A mill based instrument and software system for dissecting slide-mounted tissue that provides digital guidance and documentation

Study design, setting, participants, and methods

The development of the mesodissection system followed an ISO 9001 compliant phase review process. The breadboard prototype instrument was developed at AvanSci Bio using a modified commercial milling machine and the consumables were constructed from common hardware components and pipet tips. The laboratory prototype instrument and software were developed at AvanSci Bio, ARUP Laboratories, and the Scientific Computing and Imaging (SCI) Institute at the University of Utah, using a combination of custom and off the shelf components. Production consumables were manufactured from custom injection molded components. The imaging software was written in C++ and interfaces with various hardware drivers. The initial laboratory prototype was presented in the exhibit hall of the Association of Molecular Pathology 2011 Annual Meeting to obtain potential user feedback. This feedback was used to refine the design of the current instrument.

Milling instrument and xScisor disposable

A new platform that uses milling technology to dissect tissue from slide-mounted tissue sections was developed (See Figures 1 and 2 for a detailed description). In the basic mill design, an object is attached to a table (stage) capable of controlled X and Y axis movement such that it can be driven against a fixed rotating cutting bit thereby shaping the object. However, as milling is typically not used for the purpose of collecting fragments, a plastic mill bit termed the xScisor was developed that simultaneously dispenses liquid, cuts tissue, and aspirates the tissue fragments from the surface of the glass slide (Figure 3). Low cost manufacturing methods were developed so the xScisor could be disposable to prevent sample cross-contamination. Because tissue is relatively soft compared to glass, a spring pressure controlled system was designed such that the xScisor blade rests on the slide surface with sufficient downward force to cut through the tissue but glides across the glass slide. Milling tissue from glass slides also provided the opportunity to place a digital microscope below the slide in order to view the process, direct the dissection, and generate digital documentation.

Software and workflow

A software package was developed that was modeled after a typical manual slide-mounted tissue dissection workflow found in many molecular pathology labs (discussed above, see Figure 4 for a more detailed description). This software provides an interface to digitally indicate AOIs and save dissection reference images. Alternatively, the dissection reference image can be generated by digital scanning technology if it is saved in either .jpg, .png. or .tif file format. In the laboratory, the reference image was imported from the database, aligned to the magnified live view from a serial tissue section, the AOI transferred to overlay the live image and guide the dissection, and finally a digital report in PDF format was generated.

Tissue samples used in this study

All human FFPE tissue sections used for the studies described here were derived from left over/unused normal regions of placenta, liver, bowel, kidney, and skin specimens from the Anatomic Pathology Gross Room at the University of Utah (Department of Pathology, School of Medicine). As these tissue sections were anonymized, they are exempt from IRB approval. All human tissues were processed using standard FFPE methods, and the resulting tissue blocks and the mouse-human fusion tissue blocks were sectioned at 5 micron thickness. The formalin perfused mouse liver and kidney tissues that were used to make the mouse-human fusion blocks were surplus tissues obtained from studies of the mouse olfactory system done at the University of Utah (Department of Physiology). All the mouse olfactory tissue sections were from the same source, but had been OCT embedded, and then frozen sectioned at 10 to 60 microns thickness. These olfactory tissue sections were surplus from the intended studies. For the studies described here, the OCT was replaced with paraffin as described below.

Resolution and accuracy

Both resolution and accuracy were quantitated using Microsoft PowerPoint by visual examination of the post-dissection digital images including the scale bars generated by the mesodissection software. The accuracy values were determined by attempting to fill the difference between the intended and actual boundary with a line of relative width of 50, 100, or 250 microns (Figure 5A and B). Standard deviation was calculated using Microsoft Excel.

