Improving Yields in Multi-analyte Extractions by Utilizing Post-homogenized Tissue Debris

Abstract

In multi-analyte extractions, tissue is typically homogenized in a lysis buffer, and then DNA, RNA, and protein are purified from the supernatant. However, yields are typically lower than in dedicated, single-analyte extractions. In a two-part experiment, we assessed whether yields could be improved by revisiting the normally discarded, post-homogenized tissue debris. We initially performed additional homogenizations, each followed by a simultaneous extraction. These yielded no additional RNA, 13% additional DNA (which became progressively more degraded), and 161.7% additional protein (which changed in proteome when analyzed using SDS-PAGE). We then digested post-homogenized tissue debris from a simultaneous extraction using proteinase K and extracted DNA using silica spin columns or alcohol precipitation. An average additional DNA yield of 27.1% (silica spin columns) or 203.9% (alcohol precipitation) was obtained with/without compromising DNA integrity (assessment by long-range PCR, DNA Integrity Numbers, and size at peak fluorescence of electropherogram). Validation using a cohort of 65 tissue blocks returned an average additional DNA yield of 31.6% (silica columns) and 54.8% (alcohol precipitation). Users can therefore refreeze the homogenized remnants of tissue blocks rather than disposing of them and then perform additional DNA extractions if yields in the initial multi-analyte extractions were low.

Introduction

Fresh-frozen (FF) tissue biospecimens are commonly used for research because the nucleic acid and protein that can be extracted from them are of higher yield and integrity compared with those extracted from formalin-fixed, paraffin-embedded (FFPE) tissue. However, given that clinical pathology workflows are designed to deliver the FFPE tissue required for patient diagnosis, collecting and then storing additional FF tissue for research is expensive, time-consuming, and can be logistically difficult. It is becomingly increasingly common for researchers to perform multi-analyte extractions, whereby DNA, RNA, and protein or two of these analytes are simultaneously extracted from each tissue block. This procedure enables one piece of tissue to feed several projects and/or for multi-omic analyses to be performed using the same cellular mass, minimizing the problem of intra-biospecimen tissue heterogeneity, which can be extensive.^1,2 A disadvantage associated with multi-analyte extractions, however, is that researchers can expect yields of at least one analyte to be lower than those found in an optimized single-analyte extraction. Protein is usually recovered by precipitation/resolubilization after being centrifuged through nucleic acid–binding silica spin columns, which is less efficient than homogenizing the tissue directly in the protein buffer, and DNA yields become compromised when the proteinase K digest that forms part of a DNA-only extraction is necessarily excluded from a multi-analyte extraction. ³ These losses can become critical, especially when a research project has a minimum yield requirement of more than one analyte.

Most FF simultaneous extraction protocols start with the tissue undergoing mechanical homogenization in a lysis buffer, often using a bead mill. Then, the homogenate is centrifuged and the supernatant containing solubilized nucleic acid and protein is applied to the extraction protocol. The centrifuge tubes, containing the homogenization beads and any residual tissue debris, are discarded. The combination of physical and chemical disruption and lysis of the tissue during homogenization is not universally effective because although factors such as the volume of lysis buffer and bead mill settings (velocity, time, bead size and number) can be optimized in respect of a tissue type, other critical factors such as the size of the tissue blocks and their morphology (e.g., extent of fibrosis) are inconsistent across a cohort of tissue blocks. Insufficient homogenization results in incomplete cell lysis and compromised nucleic acid/protein yields, but excessive homogenization is also detrimental because lysis buffer converts to froth, excessive heat can be generated, and nucleic acid or protein integrity becomes compromised. So, even when optimized in respect of a particular tissue type and maximum block weight, a homogenization protocol is ultimately still a compromise, and it is commonplace to see fragmented tissue debris in some of the post-homogenized centrifuge tubes when they are discarded.

We present a study to establish whether these post-centrifuged tissue debris contain sufficient nucleic acid and protein that, if extracted, could rescue biospecimens that have failed Quality Control (QC) due to low yield. Given that neither nucleic acid nor protein yield cannot be predicted with any confidence at the onset of an extraction, it would be straightforward for a user to freeze the centrifuge tubes containing the tissue debris and homogenization beads instead of disposing of them, and then for the low-yielding blocks perform additional extractions from them. We evaluated this approach in a two-part experiment. In the first part, four sequential multi-analyte DNA, RNA, and protein extractions were performed on 11 FF tissue blocks. Each extraction involved adding lysis buffer to the tissue debris from the previous extraction and homogenizing them again in the bead mill. In the second part, we addressed the problem of compromised DNA yields in a multi-analyte extraction, but rather than using the bead mill we instead subjected the tissue debris from the multi-analyte extraction to a proteinase K digest and then extracted DNA only using either silica spin columns or alcohol precipitation.

Materials and Methods

Biospecimens and Ethical Statement

This project used FF endometrium tissue blocks that had been collected for cancer research projects. The tissue was stored at −80C and shipped to the Integrated Biobank of Luxembourg on dry ice, whereupon simultaneous DNA/RNA/protein or DNA/RNA extractions followed by quantification and QC assays as specified by the researchers performing the cancer research projects were carried out. The nucleic acid and protein extracts were then returned to their originating institutions. The work described in this article used the data generated in the course of the QC of these primary extractions (n=152) and in addition, the tissue debris that would otherwise have been disposed of following tissue homogenization (n=76). Although we were able to perform the QC assays that we considered appropriate for this study in respect of the tissue debris, comparisons between the primary and secondary extractions are restricted to those QC assays specified for the former. Written informed consent had been obtained from all patients, and the Institutional Review Boards of the institutes providing the tissue had approved the project. The work was also approved by the Integrated Biobank of Luxembourg’s ethics committee.

