SIVQ-LCM is an innovative approach that harnesses a computer algorithm, Spatially Invariant Vector Quantization (SIVQ), to drive the laser capture microdissection (LCM) process. The SIVQ-LCM workflow greatly improves the speed and accuracy of microdissection, with applications in both the research and clinical settings.
SIVQ-LCM is a new methodology that automates and streamlines the more traditional, user-dependent laser dissection process. It aims to create an advanced, rapidly customizable laser dissection platform technology. In this report, we describe the integration of the image analysis software Spatially Invariant Vector Quantization (SIVQ) onto the ArcturusXT instrument. The ArcturusXT system contains both an infrared (IR) and ultraviolet (UV) laser, allowing for specific cell or large area dissections. The principal goal is to improve the speed, accuracy, and reproducibility of the laser dissection to increase sample throughput. This novel approach facilitates microdissection of both animal and human tissues in research and clinical workflows.
Originally developed in the mid-1990s, laser capture microdissection (LCM) enables the user to precisely capture specific cells or cellular regions from a histological tissue section via microscopic visualization1,2. Many studies comparing molecular analysis of LCM versus tissue scrapes illustrate the value of the method3-12. In addition, there are three video protocol publications on the technology that are available for viewing13,14. However, despite its proven value, LCM can be tedious and laborious when the target of interest is a dispersed cell population in a heterogeneous tissue section, or when large numbers of cells are required for specific downstream applications such as proteomics. The burden placed on the human operator led us to develop a semi-automated dissection approach for LCM by combining a powerful image analysis algorithm to guide the LCM process15.
In collaboration with the University of Michigan, our laboratory at the NIH extended the previously developed and reported Spatially invariant vector quantization (SIVQ) algorithm in a manner to allow it to semi-automate the tissue selection process intrinsic to guided microdissection, thus making available a tool with the pathologist or life scientist in mind. Spatially invariant vector quantization (SIVQ) is an algorithm that allows the user to simply “click” on a histological feature of interest to create a ring vector (predicate image feature) that can be used to search the entire histological image, adjusting the statistical threshold as needed16-21. The resultant heat map displays the quality of matches to the initial predicate image feature and is subsequently converted into a single color (red) annotation map that can be imported into the LCM instrument. The automated selection software, AutoScanXT, is then used to draw a map based on SIVQ's annotation guiding the capture of the target cells from the tissue sample. The detailed protocol below describes the implementation of SIVQ into the microdissection workflow.
The described protocol was employed in accordance with NIH rules on the use of human tissue samples.
1. Tissue Preparation
2. Specimen Imaging
3. Algorithm Analysis of the Image
4. Microdissection
A FFPE human breast tissue section was immunostained for cytokeratin AE1/AE3 using a standard IHC protocol23. After staining, the tissue slide was placed on the ArcturusXT stage and the SIVQ-LCM protocol was initiated as described above. Since the tissue cannot be coverslipped for microdissection, the IHC+ stained cells can be difficult to discern visually (Figure 1A). Thus, to provide better index matching and an improved image, xylenes were added to the tissue section to create a pseudo-coverslip15 (Figure 1B). A jpeg image was then captured of the pseudo-coverslipped area and imported into SIVQ for the algorithm analysis. A predicate image feature (dark brown DAB stain) selected by the user initiated the SIVQ algorithm to analyze the image (Figure 1C). The SIVQ heatmap was then converted to a “red paint” that is recognized by the automated selection software (Figure 1D). The xylenes were allowed to evaporate and the SIVQ heatmap was imported into the automated selection software (Figure 1E) and the highlighted cells were dissected with the IR laser. The LCM cap was then moved to the QC station of the ArcturusXT instrument and visually inspected to assess dissection efficiency (Figure 1F). The remaining tissue was also inspected (Figure 1G) and the SIVQ heatmap was re-imported to further assess the microdissection efficiency (Figure 1H).
