“Out-Gel” Tryptic Digestion Procedure for ... - American Chemical Society

Article pubs.acs.org/jpr

“Out-Gel” Tryptic Digestion Procedure for Chemical Cross-Linking Studies with Mass Spectrometric Detection Evgeniy V. Petrotchenko,†,∥ Jason J. Serpa,†,∥ Ashley N. Cabecinha,† Mary Lesperance,‡ and Christoph H. Borchers†,§,* †

University of Victoria, University of VictoriaGenome British Columbia Protein Centre, 3101-4464 Markham Street, Vancouver Island Technology Park, Victoria, BC V8Z 7X8 Canada ‡ University of Victoria, Department of Mathematics and Statistics Social Sciences and Mathematics Building Room A425, 3800 Finnerty Road (Ring Road), Victoria, BC V8P 5C2 Canada § University of Victoria, Department of Biochemistry & Microbiology, Petch Building Room 207, 3800 Finnerty Road, Victoria, BC V8P 5C2 Canada S Supporting Information *

ABSTRACT: SDS-PAGE is one of the most powerful protein separation techniques, and in-gel digestion is the leading method for converting proteins separated by SDS-PAGE into peptides suitable for mass spectrometry-based proteomic studies. In in-gel digestion, proteins are digested within the gel matrix, and the resulting peptides are extracted into an appropriate buffer. Transfer of the digested peptides to the liquid phase for subsequent mass spectrometric analysis, however, may be hampered by peptidespecific characteristics, including size, shape, poor solubility, adsorption to the polyacrylamide, andin the case of cross-linking applicationsby the branched structure of the peptides produced. This can be a limitation in cross-linking studies where efficient recoveries of the cross-linked peptides are critical. To overcome this limitation, we have developed a modification to the standard in-gel digestion procedure for SDS-PAGE-separated cross-linked proteins, based on older passive diffusion methods. By omitting the gel staining and gel fixation steps, intact proteins or crosslinked protein complexes can move through the gel and into the buffer solution. Digestion of the entire protein in the buffer outside the gel increases the probability that most of the proteolytic peptides produced will be present in the final digest solution. The resulting peptide mixture is then freed of SDS and concentrated using SCX (strong cation exchange) zip-tips and analyzed by mass spectrometry. For standard protein identification studies and the recovery of noncross-linked peptides, the in-gel procedure outperformed the out-gel procedure, but for cross-linking studies with enrichable cross-linkers (such as CBDPS), the standard out-gel procedure allowed the recoveries of cross-links not recovered via the in-gel method. Labeling experiments showed that, with an enrichable cross-linker, 93% of the cross-links showed better or equal recoveries with the out-gel procedure, as compared to the in-gel procedure. It should be noted that this method is not designed to replace in-gel digestion for most proteomics applications. However, by using the out-gel method, we were able to detect twice as many interprotein CBDPS crosslinks from the histone H2A/H2B complex as were found in the in-gel digested sample. KEYWORDS: proteins, SDS-PAGE, protein complexes, extraction and digestion, cross-linking

■

INTRODUCTION

reduced solubilities and the adsorption properties of the crosslinked proteins, which also makes SDS-PAGE separation the method of choice. The techniques used for preparation of gel-separated proteins for mass spectrometric analysis include methods for the extraction of intact proteins as well as methods for in-gel digestion and extraction of the resulting peptides. Various combinations of fixing, staining, destaining, passive diffusion/ passive elution,4−6 electroelution,7−9 and blotting for the

The analysis of proteins after gel separation has been performed for the past several decades, using a variety of techniques (for reviews of specific techniques, see refs 1 and 2; for an excellent overview, see 3). An advantage of SDS-PAGE separation for cross-linking analyses is the ability to focus the analysis on specific cross-linked protein bands, allowing one to direct the search toward cross-links of a particular product of interestfor example, only the interprotein cross-linked heterodimer without interference from interprotein cross-linked homodimers or intraprotein cross-linked monomers. Moreover, LC separation of cross-linked protein complexes is usually hindered by the © 2013 American Chemical Society