Repeated Homogenizations of Tissue Block Debris and Then Additional Multi-analyte Extractions

The first part of the experiment used 11 FF endometrium tissue blocks (5–48 mg) plus a control extraction consisting of a frozen cell pellet of the MCF7 cell line. The three heaviest blocks (40, 43, and 48 mg) exceeded the maximum weight threshold of the extraction kit (30 mg). Our rationale for using them was to establish whether performing multiple homogenizations might facilitate efficient extractions from such large blocks, thereby avoiding the necessity to cut them to <30 mg. The tissue was homogenized using a TissueLyzer LT bead mill (Qiagen; Hilden, Germany), with settings previously optimized for multi-analyte extractions of this tissue type: 4-min homogenization at 50 Hz with a break after 2 min to allow for cooling in 2-ml Safelock centrifuge tubes (Eppendorf; Hamburg, Germany), each containing one 5-mm stainless steel bead and 600 µl of RLT lysis buffer (Qiagen) plus 2% 2M dithiothreitol (DTT) (Sigma-Aldrich; St. Louis, MA).

The AllPrep DNA/RNA/Protein Mini kit (Qiagen) was used for the extractions. Immediately after homogenization, the Safelock tubes were centrifuged for 3 min at 25,000 × g and then the supernatant applied to the AllPrep DNA spin column. The Safelock tubes containing the homogenization bead and any insoluble tissue debris, which would normally be discarded, were instead placed in a −80C freezer. The nucleic acid and protein extractions were performed as per the kit’s protocol, with the optional nuclease digests included to avoid the coelution of RNA in the DNA extraction and vice versa. RNA and DNA were eluted in 50-µl water and 100-µl EB buffer (10 mM Tris-Cl, pH 8.5), respectively, as per the kit protocol, but the protein pellet was solubilized in 100 µl of 4% SDS, 25 mM HEPES pH 7.6, and 1 mM DTT (all Sigma-Aldrich) rather than in the buffer provided in the kit due to the requirements of the downstream proteomic platform.

To establish whether the fragmented remnants of the original tissue blocks still contained recoverable nucleic acid and protein, the above process was repeated three additional times. Each homogenization was performed in the original Safelock centrifuge tubes using the same homogenization bead: only fresh lysis buffer was added.

Nucleic Acid Quantification and Analyses

RNA was quantified by OD260 nm photometry using a Synergy MX spectrophotometer (BioTek Instruments; Winooski, VT) and its integrity assessed using RNA Integrity Numbers (RINs) obtained by running 1 µl of RNA on Nano chips on a 2100 Bioanalyzer nanoelectrophoresis platform (Agilent Technologies; Santa Clara, CA). RINs range from RIN 1 (RNA completely degraded) to RIN 10 (RNA completely intact).

DNA was quantified by pico green fluorometry using the QuBit dsBR assay and a QuBit 4 spectrofluorometer (both ThermoFisher Scientific; Waltham, MA). OD260:280 nm purity ratios and DNA concentrations were obtained using the Synergy spectrophotometer, but the latter was exclusively used (unless stated otherwise) to calculate the percentage of total DNA that was double-stranded (% dsDNA = fluorometry concentration/photometry concentration × 100). DNA integrity was also assessed using nanoelectrophoresis by applying 1 µl of DNA to Genomic DNA ScreenTapes that were run on a 4200 TapeStation (all Agilent Technologies). DNA Integrity Numbers (DINs) range from DIN 1 (DNA completely degraded) to DIN 10 (DNA completely intact). The electropherograms plot each sample’s fluorescence intensity against size (200–60,000 bp), and in addition to DIN, the software calculates the size (in bp) of the most abundant (i.e., fluorescent) peak.

For the second, third, and fourth homogenizations only, DNA integrity was evaluated using long-range PCR (this assay was not specified for the DNA from homogenization 1 destined for the cancer research project). Aliquots of 10 ng DNA (assessed using pico green) or 18 µl (where DNA concentration was too dilute to allow 10 ng to be loaded onto the gel) were amplified using GoTaq Long PCR Master Mix (Promega; Madison, WI) and primers that generate a 15-kilobase-pair (kbp) amplicon of the plasminogen activator, tissue type (PLAT) gene, in 31 cycles of PCR in a C1000 thermal cycler (Bio-Rad; Hercules, CA) as previously described. ⁴ The PCR results were visualized on a 0.7 % agarose gel stained with GelRed (Biotium; Fremont, CA) and imaged using UV transillumination on an ImageQuant LAS4000 gel imaging system (GE Healthcare; Chicago, IL). Where 10 ng DNA had been amplified, both positive and negative results were accepted, but where 18 µl of DNA (i.e., <10 ng) had been amplified, a tissue block that returned a negative result was excluded from the analysis because the result might have been a consequence of there being insufficient DNA to amplify as well as the DNA being too degraded to generate an amplicon.

Protein Quantification and Analyses

The detergent compatible (DC) Assay (BioRad) was selected for protein quantification because it was compatible with the protein buffer. Aliquots of 5-µl protein in a total volume of 230-µl assay reagent were quantified in 96-well plates according to the manufacturer’s protocol. The plates were read on the Synergy MX at OD750 nm using a standard curve of 200–1600 µg/ml, made from solubilizing lyophilized albumin (Sigma-Aldrich) in the protein buffer.

Six tissue blocks yielded protein sufficiently high in concentration for 10 µg from the second, third, and fourth homogenizations to be evaluated using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Protein from homogenization 1 was not available as it was required for the cancer research project. Crude protein was mixed with 4× Laemmli Sample Buffer containing 50 mM DTT, heated at 70C for 10 min, and then loaded onto sequential wells of 7.2 × 8.6 cm, Mini-PROTEAN TGX, 4–20% polyacrylamide gels (all Bio-Rad). The electrophoresis was performed at 200 V in 25 mM Tris, 192 mM glycine, and 0.1 % SDS running buffer until the trace dye approached the bottom of the gel. Gels were fixed in 50% methanol and 7% acetic acid, stained overnight using SYPRO Ruby (ThermoFisher Scientific), washed in 10% methanol and 7% acetic acid (all Sigma-Aldrich), and then imaged using UV transillumination on the ImageQuant previously used for DNA gels. Automated exposure settings were used.