Figure 1. SIVQ-LCM of IHC stained FFPE breast tissue. A) Uncoverslipped image of cytokeratin AE1/AE3 stained FFPE breast tissue. B) Xylenes coated (pseudo-coverslipped) image of panel A. C) SIVQ generated heatmap using a ring vector (predicate image) captured from the brown chromagen. D) The heatmap was converted to a single color (red). In the ArcturusXT, the AutoScanXT software was trained to recognize the red annotation. E) The microdissection map generated from the AutoScanXT algorithm. F) Image of the microdissected breast epithelial cells on the LCM cap. G) Image of the tissue on the slide after microdissection. H) Overlap of the microdissection map (panel E) on the tissue area that was dissected. Please click here to view a larger version of this figure.
We present a protocol for the application of SIVQ-LCM to microdissect immunostained epithelial cells from FFPE human breast tissue. The use of an image analysis algorithm, such as SIVQ, reduces the amount of hands-on time required for the microdissection process. This is a potentially important advance for the field since operator time and effort is typically the rate-limiting step for the precise dissection of cells of interest. In the present protocol, we specifically adapted our procedure to the ArcturusXT instrument, although it likely will be possible to adapt SIVQ and other algorithms to other commercially available microdissection instruments as well. In addition to improving efficiency, this new approach may allow users who are non-histopathology experts to perform microdissection. Finally, the use of an algorithm to drive the microdissection process may enable greater reproducibility from site-to-site by removing user subjectivity in identifying cells of interest.
While SIVQ exhibits broad adaptability to multiple classes of image subject matter, including histology, it is important to recognize that it is not intended to serve as a universal image processing/segmentation tool, but rather, as a high-efficiency first-pass foreground selection tool. Used in this capacity, it has high utility for predicting whether or not a particular class of imagery subject matter is suitable for numerical segmentation approaches. When SIVQ is successful, there is high predictive power that either: a) a more refined and targeted image analysis/segmentation approach will be highly successful and/or b) an optimized, machine-selected SIVQ vector will be effective in rendering a clinical-workflow-ready solution. In either circumstance, the use of SIVQ serves as an effective triaging tool to identify imagery subject matter that poses computationally tractable solutions in terms of foreground segmentability.
While we previously described the use of the SIVQ-LCM protocol only for mRNA expression microarray experiments15, we believe the protocol is applicable to most downstream molecular assays, and may be particularly useful for molecular assays that require large numbers of cells, such as proteomics. For these studies, large amounts of cells need to be procured due to the inability to amplify proteins. In addition, the advent of newer technologies, such as Next-Gen Sequencing (NGS), emphasizes the need to begin with as pure of a sample as possible. The ability for SIVQ-LCM to isolate cells via both morphological features and/or staining qualities could enable the ability to more easily isolate pure cell populations.
It is critical to have properly prepared tissue for reproducible results for both the SIVQ analysis and microdissection. For optimal conditions for LCM it is best to perform dissects in a low humidity room and maintain properly dehydrated tissue. In addition, ensure that the staining of the tissue is reproducible. Finally, familiarity with the dissection instrument is essential to successful SIVQ-LCM.
The SIVQ-LCM protocol presented here is the first steps towards semi-automated microdissection, which may have applications in future clinical testing since current dissection methods are not easily capable of handling large workloads. In summary, we believe that SIVQ-LCM is a new foray into the field of automated microdissection and provides a starting point from which investigators and algorithm developers can further improve the technology.
The authors have nothing to disclose.
The study was supported in part by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research.
Positive Charged Glass Slides | Thermo Scientific | 4951Plus-001 | |
Xylenes, ACS reagent, ≥98.5% xylenes + ethylbenzene basis | Sigma Aldrich | 247642 | CAUTION: PLEASE USE PROPER SAFETY PROCEDURES. |
Ethyl Alcohol, U.S.P. 200 Proof, Anhydrous | The Warner-Graham Company | 6.505E+12 | CAUTION: PLEASE USE PROPER SAFETY PROCEDURES. |
Arcturus CapSure Macro LCM Caps | Life Technologies | LCM0211 | |
ArcturusXT Laser Microdissection Instrument | Life Technologies | ARCTURUSXT | |
AutoScanXT Software | Life Technologies | An optional image analysis program for the ArcturusXT Laser Microdissection Device. This is software is required for SIVQ-LCM. | |
Spatially Invariant Vector Quantization (SIVQ) | University of Michigan | This tool suite is publicly available for academic collaborations. For access to the SIVQ algorithm, please contact Dr. Ulysses Balis [Ulysses@med.umich.edu] |