Received: July 10, 2013 Published: December 19, 2013 527

dx.doi.org/10.1021/pr400710q | J. Proteome Res. 2014, 13, 527−535

Journal of Proteome Research

Article

separated by SDS-PAGE using 1D Novex gradient gels (Invitrogen, Carlsbad, CA). D-Succinimydyl suberate (DSS), an amine-reactive isotopically labeled cross-linker, and cyanurbiotindipropionylsuccinimide (CBDPS),22 an aminereactive isotopically labeled biotinylated CID-cleavable crosslinker (Figure 1), were obtained from Creative Molecules, Victoria, BC, Canada. All other materials were from SigmaAldrich, unless noted otherwise.

extraction of intact proteins, as well as digestion on the PVDF membranes,10 and in-gel digestion10−14 have been utilized. Of all of these techniques, in situ proteolysis of the SDS-PAGEseparated proteins is still the most commonly used technique for protein identification of gel-separated proteins.11,15 In situ proteolysis of proteins in nonstained gels was introduced in 1977,16 but high concentrations of SDS prevented efficient digestion of the proteins by trypsin. During the in-gel digestion process, a protein-containing gel band is fixed, excised, dehydrated, and soaked in a trypsin solution. These steps allow diffusion of the enzyme into the gel matrix and proteolysis of the proteins in situ, within the gel. The resulting peptides are extracted from the gel and are analyzed by mass spectrometry. The power of mass spectrometry for the analysis of peptides, combined with the use of SDS-PAGE for the separation of complex protein mixtures, led to the in-gel digestion method for mass spectrometric identification of proteins becoming one of the central proteomics techniques for the past 10 years.17 The in-gel digestion procedure does not usually result in complete sequence coverage, but the extracted peptides (mainly hydrophilic and shorter peptides) are usually sufficient for protein identification. In particular, there are problems with the detection of hydrophobic and larger peptides, both of which are characteristics of interpeptide cross-links. In our cross-linking studies, we have become concerned about the low degree of extraction of branched interpeptide cross-links after the in-gel digestion of cross-linked proteins. Our hypothesis was that performing the digestion of the crosslinked proteins in the solution phase rather than in the polyacrylamide gel matrix would lead to better recovery of the branched interpeptide cross-links. We have combined passive diffusion of the SDS-PAGE-separated proteins under optimized conditions with tryptic digestion of the cross-linked proteins. Although minute amounts of SDS promote efficient tryptic digestion of protein, to facilitate subsequent mass spectrometric analysis of the extracts from SDS-PAGE-separated proteins, SDS needs to be removed from solutions. This has been achieved by acetone precipitation of proteins,18 extraction with waterimmiscible organic solvents,19−21 or by removal of SDS during the in-gel digestion procedure.17 To improve the recovery of cross-links, the method we describe here involves diffusion of the intact protein out of the gel slice, followed by in-solution digestion. The resulting peptides are then concentrated and freed from SDS using SCX zip-tips. Three factors are critical for the development of an out-gel protein digestion procedure: (1) ensuring a high degree of diffusion of protein from the gel, (2) maintaining the activity of the trypsin in the presence of SDS, and (3) removal of SDS before MS analysis. These three factors were optimized during the development of the out-gel procedure. The final out-gel method was tested on cross-linked protein complexes, and a quantitative comparison of the effectiveness of the in-gel and outgel procedures was performed using isotopically labeled crosslinkers.

■

Figure 1. Structure of CBDPS-H8/D8. The features of the reagent which facilitate mass spectrometric analysis of the cross-links are indicated.

Instrumentation

All MALDI (matrix-assisted laser desorption/ionization) mass spectrometry experiments were performed on an Applied Biosystems (now AB Sciex, Concord, Ontario, Canada) 4800 MALDI-TOF/TOF mass spectrometer. All experiments for the recovery of nonmodified peptides were performed on a QStar Pulsar (AB Sciex, Concord, Ontario, Canada). Cross-Linking of Histones