Efficiency

Serially cut tissue sections were either deparaffinized with AvanSci Bio’s mineral oil/alcohol system, or left paraffinized, then dissected by hand or using the mesodissection system. In order to dissect the same AOIs using both methods, the post manual dissected tissue sections were digitally outlined and used as guides for mesodissection. The resulting tissue samples were recovered in one of the following liquids: TE (10 mM Tris, 1 mM EDTA, pH 8.5), TE + 0.1% SDS, or light mineral oil, then subject to Proteinase K digestion by adding 5 μg of Proteinase K to the samples in TE and TE + SDS, or by adding 70 μl of TE with 5 μg proteinase K to the samples in mineral oil. Incubation was performed on the Thor (AvanSci Bio) programmable heater-shaker: 65°C 30 min 1500 rpm; 95°C 10 min 1500 rpm; 25°C 1 min 450 rpm. The samples were then centrifuged and assayed for genomic DNA concentration using PicoGreen according to the manufactures’ protocol (Life Technologies) as a measure of total sample recovered.

Purity

To make the mouse-human fusion blocks, the formalin treated tissue samples described above were held in close proximity while casting in paraffin. The resulting 5 micron slide-mounted tissue sections were deparaffinized and dissected as indicated. The recovered tissue samples were Proteinase K digested as described above, then subject to both dual amplicon multiplex endpoint PCR with Taq Platinum using the recommended conditions (Life Technologies) and 2% agarose gel electrophoresis, or single amplicon qPCR using the Power SYBR Green master mix and the recommended conditions (Life Technologies). The PCR primer pair (amplicons) sequences (directed to either the mouse or human Cox1 mitochondrial genes) are: AGGGGACCCAATTCTCTACCA + CTCCGTGTAGGGTTGCAAGT (mouse) and TTCGGCGCATGAGCTGGAGTCC + AGTTGCCAAAGCCTCCGATT (human).

RNA isolation, reverse transcription, and quantitative PCR

The formalin perfused, OCT embedded, frozen sectioned, 10 to 60 microns thick mouse olfactory tissue sections that were obtained for this study were relatively poorly adhered to the slides surface and tended to be dislodged rather than cut by the spinning xScisor blade. Therefore, the tissue sections were gently rinsed in 0.1X PBS to remove the OCT, dipped in molten paraffin for 5 minutes, then the excess paraffin allowed to drain by standing the slide on edge at 70°C for 5 minutes. These tissue sections were milled using 70 μl per sample of light mineral oil, then 70 μl of TE, pH 8.5 containing 5 μg of protease K was added and the tubes mixed. Protease K digestion was carried out using the Thor heater-shaker at 60°C 30 min 1500 rpm; 82°C 15 min 1500 rpm; 25°C 1 min 450 rpm. The overlaying mineral oil was removed using Wicking Strips (AvanSci Bio), then 90 μl PKD buffer (Qiagen) was added to achieve 160 μl total volume. Next, 16 μl of DNase booster buffer and 10 μl of DNase I (Qiagen) were added to each tube, the tubes incubated at room temperature for 15 minutes and the RNA isolated as described in the RNeasy FFPE kit (Qiagen). The manual dissected material was processed using the limonene deparaffinization protocol described in the RNeasy FFPE kit. cDNA was generated using the Applied BioSystems High Capacity Reverse Transcription kit and Quantitative PCR was performed using Applied BioSystems Power SYBR Green reagent as described in the manufacturer’s protocol. The primers pairs used in the qPCR were GATGACTGAGTACCTGAACCG + CAGAGACAGCCAGGAGAAATC (mouse Bcl-2 cDNA) and GCCCTCCGTATCTTACTTCAAG + GCGGTCCAGGTAGTTCATG (mouse Cyclin D1 cDNA).

FISH

Slide-mounted tissue sections were deparaffinized using d-limonene or AvanSci Bio’s mineral oil/alcohol system, then mesodissected using TE plus 0.1% Tween-20 as the milling solution. Recovered tissue fragments were centrifuged for 2 minutes, the majority of the supernatant discarded, the fragments resuspended, and 2 μl aliquots were spotted onto Fisher Scientific Capillary Gap plus slides (130 micron) and baked at 65°C for 2 hours. These slides were then subject to tissue FISH processing using the Kreatech recommended FFPE Tissue FISH protocol.