The gel images were analyzed using ImageQuant TL software (GE Healthcare). The software automatically detected any protein bands in each lane of the gel, performed background subtraction, calculated the volume of each detected band, and exported the data to Microsoft Excel. For each gel lane (i.e., homogenization), the volume of each band was calculated as a percentage of the total volume of all bands in that gel lane. The bands in the three adjacent lanes of the gel that were sequential homogenizations from the same tissue block were then numbered, with bands present in more than one homogenization of the same tissue block having the same number: each protein band could therefore be identified as being unique to one homogenization or common to more than one homogenization and evaluated in terms of changes in volume in subsequent homogenizations.

Digesting Tissue Block Debris to Yield Additional DNA

The second part of the experiment tested the hypothesis that the post-homogenized tissue debris from a simultaneous DNA/RNA/protein extraction still contained DNA that could be extracted in a DNA-only extraction. These secondary DNA extractions included the proteinase K digest that is usual in DNA-only extractions but which cannot be included in a simultaneous extraction protocol because the heat and enzymatic activity would degrade RNA and protein. Twelve frozen endometrium tissue blocks were cut into two with a sterile scalpel whilst remaining frozen, so two DNA extraction protocols could be compared.

The blocks were homogenized (once), centrifuged, and the supernatant applied to the AllPrep DNA/RNA/protein extractions as described above. After the supernatant had been collected for the simultaneous extractions, the tubes containing any insoluble, post-homogenized tissue debris and the homogenization beads were placed in storage at −80C. DNA was later extracted from them either using silica spin columns or by precipitation [QIAamp DNA Mini kit and Gentra Puregene kit (both Qiagen) respectively].

For the QIAamp extractions, the tubes containing the tissue debris were thawed and then digested over 24 hr in 180-µl ATL lysis buffer containing 40 µl of the proteinase K provided in the kit (20 µl added initially and an additional 20 µl after 7 hr) in a Thermomixer (Eppendorf) running at 300 rpm and 56C. DNA was then extracted according to the kit’s protocol, with the RNase digest and elution in 200-µl AE buffer. For the Puregene extractions, the tissue debris were heated in the thermomixer at 65C and 300 rpm for 30 min in 750 µl of the kit’s Cell Lysis Solution. The temperature was then reduced to 55C and the proteinase K digest performed for 24 hr with 15 µl (7.5 µl added immediately and another 7.5 µl after 7 hr). The DNA extraction was as per the kit’s protocol, with an RNase digest and a step in which protein is removed by salt precipitation. The DNA was precipitated using isopropanol and then solubilized by heating at 65C/300 rpm for 1 hr and then 21C/300 rpm for 3 hr in 100 µl of the kit’s Rehydration Solution (TE buffer).

The elution volumes of 200 and 100 µl for the QIAamp and Puregene kits, respectively, were selected because these are manufacturer-specified. DNA concentrations were converted to yields for comparison purposes. DNA was quantified and its integrity assessed using the same methods as the previous multiple-homogenization study.

Validation Cohort

Validation was performed using 65 frozen endometrium biospecimens that had been stored at −80C. The tissue had been collected for a cancer research project to which the nucleic acid from the initial AllPrep DNA/RNA extractions was issued. This project stipulated that quantification be performed by OD260 nm spectrophotometry only, purity assessed by OD260:280 nm, and that an RNase digest would not be performed. Frozen cores of 3-mm diameter were cut from the frozen tissue blocks using a CryoXtract CXT353 (CryoXtract Instruments Inc.; Boston, MA). The cores were not individually weighed before extraction, but we know from experience that their weight would typically be 10–20 mg. ⁵ The centrifuge tube, containing the homogenization bead and any tissue debris, was, instead of being discarded, frozen at −80C. The tubes were retrieved 3–7 months later and DNA extracted from the tissue debris using either the QIAamp (n=33) or Puregene (n=32) kit as described above.

Statistics

Data were analyzed using SigmaPlot v. 12.5 (Systat Software Inc.; Chicago, IL). Normality was tested using the Shapiro–Wilk test. Where repeated homogenizations had been performed on the same tissue blocks, data were analyzed using one-way repeated-measures ANOVA (using ranks when appropriate), followed by the Tukey test for pairwise comparisons. Correlation was tested using Pearson’s or Spearman’s correlation. Comparisons between AllPrep extractions and subsequent DNA-only extractions were tested using the paired t-test or Wilcoxon signed-rank test when the extractions were carried out on the same blocks; otherwise, comparisons between two groups were compared using Student’s t-test or the Mann–Whitney rank-sum test.

Results

Repeated Homogenizations of Tissue Block Debris and Then Additional Multi-analyte Extractions

Mean RNA yield was 0.74 µg per block (range, 0.05–2.46 µg) in homogenization 1 and the RNA returned a mean RIN of 5.9 (SD 1.9), indicating partially degraded RNA. No tissue block yielded any additional RNA in any subsequent homogenization, so no further analyses of the impact of repeated homogenizations on RNA were performed. It is known, however, that the AllPrep simultaneous extraction kit we used is suboptimal for purification of small species of RNA such as microRNA, so these have fallen outside of the scope of this study.

There was no correlation between block weight and yield in DNA, RNA, or protein extractions (n=11). In addition, there was no tendency for individual blocks that were high- or low-yielding in respect of one analyte to be the same in respect of another, regardless of whether we used data from homogenization 1 only or the sum of homogenizations 1–4. However, we also analyzed historic data from extractions from the same wider cohort of tissues collected by the same biobanks, but where just one homogenization had been performed. With this expanded sample size, a limited positive correlation between DNA and RNA yields from the same blocks was evident in conjunction with a poor correlation between their DINs and RINs (R² = 0.58 and 0.32, respectively; p<0.001, n=132). Protein yields did not correlate with either DNA or RNA yields from the same blocks, however (R² = 0.07 and 0.08, respectively; n=87).