Histones H2A and H2B were cross-linked with an isotopically labeled cross-linker prior to SDS-PAGE separation. Briefly, proteins were cross-linked in PBS for 30 min at 25 °C, the reaction was quenched with ammonium bicarbonate, and the reaction products were separated by SDS-PAGE using NuPAGE mini-gels with MES or MOPS buffer systems (Invitrogen), as described previously.23 The MES and MOPS buffer systems produced similar results. For CBDPS cross-linking of the histone heterodimer, 7 μL of a 1 mg/mL solution of each H2A and H2B histones were mixed in 21 μL of PBS and 7 μL 0.2 M Na2HPO4. A 25 mM stock solution of CBDPS in DMSO was diluted to 1 mM in water, and 2 μL of this solution was immediately added to the rest of the reaction mixture to give a final concentration of 0.05 mM CBDPS. The reaction mixture was incubated for 30 min at 25 °C and quenched with ammonium bicarbonate (20 mM final concentration). For single-spot analyses, the reaction mixture was split in half and separated by SDS-PAGE in an MES buffer system with a final load of 3.1 μg of each histone per lane. For the replicate LC-MALDI-MS analyses, the reaction mixture was doubled (14 μL of a 1 mg/mL solution of each H2A and H2B histone, 42 μL of PBS, and 14 μL 0.2 M Na2HPO4), and 2.1 μg of each histone per lane was used. Cross-Linking of HIV-RT

For each in-gel and out-gel analysis of HIV-RT dimer, subunits p66 (4 μg) and p51 (4 μg) were cross-linked with either 0.1 mM (final concentration) of CBDPS or 10 mM (final concentration) DSS, using the conditions described above.

EXPERIMENTAL SECTION

Materials and Chemicals

The HIV-RT (human immunodeficiency virus-reverse transcriptase) from Worthington Biochemical Corp. (Lakewood, NJ), plasminogen, human serum albumin (HSA), α1-macroglobulin, (Sigma-Aldrich, St. Louis, MO), and H2A and H2B histones (all from New England Biolabs, Ipswich, MA) were

Out-Gel Extraction and Digestion

In order to permit passive diffusion of the SDS-PAGE-separated proteins out of the gel, the gel was not fixed prior to the out-gel digestion. The 6 × 4 mm gel bands, which corresponded to cross528

dx.doi.org/10.1021/pr400710q | J. Proteome Res. 2014, 13, 527−535

Journal of Proteome Research

Article

Analysis of the Mass Spectra

linked histone dimers, were excised immediately after electrophoresis, immersed in 240 μL of 20 mM ammonium bicarbonate solution (a gel/solution ratio of 1:10 (v/v), and digested with 0.5 μg of trypsin using end-over-end mixing at 25 °C, overnight. Each protein digest was concentrated using a SCX zip-tip. For the analysis of the unfractionated digest in a single MALDI spot, 1.0 μL of the 5% ammonium hydroxide/30% methanol elution was spotted directly onto the MALDI plate. Protein digests to be analyzed by LC-MALDI were eluted with 10 μL of 5% ammonium hydroxide/30% methanol, neutralized with 50 μL of 2% acetic acid, and supplemented with 360 μL of 0.17 M ammonium acetate to give a final volume of 420 μL at pH 6.5. The CBDPS cross-links were affinity-purified using monomeric avidin beads (Thermo Scientific, Waltham, MA). The beads were washed with 100 mM ammonium acetate, 500 mM ammonium acetate, and then with water, and the affinity-bound material was eluted first with 100 μL of 0.1% TFA and then with 100 μL of 0.1% TFA/50% acetonitrile. The eluates were combined before mass spectrometric analysis.22 In-Gel Digestion. In-gel digestion was performed using our standard in-gel digestion protocol.24 Once the out-gel samples had been extracted from the gel, the remaining portion of the gel was fixed, stained with Coomassie, destained, and the stained protein bands were excised and subjected to standard in-gel digestion techniques.24 For the analysis of the unfractionated digest, each protein digest was concentrated using a C18 zip-tip, eluted with 1−2 μL 0.1%TFA/50%ACN directly onto the MALDI plate. Protein digests to be analyzed by LC-MALDI were affinity-purified as described above and then analyzed by LCMALDI as described below. LC-MALDI analysis was performed as described previously.25 Briefly, the combined eluates from the enrichment procedure were concentrated by lyophilization and were separated by nanoflow reversed-phase HPLC on a Ultimate nano-LC system (LC Packings, Sunnyvale, CA), equipped with an 0.3 × 5 mm C18 PepMap trap column (5 μm particle size, 100 Å pore size, LC Packings), and a 75 μm × 15 cm capillary column packed inhouse with Magic C18 Aq (Michrom Bioresources, Inc., Auburn, CA) particles (5 μm, 100 Å). This capillary LC system was operated at a flow rate of 300 nL/min, using a 55 min gradient from 5 to 60% acetonitrile (0.1% TFA). The column effluent was spotted at 1 min intervals (300 nL/spot) onto a stainless steel MALDI target using an AccuSpot NSM-1 spotter (Shimadzu, Kyoto, Japan). The spots were dried, overlaid with 0.7 μL of a 4 mg/mL solution of α-cyano-4-hydroxycinnamic acid in 0.1% TFA/50% acetonitrile, and analyzed by MALDI-MS and MS/ MS, using a 4800 MALDI-TOF/TOF (AB Sciex, Concord, ON, Canada). MS/MS spectra were acquired using “CID off”, 50 FHWM gate width, and a 1 kV MSMS method. Comparison of Out-gel and In-Gel ProtocolsQuantitative Study. For the comparative recovery experiments, histones H2A and H2B were cross-linked with either the light H8 or the heavy D8 isotopic forms of the CBDPS cross-linker. Each cross-linking reaction was split and processed in triplicate using either the in-gel or the out-gel procedure, giving a total of 12 lanes on an SDS-PAGE gel. Digests from the in-gel and outgel procedures were combined pairwise and analyzed by LCMALDI-MS, as described above for the out-gel procedure. The quantitative experiments were also performed with the labels reversed to be certain that there was no bias due to the labeling.

