Supplementary Figure 1: Data analysis workflow.
From ProteoPlex: stability optimization of macromolecular complexes by sparse-matrix screening of chemical space
- Ashwin Chari1, n1
- David Haselbach1, n1
- Jan-Martin Kirves1, n1
- Juergen Ohmer2,
- Elham Paknia2,
- Niels Fischer1,
- Oleg Ganichkin3,
- Vanessa Möller4,
- Jeremiah J Frye5,
- Georg Petzold6,
- Marc Jarvis6,
- Michael Tietzel7,
- Clemens Grimm2,
- Jan-Michael Peters6,
- Brenda A Schulman5, 8,
- Kai Tittmann7,
- Jürgen Markl4,
- Utz Fischer2,
- Holger Stark1,
- Journal name:
- Nature Methods
- Volume:
- 12,
- Pages:
- 859–865
- Year published:
- DOI:
- doi:10.1038/nmeth.3493
- Received
- Accepted
- Published online
First step: Identification of valid datasets. Amplitudes and number of valid maxima are investigated. If more than 6 local maxima are identified the curve is neglected. Second step: Background subtraction. The signal from both buffer references are averaged and subtracted from all individual curves in the dataset. Third step: Normalization. All datasets are normalized to a range of 0 to 1000, by setting the lowest local minimum to 0 and the highest relevant local maximum to 1000. Fourth step: Identification and removal of datasets containing air bubbles. The slope from the first data-point to the first relevant local minimum is evaluated. Curves with large slopes are omitted. This also removes transitions of proteins, which were already aggregated in solution prior to measurement. Fifth step: Data approximation according to the thermodynamic framework presented here for two- to five-state models. Sixth step: Parameter extraction for two-state and best fitting models. The values for Tm, ∆Hm, R² are extracted for the two-state model and the best fitting model. These extracted parameters are then hierarchically sorted to evaluate the most stabilizing conditions.
Additional data
Author footnotes
These authors contributed equally to this work.
- Ashwin Chari,
- David Haselbach &
- Jan-Martin Kirves
Affiliations
-
Research Group of 3D Electron Cryomicroscopy, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany.
- Ashwin Chari,
- David Haselbach,
- Jan-Martin Kirves,
- Niels Fischer &
- Holger Stark
-
Department of Biochemistry, Theodor-Boveri Institute, University of Würzburg, Würzburg, Germany.
- Juergen Ohmer,
- Elham Paknia,
- Clemens Grimm &
- Utz Fischer
-
Institut für Chemie und Biochemie, Freie Universität Berlin, Berlin, Germany.
- Oleg Ganichkin
-
Institut für Zoologie – Abteilung für Molekular Tierphysiologie, Johannes Gutenberg Universität Mainz, Mainz, Germany.
- Vanessa Möller &
- Jürgen Markl
-
Department of Structural Biology, St. Jude Children's Research Hospital, Memphis, Tennessee, USA.
- Jeremiah J Frye &
- Brenda A Schulman
-
Research Institute of Molecular Pathology, Vienna, Austria.
- Georg Petzold,
- Marc Jarvis &
- Jan-Michael Peters
-
Department for Bioanalytics, Georg-August University Göttingen, Göttingen, Germany.
- Michael Tietzel &
- Kai Tittmann
-
Howard Hughes Medical Institute, St. Jude Children's Research Hospital, Memphis, Tennessee, USA.
- Brenda A Schulman
Contributions
A.C. and H.S. designed research; A.C. and D.H. performed most experiments and data analysis with the support of J.O., E.P., O.G. and V.M.; D.H. derived the thermodynamic theory used for the analysis; J.-M.K. designed an analysis software and performed analysis of data; N.F., C.G., J.J.F., M.J., M.T. and G.P. performed further experiments; J.M., J.-M.P., B.A.S. and K.T. supervised research partially; initial experiments were performed in the lab of U.F.; A.C., D.H., J.-M.K. and H.S. analyzed the data and prepared the manuscript.