Mean DNA yield (assayed using pico green) was 5.72 µg per tissue block (range, 0.88–13.70 µg) in homogenization 1, and in 6 of the 11 blocks, no additional DNA was extracted in any subsequent homogenization (Fig. 1). The cell pellet–positive controls yielded the expected quantity of nucleic acid, so we are confident that any failure to extract nucleic acid was not a consequence of a technical problem or user error. In the five blocks that yielded DNA after homogenization 1, additional DNA was recovered in all three subsequent homogenizations. However, the DNA yield in homogenizations 2, 3, and 4 combined was always lower than that from homogenization 1 (mean, 1.03 compared with 5.72 µg DNA, respectively; p=0.01). With one exception, additional DNA was only recovered when the DNA yield in the initial homogenization was above average (>7 µg). Additional DNA was obtained from all three blocks that exceeded the kit manufacturer’s maximum weight threshold of 30 mg.

Figure 1.

DNA yield from tissue blocks that underwent four sequential homogenizations (Hom 1–4), each followed by an AllPrep simultaneous DNA/RNA/protein extraction.

There was evidence that repeated homogenizations of the tissue blocks caused DNA degradation because the % dsDNA, DIN, and the percentage of blocks returning 15-kbp amplicons all declined when more than one homogenization had been applied (p<0.05): mean percentage DNA values were 72 %, 41%, 25%, and 30%, and mean DINs were 6.5, 5.4, 4.1, and 5.8 after homogenizations 1–4, respectively. In the long-range PCR assay (which was only performed for homogenizations 2, 3, and 4), the 15-kbp amplicon was present in 86%, 71%, and 57% of blocks after homogenizations 2, 3, and 4 respectively.

Performing additional homogenizations always resulted in additional protein being recovered (Fig. 2A). Homogenization 1 returned higher yields than homogenizations 2 and 3 (mean, 574, 329, and 150 µg protein, respectively; p<0.03), but the difference between homogenizations 1 and 4 fell slightly short of statistical significance (mean, 574 and 264 µg protein, respectively; p=0.06). There was no correlation in protein yield between the initial homogenization and the second one, between the initial homogenization and all three subsequent homogenizations added together, or between total protein yield and block weight (R² = 0.05, 0.08 and 0.0, respectively; p>0.05). The band patterns following SDS-PAGE (only performed in respect of homogenizations 2, 3, and 4) were visually very similar, so it was easy to match bands between the different homogenizations on adjacent lanes of the gel. One-dimensional electrophoresis inherently lacks resolution, and we found the entire lane of the gel to contain protein, with the software detecting bands only where proteins of high abundance had stained more intensely than the background protein. Our analysis was therefore restricted to these highly abundant protein bands (n=15–26 per tissue block, median = 22), and we have to make the assumption that the level of background protein staining would have been consistent in each band detected.

Figure 2.

Yield and consistency in protein extracted from tissue blocks after four sequential homogenizations (Hom 1–4), each followed by a simultaneous DNA/RNA/protein extraction. (A) Protein yield (n=11). (B) Commonality in protein bands detected following SDS-PAGE when tissue blocks (n=6) underwent three sequential homogenizations (Hom 2–4). (C) The magnitude of change in volume (either increasing or decreasing) of the protein bands of Fig. 2B following Hom 3 and 4 compared with Hom 2. Data are mean number of bands per tissue block. (D) Gel image of the banding pattern of a tissue block following Hom 2, 3, and 4. The black arrow denotes a band detected in Hom 2 and 3 only, the gray arrow a band with volume increases of 13% between Hom 2 and 3 and then 15% between Hom 2 and 4, and the white arrow a band with a volume decrease of 34% between Hom 2 and 3 and then a volume increase of 28% between Hom 2 and 4.

A total of 274 bands were identified and analyzed for the six tissue blocks applied to SDS-PAGE. We numbered the bands such that each band number represented a band of distinct molecular mass that was unique to a tissue block, but which could be present in more than one homogenization of that tissue block. There were a total of 127 bands numbered [mean, 17.5 per block (range, 14–22)], of which 5.5 % were only detected in homogenization 2, 5.5% only in homogenization 3, and 3.1% only in homogenization 4. Most were present in all three homogenizations (73 % of the total, n=93), and the remainder were in two homogenizations only (Fig. 2B). For bands present in more than one homogenization, there was a greater similarity in band volume between homogenizations 2 and 3 than between homogenizations 2 and 4 (median coefficient of variation (CV)s were 5.8% and 8.3%, respectively; p=0.04). Generally, the magnitude of the percentage change in band volume was limited, but it increased with ongoing homogenization. Between homogenizations 2 and 3, 62.9% (n=66) of the total number of bands changed <10% in their volume (mean 11 bands per block) and 85.7% changed <20% (mean 16 bands per block), but between homogenizations 2 and 4, 23.2% of the total number of bands (mean, 7.2 bands per block) changed <10% in volume and 44.1% changed <20% (Fig. 2C).

Approximately equal numbers of bands increased and decreased in volume with ongoing homogenizations (43% of the bands decreased in volume between homogenizations 2 and 3, and 50% between homogenizations 2 and 4). Using absolute values, the change in band volume between homogenizations 2 and 3 was statistically significantly lower than that between homogenizations 2 and 4 (median = 7.3% and 12.0%, respectively; p<0.001). There was a high degree of commonality in the band patterning between different tissue blocks, and we could not see a pattern whereby a particular band was more susceptible to homogenization-induced changes across the different tissue blocks. There was no relationship between tissue block weight and either the number of gel bands identified or the magnitude of the change in band staining intensity with ongoing homogenization.