The MALDI spectra of the combined digests were analyzed using the DX and ICCLMSMS programs of the ICC-CLASS software suite for the detection of the isotopic signatures of the cross-links in the MS and MS/MS spectra, respectively.26 Mass spectra peak lists were generated using Data Explorer Version 4.0 (AB Sciex). Interpeptide cross-links were recognized on the basis of their CID cleavage patterns in the MS/MS spectra and were identified using the DXMSMS module of the ICC-CLASS software package, using the following settings: 100 ppm mass tolerance for the precursor mass; 300 ppm mass tolerance for MS/MS fragment ions; trypsin cleavage sites; and K as the allowed crosslinking site (all of these software programs are available as free downloads from www.creativemolecules.com). Data Analysis Methods for Comparing the Two Procedures

Method for the Qualitative Comparison. Digests from the in-gel and out-gel procedures were combined pairwise and analyzed by LC-MALDI-MS. Peak heights were obtained from the peak lists for each spectrum, as generated by the Data Explorer software. Peaks not detected by the software’s default parameters (valley-to-baseline selected for integration baseline settings; peak detection settings: centroid 50%, S/N threshold 3, noise window width 250 m/z, and threshold S/N after S/N recalculation 10), were designated as nondetectable (ND). Method for the Quantative Comparison. Direct comparison of the peak heights led to a “divide-by-zero” problem when the peak heights for nondetectible peaks were in the denominator. To circumvent this issue, the S/N values were used. In this case, the peaks designated as “ND” above can be assigned a S/N value of 1 (i.e., the peak height was the same as the background noise), and the ratio of the S/N ratios can be calculated. For automated determination of these signal-to-noise ratios (S/N), the Data Explorer peak detection parameters were changed to S/N threshold = 1 and threshold after S/N recalculation = 1. In cases where software could not detect a peak with the revised parameters, a S/N value of 1.0 was assigned. Although this may over-represent the peak heights of small signals, it avoids the divide-by-zero problem, which occurs when ratios are calculated with zero in the denominator. For calculation of the average “ratios of the S/N ratios” for the three replicates, the geometric mean of the S/N ratios of the triplicate repeat values was used. Recovery of Nonmodified Peptides

Three lanes for SDS-PAGE separation were prepared, each lane containing plasminogen (2 μg), HSA (5 μg), and α1-macroglobulin (5 μg). After SDS-PAGE separation, the bands containing each protein were excised separately from one of the lanes . Each of these samples was then subjected to the outgel protocol as described above. The proteins in the remaining two lanes were fixed, Coomasie stained, destained, and then subjected to the in-gel procedure, as described above. The three protein digests from one of the in-gel lanes were then put through a final SCX step to determine the extent of peptide loss. All digests were submitted to LC/ESI-MSMS then submitted to Mascot for Mascot score and percent sequence coverage.