Competing financial interests
A.C., D.H., J.-M.K. and H.S. have filed a patent application (WO2013034160 A1: “Methods for analyzing biological macromolecular complexes and use thereof”) and are currently developing a relevant product in cooperation with FEI Company.
Author details
Ashwin Chari
Search for this author in:
David Haselbach
Search for this author in:
Jan-Martin Kirves
Search for this author in:
Juergen Ohmer
Search for this author in:
Elham Paknia
Search for this author in:
Niels Fischer
Search for this author in:
Oleg Ganichkin
Search for this author in:
Vanessa Möller
Search for this author in:
Jeremiah J Frye
Search for this author in:
Georg Petzold
Search for this author in:
Marc Jarvis
Search for this author in:
Michael Tietzel
Search for this author in:
Clemens Grimm
Search for this author in:
Jan-Michael Peters
Search for this author in:
Brenda A Schulman
Search for this author in:
Kai Tittmann
Search for this author in:
Jürgen Markl
Search for this author in:
Utz Fischer
Search for this author in:
Holger Stark
Search for this author in:
-
Supplementary Figure 1: Data analysis workflow.
Hover over figure to zoomFirst step: Identification of valid datasets. Amplitudes and number of valid maxima are investigated. If more than 6 local maxima are identified the curve is neglected. Second step: Background subtraction. The signal from both buffer references are averaged and subtracted from all individual curves in the dataset. Third step: Normalization. All datasets are normalized to a range of 0 to 1000, by setting the lowest local minimum to 0 and the highest relevant local maximum to 1000. Fourth step: Identification and removal of datasets containing air bubbles. The slope from the first data-point to the first relevant local minimum is evaluated. Curves with large slopes are omitted. This also removes transitions of proteins, which were already aggregated in solution prior to measurement. Fifth step: Data approximation according to the thermodynamic framework presented here for two- to five-state models. Sixth step: Parameter extraction for two-state and best fitting models. The values for Tm, ∆Hm, R² are extracted for the two-state model and the best fitting model. These extracted parameters are then hierarchically sorted to evaluate the most stabilizing conditions.
-
Supplementary Figure 2: Schematic representation of possible unfolding scenarios.
Hover over figure to zoomSchematic temperature vs. fluorescence curves are shown, with different colors representing three phases of the entire unfolding transition. The left white part represents the folded phase, the middle orange part the unfolding phase and the right white part the aggregation phase. The half-maximal intensity corresponding to the inflection point of the curve and thus the melting temperature is depicted as dotted line. The dashed line curve insets represent an idealized two state unfolding behavior, with a steep transition. Cartoons of the behavior of proteins are depicted below the graph. (a) A typical unfolding transition curve of a single domain protein is shown. Note, that it overlays well with the dashed line and therefore it is assumed to represent two-state unfolding. (b) The unfolding curve for an instable protein complex is shown. The complex disassembles before it unfolds, yielding multiple transitions and a considerable divergence from two-state unfolding. (c) The unfolding transition for another instable protein complex is shown. Disassembly and unfolding occur in a narrow temperature range yielding multiple transitions which superimpose in a way that they cannot be distinguished anymore. However, this yields a shallow transition in comparison to the dashed line curve (i.e. two-state unfolding). (d) The unfolding curve of a stable complex is shown. Disassembly and unfolding occur in a near concerted manner. Thus, the curve resembles the two-state transition shown in (a).
-
Supplementary Figure 3: Theoretical unfolding transition behaviors.
Hover over figure to zoomTemperature – fluorescence unfolding curves of a theoretical three-subunit complex are shown. Left: The single-transitions of the individual components are depicted. Right: Graphs depict a normalized sum of three individual curves. (a) The three components unfold independently from each other at different temperatures. The sum is a curve showing several independent transitions. (b) The components unfold at similar temperatures but still independently from each other. The resulting sum resembles a two-state unfolding curve. (c) The components unfold cooperatively at the same temperature. This also results in a two-state curve, which is steeper than in the middle case. However, the melting temperature of the shallower uncooperative unfolding transitions results in an apparently higher melting point Tm1, than the cooperative transition (Tm2) as visualized by the dotted lines.