To summarize, performing multiple homogenizations on a tissue block using the AllPrep kit to gain more yield resulted in no additional RNA being recovered. Although additional DNA was recovered, this was only from larger tissue blocks and the DNA became progressively more degraded with each additional homogenization. Meaningful quantities of protein could be recovered with additional homogenizations, but the quantity of protein progressively declined with additional homogenizations, accompanied with limited changes in proteomic profile.

Digesting Tissue Block Debris to Yield Additional DNA

The 12 tissue blocks used for this experiment had been cut into two pieces, so the AllPrep/QIAamp and AllPrep/Puregene extractions could be performed using the same starting material. In the initial AllPrep extractions, there were no statistically significant differences between the paired blocks in respect of DNA yield (p=0.40), 260:280 (p=0.21), or DIN (p=0.92). However, the DNA from the AllPrep extractions where QIAamp was subsequently to be applied had slightly higher % dsDNA values than their paired counterparts destined for Puregene extractions (mean, 87% compared with 82%, respectively; p=0.04). With the exception of % dsDNA values, we are therefore confident that any differences in the downstream DNA-only extractions were not consequentially influenced by differences in the starting material.

In the AllPrep/QIAamp combination, median DNA yields (assessed by pico green fluorometry) were 5.5 µg of DNA in the initial AllPrep extractions and 2.6 µg in the subsequent QIAamp extractions, with 11 of the 12 QIAamp extractions delivering additional DNA (Fig. 3A). The block that failed to deliver any additional DNA in the QIAamp extractions was the lightest prior to its initial homogenization (6 mg), with the next lightest (7 mg) only delivering a very low yield in the QIAamp extraction (0.27 µg). Overall, the QIAamp extractions yielded a median 27.1% additional DNA compared with the preceding AllPrep extraction. Switching from QIAamp to Puregene for the DNA-only extraction also resulted in 11 extractions returning DNA, and although the block that failed to yield additional DNA had a different parent block to the one that failed to yield DNA in the QIAamp extractions, it was again the lightest one prior to its initial homogenization (4 mg). The median additional DNA yielded in the Puregene extractions compared with their preceding AllPrep extractions was 203.9%, which was a statistically significantly higher additional yield than QIAamp (p=0.032) (Fig. 3B). For neither QIAamp nor Puregene extractions was there any correlation between the yield in the initial AllPrep extraction and that in the subsequent DNA-only extraction (R² = 0.15 and 0.25, respectively; p>0.09).

Figure 3.

DNA yield from tissue blocks that were homogenized for AllPrep simultaneous DNA/RNA/protein extractions, followed by DNA-only extractions from the tissue block debris. The DNA-only extractions were performed using either the QIAamp kit (A) or the Puregene kit (B), after the tissue debris had undergone a proteinase K digest.

The median 260:280 purity ratios were within the range that denotes pure DNA in all three extraction methods: 1.90 (AllPrep), 1.94 (QIAamp), and 1.85 (Puregene) (Fig. 4A). Although small in magnitude, the differences between AllPrep and Puregene 260:280 were statistically significant (p=0.003), with Puregene showing a greater range of values that fell below optimal (a 260:280 of 1.8). The reason for lower purity ratios in Puregene extractions is likely because contaminant removal by pipetting and evaporation of the wash solutions in the precipitation-based Puregene protocol is less effective than by centrifugation in the silica spin-column AllPrep and QIAamp protocols.

Figure 4.

Purity and integrity of DNA extracted from homogenized tissue blocks using the AllPrep simultaneous DNA/RNA/protein extraction kit compared with that extracted from the post-homogenized tissue debris using either the QIAamp or Puregene DNA-only kits. (A) 260:280 purity ratio, (B) DIN, and (C) % dsDNA. Data from the two cohorts of AllPrep extractions were very similar, so were pooled for the purposes of brevity in the Figure, but analyses were performed using non-pooled data. The boxes extend through the second and third quartiles, intersected with the median, and the whiskers extend to the 10th and 90th percentile and the dots are the remaining data points. * denotes statistical significance. (D) Overlaid electropherograms from TapeStation DIN analysis for two representative samples: the peak at 100 bp is the lower marker, the center peak is a DNA sample of 233 ng/µl with DIN 6.2 and maximum fluorescence assigned at 4640 bp, and the right peak is a DNA sample of 152 ng/µl with DIN of 9.1 and maximum fluorescence assigned at >60,000 bp. Abbreviation: DIN, DNA Integrity Number; dsDNA, double-stranded DNA.

DINs were higher in the DNA-only extractions compared with their preceding AllPrep extractions, increasing from 6.8 to 7.6 for the AllPrep/QIAamp extractions and from 7.2 to 9.3 for AllPrep/Puregene extractions (Fig. 4B). The difference between AllPrep and QIAamp was not statistically significant, but the higher DINs in Puregene compared with both AllPrep and QIAamp were statistically significant (p<0.01). Analysis of the size (in kbp) of the summit of the peak (i.e., maximum fluorescence) in the electropherograms showed a similar pattern, with the median maximum peak being at 12.0 kbp for AllPrep extractions, 22.7 kbp for QIAamp extractions and “>60.0 kbp” (i.e., above the maximum quantification size of the software) for Puregene extractions. As in the DIN, the differences between AllPrep and QIAamp were not statistically significant but between Puregene and both AllPrep and QIAamp were (p≤0.02). We speculate that the reason why DINs and maximum peak sizes were lower in AllPrep than the subsequent DNA-only extractions was because in the former the DNA became slightly fragmented during homogenization by the action of bead, while in the DNA-only extractions the DNA had remained intracellular in the tissue debris, so protected during homogenization, to be liberated during the subsequent proteinase K digests. The reason DNA had higher DIN and maximum peak sizes in Puregene extractions was likely because DNA of very high integrity is difficult to elute from silica spin columns while DNA of very low integrity is difficult to precipitate, meaning that, compared with QIAamp, Puregene extractions are enriched in DNA of high integrity but depleted in DNA of low integrity.