■

RESULTS AND DISCUSSION

Development of the Out-Gel Digestion Procedure

Principle behind Out-Gel Digestion. The gel-fixing step causes denaturation, aggregation, and precipitation of the proteins, and it is known to reduce recoveries of intact proteins 529

dx.doi.org/10.1021/pr400710q | J. Proteome Res. 2014, 13, 527−535

Journal of Proteome Research

Article

analyzed. In the case of cross-linking applications, the position of the cross-linked protein band can be visualized if the reagents used contain fluorescent groups or other chromophores.23,36 The high degree of diffusion of the protein out of the gel and into the solution is probably attributable to the fact that, at the end of the SDS-PAGE separation process, the proteins were still complexed with SDS and were therefore efficiently solubilized. Since the gel was not fixed, these SDS−protein complexes were not disrupted, which facilitated efficient transfer of the intact proteins from the gel matrix into the liquid phase. Using an extraction buffer volume large enough to dilute the SDS to the point where trypsin remains active allows the use of a convenient one-step format for this procedure. Partial unfolding of the proteins due to residual SDS in the extraction buffer also may promote efficient digestion.

from a gel. This is thought to be due partly to reduced solubility and partly due to the size of the resulting aggregate, which may become physically trapped inside the gel; strands of the gel polymer may also become trapped within the aggregate.27 The commonly used technique of in-gel digestion relies on proteolysis to generate smaller pieces of the protein, which can be removed from the gel matrix by passive diffusion. This paper explores an alternative option of leaving out both the gel-fixing step as well as the staining step, which we have previously shown can reduce recoveries.28 If these steps are omitted, the protein or the protein complex can be removed from the gel by passive diffusion. Trypsin can be added either after the protein has diffused out of the gel, or it can be added to the diffusion buffer. Because the gels are not stained, cutting of the gel bands is based on a location relative to prestained gel markers. Efficiency of Passive Diffusion and Digestion of the Proteins. As part of this work, we compared a two-step digestion procedure (passive diffusion followed by addition of trypsin) and a one-step digestion procedure, where trypsin was immediately added to the extraction buffer. These procedures produced similar results, but the one-step protocol was found to be simpler and more rapid (Supporting Information Figure 1). The effectiveness of the extraction/digestion was examined using prestained protein molecular weight markers (Invitrogen). After incubation of the unfixed and unstained gel slices in the out-gel extraction buffer (which contained trypsin), the gel pieces were fixed and stained with Coomassie. Also, in order to visualize that proteins were diffusing out of gel slices, an entire molecular weight marker lane was excised and placed in the out-gel extraction buffer. No staining was observed in either experiment after the out-gel extraction and digestion, confirming that most of the protein had been removed from the gel (Supporting Information Figure 1). Determination of the Optimum Concentration of SDS. Because SDS is present in the gel running buffer, and because there are no gel fixing or staining/destaining steps which would have removed this SDS, it was important to ensure that the enzymatic activity of the trypsin was preserved. The optimum concentration of trypsin was determined by examining the degree of digestion of HIV-RT and albumin, which were monitored by SDS-PAGE analysis (Supporting Information Figure 2). In agreement with earlier studies,29−31 we determined that SDS had no inhibitory effect on in-solution tryptic digestion of proteins at SDS concentrations of 5 times the volume of the gel piece (1 mm3 gel: 5− 10 μL buffer). All subsequent out-gel digestions were performed using a 1:10 ratio. Removal of SDS before MS Analysis. To remove the SDS and to concentrate the peptides before MS analysis, the out-gel digest samples were purified with strong cation exchange (SCX) zip-tips. After this procedure, MS spectra were produced that were virtually identical to spectra obtained from in-solution digested proteins, indicating adequate removal of SDS as well as successful concentration of the sample. As has been previously demonstrated,35 excision of an unstained protein band can be performed by using the known positions of the band relative to the prestained molecular weight markers, or an entire lane of interest can be excised, sliced, and all of the gel slices can be

Recovery of Noncross-Linked Peptides

The method does not seem to improve the recovery of noncrosslinked peptides from proteins. This was determined by digesting three proteins of varying molecular weights using the in-gel and out-gel methods, followed by LC-MS/MS analysis using a QStar Pulsar I (AB SCIEX). We then looked at the Mascot scores and the percent sequence coverage generated by the peptides from each digest to determine which method led to the recovery of more peptides. In order to determine the extent of losses that occurred as a result of the SCX step, we performed the in-gel procedure with and without an SCX step. The results show that the standard in-gel approach outperforms the out-gel approach for the identification of noncross-linked peptides (Supporting Information Table 1). The results also show that the SCX step can lead to significant losses of nonmodified peptides when performed as a part of the in-gel and out-gel procedures, as has been reported by others.37 Enrichable versus Nonenrichable Cross-Linkers