-
Supplementary Figure 4: Evaluation of the quality of ProteoPlex data approximation.
Hover over figure to zoomTwo experimental data curves exemplifying the curve fitting process are shown as blue dots. In DSF only the transition part of the data is fitted by a Boltzmann model as shown in yellow13. The best data approximation from ProteoPlex is shown in green. While Boltzmann data approximation still gives acceptable results for a near two-state unfolding behavior (left), multiple transitions cannot be approximated by a simple Boltzmann model (right). In contrast, ProteoPlex still describes the obtained curve well. Of note: The ProteoPlex model is able to fit the whole curve and thus will obtain more accurate parameters.
-
Supplementary Figure 5: Screening for additives.
Hover over figure to zoomAnalysis of Biomphalaria glabrata hemoglobin complex (BgHb, 1.5 MDa native molecular weight) – a protein of unknown structure. Under standard purification conditions BgHb is mostly present as aggregated particles in negatively stained EM images (left panel, scale bar = 50 nm). The upper pane of the middle panel depicts a subset of unfolding transitions from a ProteoPlex pH screen. The stability of BgHb is gradually increased from alkaline to acidic Imidazole buffer conditions, with a final enhanced stabilization of 45 K at pH 5.8 compared to pH 8.2. The lower pane of the middle panel reveals that interpretable unfolding transitions of the complex only occurs in Imidazole buffer, which suggests the role of Imidazole additionally as a stabilizing ligand. In Imidazole pH 5.8, negatively stained EM analysis reveals a monodisperse field of compact particles (right panel, scale bar 50 nm).
-
Supplementary Figure 6: Finding optimal conditions for reconstitution experiments.
Hover over figure to zoom(a) Reconstitution of PDHc from individually purified subunits (E1, E2, E3). A constant concentration of the core E2 component was mixed with increasing amounts of E1 along the y-axis and increasing E3 amounts along the x-axis (5, 4 and 2.5, 1.5, 1, 0.5, 0.25 and 0.125 fold molar excess of E2 subunit) and assayed with ProteoPlex. Experimental curves (blue dots) and fits (green) from the screen (middle) show two-state unfolding behavior in the case of a high excess of E1 over E2 and low amounts of E3 and mono-disperse, compact particles in EM images (right panel). Whereas, low amounts of E1 and E3 in comparison to E2 yields polyphasic transitions and aggregated particles (left panel).The scalebar corresponds to 50 nm in the EM images. (b) SDS-PAGE of the peak fractions of the reconstituted sample and a sample purified from native source are shown. Asterisks denote impurities in the sample purified from native source. Note the stoichiometry of both samples agree well with each other.
-
Supplementary Figure 7: Automation.
Hover over figure to zoom(a) The robotic platform consists of a liquid handling system a plate sealing device and a RT-PCR machine. The setup allows full automation of liquid handling, thermal melt measurement and data evaluation. (b) Reproducibility tests using manual (top panel) and robotic pipetting (lower panel) of 96 identical conditions. The robotic liquid handling reveals a significant increase in reproducibility as shown by an almost perfect overlay of the normalized curves.
-
Supplementary Figure 8: Test of different real-time PCR (RT-PCR) machines.
Hover over figure to zoomTwo of the most commonly used RT-PCR machines were tested towards their applicability with ProteoPlex: the ABI Via7 and the Bio-Rad CFX connect. First the background signal was analyzed. While it was uniform for the Bio-Rad machine (upper row, right panel), strong discrepancies could be seen for the ABI device. A very similar result can be seen by just measuring Sypro Orange diluted to the concentration used for screening (second row). Lastly the machines were tested with Lysozyme and the amplitude of the noise as well of the protein in optimal concentration was measured. A simple SNR was estimated by dividing these two amplitudes. The Bio-Rad machine reaches a SNR that is twice higher than the Via7.