Percent dsDNA values in both DNA-only extractions were lower than in their preceding AllPrep extractions, although the magnitude of the difference was much more pronounced when QIAamp rather than Puregene was used (Fig. 4C). In the QIAamp extractions, median % dsDNA was 43.4%, reduced from 85.8% in the preceding AllPrep extractions (p<0.001). In Puregene, however, the equivalent values were 83.5% and 76.6%, respectively (p=0.004), despite the AllPrep extractions destined for Puregene DNA-only extractions having a slightly lower % dsDNA values than the AllPrep extractions destined for QIAamp extractions (see initial paragraph). Given that all extractions incorporated an RNase digest, RNA cannot account for the difference. We think these results are a consequence of the higher yields of dsDNA in Puregene compared with QIAamp in combination with a tendency for single-stranded DNA to being less amenable to precipitation than purification on a silica spin column.

Long-range PCR was performed for all extractions, except those that had failed to generate DNA. In all extractions, the 15-kbp amplicon was generated, with the exception of one of the blocks, in which the amplicon was generated in both the QIAamp and Puregene extractions but not in either preceding AllPrep extraction. While this block was of average weight, the DNA was low in DIN compared with the cohort of tissue blocks as a whole, with a greater disparity in DIN between the AllPrep extractions (DIN of 2.6 and 2.3) and the DNA-only extractions (DIN of 4.3 and 4.9).

To summarize, performing an additional DNA-only extraction incorporating a proteinase K digest of the homogenized tissue debris from an initial AllPrep extraction was considerably more successful in delivering additional DNA than performing additional homogenizations, yielding an average 27% or 204% additional DNA, depending on the DNA extraction method selected. DNA purity or integrity was not compromised in the subsequent DNA-only extractions when assessed using DIN or long-range PCR, but the % dsDNA values were lower. Using the precipitation-based Puregene method for the DNA-only extractions was optimal, returning higher yields, DINs, and % dsDNA values than the silica spin column–based QIAamp method. However, QIAamp was more user-friendly and reproducible, requiring less technician time than the Puregene method, and could, unlike Puregene, be automated.

Validation Cohort

We validated the above method using a much larger cohort of 65 endometrium tissues: initially AllPrep simultaneous DNA/RNA extractions were performed, then QIAamp or Puregene DNA-only extractions on the tissue debris. As before, Puregene returned higher yields than QIAamp, but unlike in the initial experiment, the validation cohort yielded additional DNA from every block. When pico green was used for quantification, the median dsDNA yields extracted from the post-homogenized tissue fragments of the AllPrep extractions were 6.06 µg per block for QIAamp (n=33) and 7.77 µg for Puregene (n=32), with the difference between the two methods falling well short of statistical significance (p=0.93) on account of the magnitude of variability (SDs were 15.1 and 13.6, respectively).

In the validation cohort, DNA quantification of the initial AllPrep extractions was by OD260 nm spectrophotometry only, and using this method of quantification, the median percentage additional DNA yields gained per block by performing QIAamp and Puregene extractions on tissue fragments from AllPrep extractions were 31.6% and 54.8%, respectively (Fig. 5A). However, the percentage additional DNA yield calculations will be underestimated because the DNA yields in AllPrep would have been overestimated to an unknown extent because coeluting RNA will also absorb light at OD260 nm (QIAamp and Puregene both included RNase digests). We evaluated the phenomenon of RNA coeluting with DNA in a previous study (using the QIAamp extraction kit) and found that coeluting RNA accounted for 28–52% of the spectrophotometry-derived “DNA” concentration, depending on tissue type (but endometrium was not one of the tissue types evaluated). ⁶ The quantity of coeluting RNA in an AllPrep extraction is likely to be lower, however, because ethanol is not added to the lysis buffer at the point in which it is centrifuged through the AllPrep DNA column, and we have previously evaluated it at 16.0%, although with a small cohort of eight tissue samples, none of them endometrium (data not shown).

Figure 5.

DNA yield and integrity in the validation cohort. (A) DNA yield from AllPrep extractions from homogenized tissue blocks, followed by either QIAamp (n=33) or Puregene (n=32) extractions from proteinase K–digested, residual tissue debris. Data are presented in ascending order of DNA yield in the initial AllPrep extraction. (B) DINs in the QIAamp and Puregene extractions. (C) Size (in kbp) at the point in the electropherograms at peak fluorescence (i.e., at the greatest abundance of DNA). The parameters of the box plots are as in Fig. 4. The differences between QIAamp and Puregene were not statistically significant in either B or C. Abbreviation: DIN, DNA Integrity Number.

The percentage increase in yield was more consistent for QIAamp compared with Puregene (CV = 85.8% compared with 211.4%), potentially because solubilizing DNA from the precipitated pellet in Puregene extractions over 4 hr was more variable than from the silica spin column in a 1-min incubation in QIAamp extractions. AllPrep yields and subsequent QIAamp or Puregene yields did not correlate (R² = 0.18 and 0.05, respectively), and there was no indication that yield or DIN was compromised when the period of time that the tissue debris had been stored at −80C was longer (storage time was 3–8 months). AllPrep yields in the cohort of blocks used for QIAamp extractions were higher than in the cohort of blocks used for Puregene extractions, but not statistically significantly so (median = 36.0 µg compared with 25.5 µg, respectively; p=0.13).

Given that a user’s requirement to undertake an additional extraction will depend on the minimum yield required for the downstream analytical platform, Table 1 presents a breakdown of the additional DNA yields obtained using QIAamp and Puregene at different initial yield thresholds from AllPrep extractions.

Table 1.

The Mean Additional DNA Yield Obtained From 65 Endometrium Tissue Blocks After Applying a QIAamp or Puregene Extraction to the Post-homogenized Tissue Debris of an Initial AllPrep DNA/RNA Extraction.