Mesodissection capabilities and performance metrics

Four performance metrics were established to test the mesodissection technology: resolution, accuracy, efficiency, and purity. The resolution (Figure 5A) is the minimum dissectible area determined using point dissections and is primarily dependent on the diameter of the cutting blade but is also dependent on any non-concentricity of the blade (runout). Currently, the minimum resolution is a circle about 170 microns in diameter. xScisors with 100 micron plastic cutting blades have been made but the wear rate was excessive; a steel tip blade is currently under development. A second metric, accuracy (Figure 5B), is a measure of how well the user can guide stage movement using the joystick (transverse movement). The average accuracy from this test, performed by a single user, was 60 microns. Most users with practice can achieve 100 micron accuracy, and point dissection accuracy is often better as the user can precisely position the blade before lowering it onto the slide surface. The third metric, recovery efficiency (Figure 5C), is determined by Pico Green quantitation of DNA following Proteinase K digestion of the recovered tissue. The recovery efficiency of both mesodissection and manual macrodissection are similar and also appear to be near 100% by visualization. The fourth metric, purity (Figure 5D) is a measure of potential contamination from neighboring undissected tissue (for example, if DNA were to leach out of the surrounding tissue and be picked up by the milling solution). Purity was determined by dissecting immediately adjacent human and mouse 5 micron liver tissue sections at variable distances from the intersection. The recovered samples were treated with Proteinase K, then subjected to multiplex endpoint PCR/gel electrophoresis, and single amplicon quantitative PCR. The PCR primer pairs (amplicons) were directed to either the human or mouse Cox1 mitochondrial sequence. As seen in the gel image, while both bands are present from a control containing an equal amount of each tissue (H + M), only the expected band is detectable from the dissected samples indicating little to no contaminating template. For the qPCR titration experiments, the entire section of human and mouse tissue were dissected separately, treated with Proteinase K, and then mixed at the ratios shown. While the human Cox1 amplicon amplified earlier than the mouse, the response to relative template ratios is clearly visible. In the single sample dissections, very little contaminating template (the titration experiments indicate it is significantly less than 1%) is recovered even from dissections immediately adjacent to the tissue intersection. The results from both multiplex endpoint PCR and real time qPCR indicate purity is greater than 99%, but this result could potentially be sample dependent.

Applications

The mesodissection system was applied to two common techniques in molecular diagnostics, RNA expression analysis and anatomical FISH (Fluorescent In Situ Hybridization) analysis in order to investigate capability. Because RNA is labile, many labs prefer to leave the tissue sections paraffinized during dissection to minimize exposure to moisture and atmospheric oxygen. Therefore, mineral oil was used as the milling solution, as it can aspirate paraffinized tissue and maintain a non-aqueous environment (used in Figure 5C). The recovered mineral oil/tissue fragment mixture was added directly to an equal volume of aqueous Proteinase K reaction solution where it formed a separate upper phase. When the tube was subject to heating and shaking, the tissue fragments migrated to the lower aqueous phase and were digested by the Proteinase K, which also digests RNase. Following the digestion, the organic phase was removed and RNA isolation, cDNA synthesis and quantitative PCR were performed using commercial kits. The results indicate the mesodissected material produced a larger quantity of RNA, as assayed by Reverse Transcription-quantitative Polymerase Chain Reaction (RT-qPCR), but sample loss during the additional centrifugation steps used for deparaffinization of the manual dissected sample likely accounts for the difference (Figure 6A). A link to a poster describing an endpoint RT-PCR analysis of these same samples is available at http://www.avanscibio.com.