Nonenrichable cross-linkers such as DSS can be used for both the in-gel and out-gel procedure, but there is a limit to the amount of peptide that can be loaded onto the column for analysis using nano-LC. Without enrichment, the cross-linking results for the HIV-RT heterodimer cross-linked with DSS were poor for both the in-gel and the out-gel methods. Using the in-gel procedure, three interprotein cross-links were identified, while only one was identified using the out-gel approach (Supporting Information Table 2). When the experiment was performed on HIV-RT, using the CBDPS cross-linker with enrichment, the in-gel method yielded eight interpeptide cross-links, representing five cross-linking sites (none of which were unique to the method). The out-gel method yielded seven interpeptide cross-links, representing six cross-linking sites, one of which was unique to the out-gel method (Supporting Information Table 3). One hypothesis is that CBDPS cross-linking produces bulky branched peptides, which are preferentially detected with the out-gel method. This study supports our hypothesis that the in-gel and out-gel methods are complementary, and that both in-gel and out-gel methods should be used to provide a more complete set of cross-links because one method may yield cross-links that the other does not. Application to a Cross-Linked Protein Complex

To determine the effectiveness of the out-gel digestion procedure, we performed cross-linking studies on the HIV-RT heterodimer with both DSS and CBDPS cross-linkers and on a complex of H2A and H2B histones with CBDPS, looking specifically for cross-links between the two histone isoforms. In 530

dx.doi.org/10.1021/pr400710q | J. Proteome Res. 2014, 13, 527−535

Journal of Proteome Research

Article

results were only qualitative because the peaks being compared were in different spectra. We therefore designed a quantitative recovery experiment, taking advantage of the isotopic coding of the CBDPS cross-linker (Figure 4A). Briefly, histones H2A and H2B were cross-linked using either the H or the D form of the cross-linker. Both forms were analyzed by SDS-PAGE in separate lanes on the same gel. One band was extracted and digested using the in-gel digestion procedure, while the other band was extracted and digested using the out-gel procedure. The extracts were then combined before MALDI analysis; thus, each interpeptide cross-link appears in the mass spectrum in two forms, resulting in a doublet (Figure 4B). One isotopic form represents the cross-link produced when the in-gel procedure was used, while the other represents the same cross-link produced when the out-gel procedure used. This experiment was performed in triplicate, and to account for any bias introduced by the label, the experiment was repeated with the labeling reversed. In this quantitative recovery experiment, where the cross-link signals from the two methods were compared within the same spectrum, only 1 out of the 28 (∼4%) of cross-links detected showed better recoveries using the in-gel procedure, while 8 (∼29%) showed the same recoveries with both procedures, and 19 (∼68%) of cross-links showed better recovery using the outgel procedure. These results are displayed graphically in Supporting Information Figure 3. Comparison of Peak Heights within the Same Spectrum. Although we had originally intended to calculate the ratios of the peak heights for the isotopically labeled peaks corresponding to the in-gel and out-gel procedures, some of the peaks were not detected. This led to a divide-by-zero problem, so in Supporting Information Table 6, we have noted only the relative peak heights (), which indicates which method gave higher cross-link ion signals (i.e., higher recoveries of the crosslinked peptide). Comparison of S/N Ratios. To avoid the divide-by-zero problem for nondetected (ND) peaks, a “ratio of ratios” was calculated for each ion corresponding to a cross-link, by dividing the S/N ratio for the ion representing the out-gel procedure by the S/N ratio for the corresponding ion from the in-gel procedure, (S/N) OG /(S/N) IG (Supporting Information Table7). In this treatment of the data, ND peaks have a S/N ratio of 1. A (S/N)OG/(S/N)IG ratio of 1 indicates better recovery when the out-gel procedure was used. A (S/N)OG/(S/N)IG of 1 indicates no difference in recoveries between the two procedures. The ratios of the signal-to-noise ratios of corresponding in-gel and out-gel signals were divided into three groups, and they were plotted in Figure 4C, as a stacked-column bar graph (note that where the geometric mean was