Initial AllPrep Yield (µg) by Photometry	Subsequent QIAamp Extraction			Subsequent Puregene Extraction
	n	Mean Additional Yield (µg) by Photometry	Mean Additional Yield (µg) by Fluorometry	n	Additional Yield (µg) by Photometry	Additional Yield (µg) by Fluorometry
<2.5	0	N/A	N/A	2	0.23 (12.3%)	0.23 (10.2%)
2.5–5	0	N/A	N/A	0	N/A	N/A
5–10	2	8.0 (9.2%)	6.4 (7.9%)	2	15.4 (120.3%)	10.7 (139.3%)
10–20	9	5.5 (85.3%)	3.5 (85.0%)	10	25.1 (91.2%)	14.9 (80.0%)
20–30	3	8.1 (476.4%)	5.8 (219.0%)	4	21.0 (68.8%)	13.5 (91.3%)
30–40	5	8.0 (46.5%)	6.0 (48.5%)	3	8.9 (80.4%)	4.1 (94.0%)
40–50	4	10.8 (68.0%)	7.6 (53.5%)	7	18.2 (92.1%)	7.2 (96.7%)
>50	10	24.1 (120.4%)	20.5 (120.0%)	4	68.4 (135.0%)	22.1 (127.2%)

The percentages in parentheses are CV.

DINs were not obtained for the AllPrep extractions, but for the subsequent DNA-only extractions, unlike what we found in the initial experiment, DINs were similar for QIAamp and Puregene (median = 7.7 and 7.2, respectively). There was more variability in DIN from Puregene, however (Fig. 5B). The median maximum peak of fluorescence in the electropherograms (assigned to each electropherogram by the software) was at 26.7 kbp for QIAamp and 23.8 kbp for Puregene (Fig. 5C). The differences between QIAamp and Puregene fell well short of statistical significance (p=0.71), but as in DIN, Puregene had a greater range in the size at which the peak of the electropherogram occurred (Fig. 5C). As before, QIAamp extractions were enriched in dsDNA (median = 74.4 % dsDNA for QIAamp and 52.5 for Puregene, p=0.002).

Discussion

We demonstrate that in a multi-analyte extraction, the tissue debris following homogenization can be frozen in its centrifuge tube (rather than being discarded) and then digested by proteinase K to yield additional DNA of similar integrity to that found in the initial extraction. Additional protein can be obtained by adding fresh buffer and repeating the mechanical homogenization. Yields in the initial and subsequent extractions did not correlate for either the DNA or the protein, so we could not predict what additional yield would be gained by performing the second extraction from the results of the first extraction.

The lack of correlation between DNA, RNA, and protein yields from the same blocks is likely a consequence of both technical and biological factors. Yields will be partially dependent on the number of cells ruptured, but many other factors are salient. For DNA, yields will be compromised because a proteinase K digest cannot be included in the AllPrep extraction protocol. Additional homogenizations failed to extract additional DNA from blocks <7 mg (with one exception), suggesting that proteinase K might be more critical in DNA extractions for larger blocks. It is logically easier to liberate RNA from tissue than DNA during homogenization because most RNA resides in the cytoplasm, so only the cell membrane needs to be ruptured. RNA is also less likely to become entangled in the silica matrix of the extraction columns than genomic DNA on account of its smaller size. However, RNA can bind to the upstream DNA extraction column and be lost. Protein yield will be reduced if clogging or binding in the upstream DNA and RNA purification columns occurs, and the precipitation/resolubilization steps will not be entirely efficient. Analyte yields will also depend on biological factors such as transcription/translation rates, concentration of ribosomes per cell, RNA and protein degradation rates, percentage tumor, differences in tumor microenvironment, and necrosis.

Repeated mechanical homogenization in fresh buffer followed by additional multi-analyte extractions yielded no additional RNA, and the limited quantity of additional DNA obtained became increasingly fragmented with each additional homogenization. The phenomenon of excessive homogenization causing DNA fragmentation is known, and indeed the kit manufacturer explains this in the kit’s handbook. ⁷ Therefore, we do not recommend this approach and instead promote a dedicated DNA-only or protein-only extraction (the latter directly in the buffer of choice for the downstream proteomic platform). It would be straightforward for users to freeze all the centrifuge tubes in a cohort of extractions and then just revisit those where the initial DNA or protein yield proved to be insufficient. We found no decline in DNA yield or integrity when tissue debris were stored in their centrifuge tubes at −80C for 8 months. It is possible, however, that recording the return of biospecimen debris to storage in addition to the production of nucleic acid and protein aliquots from the same initial tissue block, followed by the production of additional DNA aliquots from said debris, might be challenging for some Laboratory Information Management Systems.

Our finding that even a second homogenization followed by a multi-analyte extraction failed to yield additional nucleic acid uncompromised in integrity demonstrates, in our opinion, that the homogenization protocol we applied (in respect of time, velocity, and bead size) had been effectively optimized. Indeed, we see the value in performing additional homogenizations and extractions when optimizing a multi-analyte extraction protocol in respect of a particular tissue type: the user will know that the homogenization protocol is optimal in a simultaneous extraction in respect of DNA when additional homogenizations cause the DNA to decline in integrity. Laboratories that have not undertaken optimization of the homogenization settings may find greater yield increases than we did in a second extraction from post-homogenized tissue debris because inefficiency in homogenization might equate to a greater proportion of the total nucleic acid and protein remaining in the tissue debris.

There are several multi-analyte extraction kits commercially available from different manufacturers, and they have different strengths and weaknesses. For example, the AllPrep kit from Qiagen that we used has been shown to return higher RNA yields but lower DNA yields than the equivalent TriplePrep kit from GE Healthcare. ³ It is possible, therefore, that if we had selected the TriplePrep kit for our study, we would have extracted some RNA but less DNA from the post-homogenized tissue debris. We do not make the assumption that the results we obtained in this study would apply in respect of other kits.