Tissue fragments that had been mesodissected from slide-mounted tissue sections were re-adhered to slides and used for FISH analysis (Figure 6B). The majority of the fragments were present in a single thickness (not folded or stacked) with minimal obvious damage. The tissue fragments remained bound to the slides throughout the subsequent FISH processing steps and produced good quality FISH chromosomal signals. Testing using a variety of human tissue samples (described above, all 5 microns thick) and dissection conditions found that fragment size (determined by visual inspection using a scale bar) was variable ranging from just a few cells to over a millimeter in diameter. However, for a given dissection condition, over 90% of the total tissue area was within a 2-3 fold size range. Dissection of well adhered tissue (for example paraffinized tissue dissected using mineral oil) produced small fragments (2-10 cells in diameter) whereas dissection of less well adhered tissue (for example deparaffinized pre-moistened tissue) produced larger fragments (half to one millimeter in diameter). No obvious tissue type specific effects on tissue fragment size were noted, but we did observe that the tissue tended to tear along natural boundaries. For example, cross sections of kidney tubules remained intact as small rings of cells. We also noted some physical dissection and handling effects on fragment size. For example, larger slower rotating xScisor blades and gentle use of pipet tips with larger openings in the subsequent handling steps tended to produce larger tissue fragments.

An FFPE tissue block was mounted onto the mesodissection instrument stage using tape and the Z-axis contact position of the xScisor adjusted using potentiometers on the electronics board (an earlier version instrument). The tissue block was then dissected using mineral oil and overhead visual guidance (Figure 6C). The depth of the cut appeared to be related to the depth of the blade as the end of the inner syringe barrel appeared to ride on or very near the upper surface of the tissue block. Tissue fragments along with the expected volume of mineral oil were recovered suggesting sufficient dissection liquid fluid flow exists, possibly through the path cut in the tissue block. A significant problem with this approach is difficulty visualizing the dissection process in real time; the tissue block is opaque making visualization from beneath the block very difficult.

Biochemical considerations

The mesodissection system is compatible with both paraffinized and deparaffinized FFPE tissue sections mounted on standard glass slides. A variety of milling solutions have been tested and many more presumably can be used as long as they do not degrade the plastic xScisor and can hold the tissue fragments in suspension. For the latter reason mineral oil or SDS containing solutions were used to hold paraffinized fragments in suspension. When using mineral oil, the subsequent Proteinase K reaction can be performed by adding an aqueous solution to the recovered fragments and incubating on a heater-shaker. In these conditions, the tissue fragments migrate from the upper organic phase into the lower aqueous phase where the digestion occurs immediately. For expression analysis studies, dissection of still paraffinized tissue using organic milling solution is desirable as it minimizes exposure of the tissue to air and moisture prior to the Proteinase K step. A problem with the use of mineral oil is dissection visualization is somewhat impaired. Dissection visualization of paraffinized tissue sections is better using solutions containing SDS. SDS solutions can aspirate paraffinized tissue to a modest degree and SDS is compatible with Proteinase K. However, SDS must be removed prior to most subsequent biochemical reactions whereas post Proteinase K tissue lysates recovered from mineral oil can go directly into many downstream biochemistries.

Every type of transparent slide tested including both plain and charged glass was compatible with mesodissection as long as the tissue sections were reasonably well-adhered to the surface. Standard FFPE sections of most tissue types and thicknesses worked well, but occasional difficulty was encountered dissecting calcified tissue (the blade tends to ride over the top of the tissue). We found that frozen sectioned OCT embedded tissue was sometimes poorly adhered to the glass surface and when contacted by aqueous milling solution tended to dislodge rather than dissect. The problem tends to be worse with thicker sections (greater than 10 microns). In these cases, rinsing with an aqueous buffer to remove the OCT, dehydrating in alcohol, then dipping in molten paraffin and allowing to drain at an elevated temperature in order to remove the majority of molten paraffin greatly improved adhesion and provided good quality RNA dissection. Partial deparaffinization using AvanSci Bio’s mineral oil/alcohol system allows for tissue staining and improved visualization.

