated that a high-resolution flatbed scanner and 3DR software can be successfully used to create reproducible 3DRs of breast tumour metastases within entire SLNs. Pilot experiments indicated that an intersection distance of 20 μm yielded the most acceptable compromise between labour intensity and accuracy in subsequent 3DR. Furthermore, that the reconstructions achievable using a high resolution flatbed scanner were comparable with those created using digitally scanned images at higher resolutions created using a system of digital camera and microscope with automated stage (E C Paish, unpublished observations). The use of a readily available high-resolution flatbed scanner for the purpose of creating the digital images for reconstruction proved to be economical not only financially, but also in time, since it was optimally found that at a resolution of 1200 dpi, 10 slides could be scanned simultaneously. This concurs with the findings of Jones et al (2006)21 who used a similar scanning method in preference to a digital camera and automated stage system such as that used by Tyrer et al (2000)22 and Kurien et al (2005).23

It has been shown that the 3DR findings concur with the initial H&E and IHC findings, and in some instances that they can provide additional information about the nature of metastatic growth and development within the node. For example, in fig 3, a second area of metastasis was identified when viewing the node in 3D; this may have been missed by routine screening of 2D H&E-stained and AE1/AE3-stained sections alone. This indicates that, although the current routinely adopted protocol of screening single sections from 2 mm slicing of LNs does in most cases provide a representative sampling of the node, some FN diagnoses can occur using this method. Clearly the protocol reflects the best compromise between accuracy of the diagnosis possible, and the time and resources necessary to provide this. However, if a particular part of the node could be identified that had the greatest probability of harbouring metastatic tumour, histological and pathological examination of the node could be focussed on this area increasing the likelihood of identifying metastases and thereby reducing the FN rate.

Furthermore, since the 3DR of tumour-involved nodes enables areas of metastasis to be visualised from all angles, rather than just 2D (the situation in routine evaluations currently performed by pathologists), tumour cells can be seen actually infiltrating the node. A 3DR is therefore able to illustrate the spatial growth and development of a metastasis; assessment of 2D sections alone cannot do this.

The process of 3DR as a whole can be considered laborious and time consuming. Nonetheless, it has clear roles in the research setting and as a tool for teaching. The findings of this study have confirmed that a faster, more financially economical scanning method can be used to create successful 3DRs of SLNs, and, with concurrent findings of 3DR of breast tumours already reported,21 this method could find added applications for many other tumours/tissues in the future.

The use of SLN technology has shown that orientation can be assigned to nodes, and the findings of the spatial distribution of metastases within the nodes concur with those of authors such as Cserni (2000)14 and Diaz et al (2003).15 In accordance with previous studies, it has been assumed that the efferent lymphatic is located at the opposite side of the node to the afferent lymphatic. It is acknowledged, however, that there may be some anatomical variance. It appears that metastatic breast tumour cells more frequently localise near the inflow junction of the afferent lymphatic vessel, where they first enter the node if travelling via the lymphatic system. It would be of interest to investigate if there are any factors present within this part of the LN that may be causing this attraction or supporting establishment of metastatic deposits, or whether the tumour cells simply localise themselves at the first point of entry. Cserni (2000)14 and Diaz et al (2003)15 reported metastases present in the same half of the LN in which the inflow junction of the afferent lymphatic vessel was located in 23/32 (72%) and 52/55 (94.5%) LNs respectively. Our study reports 17/19 (89.5%) LNs containing metastases present in the same half. Therefore, it seems pertinent therefore to propose that more detailed analyses of this half of the node should be advocated in the routine diagnostic setting, to facilitate increased detection rates of metastases, and reduce false-negative diagnoses. Larger studies are also needed to confirm the significance of this relationship. If confirmed, then this observation could be incorporated into the much-needed single standardised protocol for the optimum method of histological examination of SLNs.

Our incidental findings of distinct metastatic growth patterns in LNs are interesting as they have not previously been reported. The nodular growth pattern is the most frequent in this study, and it appears to be commonly associated with the presence of both of the other patterns. The sinusoidal pattern was not common and, although it was seen in combination with the nodular pattern, it was not seen in combination with the diffuse pattern. The preliminary associations found between metastatic patterns and clinicopathological factors have suggested that metastatic patterns may have prognostic significance. Since NST tumour type, high grade and presence of vascular invasion are all regarded as indicators of poor prognosis, the results suggest that the sinusoidal metastatic pattern may be associated with a poor prognosis, and the diffuse pattern with a good prognosis. Larger and more detailed studies are required to confirm these findings and to determine the relationships between the three patterns, the underlying mechanisms of their development and any associations with clinical outcome. We therefore intend to investigate these metastatic patterns in a larger cohort of samples. Since such studies would not require 3DR, a large retrospective case series could be collected and analysed by microscopical analysis of immunohistochemically stained single sections.

Exclusion of cases with micrometastatic disease from this study may have influenced the frequency of detection of the sinusoidal pattern alone as this site is recognised as a common site for identification of micrometastatic disease and is believed to be the main route for entry of metastatic cells into the parenchyma of the LN. However, since the SLNs collected that were reported as containing only micrometastastic deposits contained only single/tiny groups of cells in the subcapsular sinus, it was felt that the limited information regarding metastatic development available from the resulting 3DRs would not justify the time and effort necessary to create them.

There is, however, a requirement to address an approach to increase the likelihood for detecting small metastases in SLNs. We aim to investigate these issues in a further case series of micrometastatic cases, whereby the afferent lymphatic pole of each SLN will be marked in order to assess the site of the earliest metastatic deposits and their relation to the afferent lymphatic pole. The results of these investigations are expected to provide insights into the early metastatic disease process, and may enable more detailed analyses to be concentrated on this area of the node, so as to improve detection rates of micrometastases.

In conclusion, this study has successfully applied 3DR techniques in order to investigate and evaluate the metastatic growth of breast cancer in entire lymph nodes. It has provided an interesting insight into the nature of the metastatic disease process of breast cancer and potentially could be used to develop strategies to improve techniques for routine detection of early metastatic disease. By implementing the methods described, future study of other histological structures and disease growth patterns are possible.