Additional homogenizations yielded meaningful quantities of protein, but there were inconsistencies within our cohort of tissue blocks as to the magnitude of the additional yield and also whether yields declined with each subsequent homogenization. When visualized following SDS-PAGE, >90% of the bands were present in all homogenizations, but there were clear differences in band staining intensity with ongoing homogenization. These differences could be a consequence of heat-induced degradation, differences in the proteome in harder-to-lyze parts of the tissue, or due to inconsistencies in the downstream sample manipulations, including the protein precipitation and resolubilization steps, rather than in the homogenizations themselves. Although it is also possible that proteases had additional time to be active during ongoing homogenizations, the RLT homogenization buffer contains the protein denaturant guanidine-isothiocyanate and the reducing agent DTT. ⁷ Nevertheless, adding protease inhibitors could be beneficial.

The impact of tissue homogenization on the proteome has not been widely studied, but Wang et al. ⁸ found the ratio and electrophoresis banding pattern of phosphorylated and unphosphorylated myosin light chain 20 protein in trachea muscle tissue changing over four sequential homogenizations/extractions, and Ryan and Walker ⁹ demonstrated that repeated homogenizations of breast tissue rapidly caused estrogen and progesterone receptors to degrade.

The electrophoresis banding pattern we observed would also depend on the efficiency of the protein precipitation/resolubilization process that is an inherent part of the simultaneous extraction protocol. A previous evaluation found the protein precipitation/resolubilization process of the AllPrep kit only yielded 20% of the protein quantity of a direct homogenization into buffer, with 11% of the proteome being more than 2-fold altered in staining intensity and a tendency for the depletion of acidic proteins. ³ It is unlikely that this 5-fold reduction in yield is entirely due to the precipitation and resolubilization process itself, however, because previous studies that have evaluated protein precipitation and resolubilization have found the reduction in yield to be in the region of 20%.^10,11 It is likely that protein is also lost during centrifugation through DNA-binding and RNA-binding columns, and that the chemistry of the lysis and resolubilization buffers might prevent highly efficient protein precipitation and resolubilization.

It is important to acknowledge that we did not separately assess the reproducibility of the protein precipitation and resolubilization step independently of the repeated homogenizations, nor were we able to compare the proteome following the initial homogenization with the subsequent homogenizations (the former was required for a separate project). Furthermore, the electrophoresis we performed lacks resolution, with each protein band we analyzed containing many different proteins. Consequently, all we can conclude is that performing subsequent homogenizations and extractions using the AllPrep kit and then solubilizing the precipitated protein in the buffer we selected result in changes in the proteome. We therefore advise users who need to recover additional protein following a simultaneous extraction to perform a direct extraction into buffer, thereby avoiding precipitation and resolubilization entirely, but to be aware that whatever approach they adopt, the proteome of the second extraction will very likely differ to some extent from that of the initial extraction.

For the DNA-only extractions from the tissue debris, the Puregene kit (DNA precipitation) returned higher yield, DINs, and % dsDNA values than the QIAamp kit (silica spin columns). We do not think the lower yield in the spin columns was because they had become saturated with DNA, because the quantity of DNA recovered was always much lower than the 100-µg binding capacity of the columns. ⁷ The mechanisms that drive the ability of DNA to be eluted from silica are complex and involve an interplay between temperature, pH, and ionic strength, but it is known that the structure of the DNA is also important.^12,13 It is therefore possible that dsDNA of very high integrity might bind to the silica particles of the spin column with an intensity that compromises elution efficiency or that long strands of DNA might become entangled in the silica matrix and be lost to the extraction, or become sheared during centrifugation.

We do not think that the lower DNA integrity in the spin columns is consequential, however: the electropherograms generated by the TapeStation had a median peak of 26.7 kbp (consistent with the manufacturer’s statement that eluted DNA will be predominantly 20–30 kbp) and all extractions returned DNA peaking at >13.5 kbp. ¹⁴ Long-range PCR demonstrated that 15-kbp amplicons could be generated from the DNA extracts, indicating amenability to sequencing.^15,16 Improved yields and higher DNA integrity with Puregene compared with QIAamp have both been demonstrated before.^17–19 The difference between the two kits in respect of DNA yield was extensive, both in the initial study using 11 blocks per method and in the validation cohort. However, the Puregene method was more labor-intensive, more time-consuming, and less consistent in its results than QIAamp, and unlike QIAamp, it is not amenable to automation.

We selected the Puregene and QIAamp kits for this study because they represent two fundamentally different methods of DNA extraction, and both are commonly used, but we see no reason why users cannot apply alternative DNA extraction methods appropriate to their workflow, provided the method incorporates a protease K digestion as an initial step. Tissue stabilized in media such as RNAlater, Allprotect, or PAXgene Tissue Fixative are also potentially amenable to the method we describe.

Protease K enhances DNA extraction efficiency by degrading DNases and protein (including histones). Proteinase K has been the protease of choice for decades on account of its relatively high levels of activity, its broad substrate specificity, and because it is resilient to the denaturants that are present in lysis buffers. ²⁰ DNA-only extraction protocols will typically include a homogenization step followed by a proteinase K digest, so it is not surprising that excluding the latter (a necessary restriction in a multi-analyte extraction) compromises DNA extraction efficiency, and therefore applying the proteinase K digest to the tissue debris would yield additional DNA. It has been shown using several biospecimen types that optimization of the proteinase K digest (in terms of quantity of protease, digest length, and digest temperature) can be instrumental in increasing DNA yield.^21–23 However, other proteases such as papain and bromelain have also been used and shown to be superior to proteinase K in some biospecimen types. ²⁴ Therefore, it is likely that performing additional optimization in respect of their tissue types and extraction kits would enable other labs to further improve the DNA yield that can be obtained from the residual tissue debris following a multi-analyte extraction.