We used the mesodissection system to generate samples for two common applications, RT-qPCR and FISH. The results show that mesodissected samples are compatible with RNA expression analysis using RT-qPCR. The recovered sample-mineral oil mixture can be added directly to a Proteinase K reaction without an extraction and centrifugation step. Elimination of this step likely improved the RNA yield compared to a commonly used limonene deparaffinization protocol. We demonstrate that fragments of tissue generated by mesodissection can be re-adhered to slides and used for FISH analysis. This application could be used to increase FISH throughput as well as minimize hybridization probe and precious sample usage. A collaborator has used the mesodissection system to obtain samples for protein analysis by mass spectrometry and reports the system is comparable in quality and yield to other methods of tissue sample retrieval (Dr. David Krizman, Oncoplex Diagnostics). We also showed that the mesodissection system is capable of dissection directly from tissue blocks. Dissection directly from tissue blocks produces a lot of sample since the cut depth approaches 100 microns without the need to generate and dissect individual slide-mounted tissue sections. A possible downside is that the deeper the dissection, the less information one has about the material being recovered. Our further development of this capability is dependent on user interest, as it will require engineering of clamps mounted on the X-Y stage to properly position the tissue blocks and improvements to the dissection visualization methods.

Factors improving dissection performance

In addition to the factors discussed above, the following list of milling variables can improve mesodissection performance and should be considered when operating the instrument: 1) The cutting effectiveness, diameter of the cut area, and wear rate of the blade is a function of the downward force of the mill head, controlled using the adjustable position spring stop. Once properly set, it is usually not necessary to readjust the spring pressure. However, sometimes it is necessary to increase pressure when using larger blades, particularly when cutting paraffinized tissue, and in this situation, going over the same area multiple times can be helpful. 2) Coordinated movement of the digital AOI with the live image of the tissue section is dependent on precise calibration of the instrument’s slide stage with the 2iD software. 3) The aspiration rate can affect recovery efficiency; in general, increase the aspiration rate to the point where most of the dissection solution is used for a given sample, then centrifuge to pellet the recovered sample and discard excess dissection solution. 4) When doing point dissections, it is necessary to depress the pulse button for a couple seconds in order to aspirate the dissected tissue, otherwise aspiration occurs only with transverse stage movement. 5) To obtain clean cut edges, the leading edge of the blade should be rotating into the leading edge of dissection. For example, because the blade rotates counter-clockwise, dissection around the outside edge of an AOI should proceed in a counter-clockwise direction. 6) When dissecting deparaffinized tissue with TE or water, the dissected tissue occasionally forms clumps, which can be flung outside of the aspiration area by the rotating xScisor rather than being aspirated. Addition of a small amount of detergent (i.e. 0.1% Tween-20) to the milling solution can greatly reduce the formation of tissue clumps and improve aspiration. 7) The design of the xScisor results in simultaneous dispense and aspirate functions because both functions are controlled by the same movement of the plunger. The aspiration volume is greater than the dispense volume by the volume of the xScisor plunger. The effect is the column of collected liquid is regularly broken by small air bubbles. If a small amount of milling liquid is pre-aspirated, these air bubbles can be used as boluses to physically separate dissected areas or place an air bolus between the plunger and the recovered liquid, which can be used to increase recovery efficiency. The disadvantage of pre-aspiration is a reduction of total aspiration capacity.

Next generation instrument

While the current mesodissection instrument provides a number of advantages related to the digital workflow, it is joystick driven and thus typically does not result in a time savings compared to hand dissection of larger AOI’s. A number of laboratories have expressed interest in automation and therefore, we are developing the next generation instrument with a fully automated stage. The AOIs will be transferred manually as this step requires user verification. However, once transferred, the software will automatically design a milling path for each AOI and perform the dissection at a significantly faster rate. This additional capability not only frees user from dissecting using the joystick, but minimizes any impact of impaired dissection visualization when using unstained and still paraffin embedded tissues.