Loading menubar.....

Team:Tuebingen/Results

PacMn

Results

Cloning

Successful cloning depended on three main steps: PCR, Gibson Assembly, and chemical transformation of E. coli. These steps were gradually optimized until we finally perfected the procedure to clone all four of our main composites: GA1 (BBa_K3510002), GA2 (BBa_K3510003), GA3 (BBa_K3510004), and GA4 (BBa_K3510005) (Fig. 1). These composites can now be easily assembled by anyone (see Protocols and Notebook). For cloning of GA5 and GA6, we have proposed an experimental procedure in our Outlook.
Figure 1 Plate 1 Figure 1 Plate 2 Figure 1 Plate 3: Agarose gel electrophoresis of colony PCR of GA1, GA2, GA3,
                    GA4
                    (corrected sequences for iGEM compatibility). The products were separated by electrophoresis in
                    1 %
                    agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 1800 bp
                    (GA1),
                    ca. 1200 bp (GA2), ca. 1600 bp (GA3), and ca. 1000 bp (GA4). Cell material of the old GA1-4
                    sequenced
                    clones (containing a PstI restriction site) was used as positive control of PCR and gel
                    electrophoresis.
                    The negative control was the PCR master mix without template DNA. Approximately 88 % of the GA1-4
                    test
                    samples are correct.

Figure 1: Agarose gel electrophoresis of colony PCR of GA1, GA2, GA3, GA4 (corrected sequences for iGEM compatibility). The products were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 1800 bp (GA1), ca. 1200 bp (GA2), ca. 1600 bp (GA3), and ca. 1000 bp (GA4). Cell material of the old GA1-4 sequenced clones (containing a PstI restriction site) was used as positive control of PCR and gel electrophoresis. The negative control was the PCR master mix without template DNA. Approximately 88 % of the GA1-4 test samples are correct.

Comparing the high cloning efficiency represented above by the colony PCR results with the cloning efficiency in our first attempt (Fig. 2), we can say with confidence that our cloning experience was a success. We were able to increase the efficiency by a large margin, demonstrating how important optimization is.
Figure 2: Gel electrophoresis of colony PCR from GA1 (old sequence, first attempt)</b>. The products
                    were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was
                    used. pUC19 is a negative control, where only the empty plasmid was transformed. Samples 12 and 13
                    are labelled, as these were positive.

Figure 2: Gel electrophoresis of colony PCR from GA1 (old sequence, first attempt). The products were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. pUC19 is a negative control, where only the empty plasmid was transformed. Samples 12 and 13 are labelled, as these were positive.

A major part of our optimization consisted in perfecting PCR to create fragments with the proper overhangs for Gibson Assembly®. Our most troublesome fragment was F (FAST - Chromoprotein - Terminator fragment), which did not show any specific gel bands in the first attempts with the UCP HiFidelity PCR Kit (QIAGEN) (Fig. 3). By switching to the AllTaq PCR Master Mix Kit (QIAGEN), a more robust PCR kit that enhances primer binding, we finally obtained a specific gel band for fragment F (Fig. 4 B) but also many unspecific bands. Since unspecific gel bands were observed for multiple other fragments (Fig. 4 A), we decided to exclude these through gel extraction of the specific gel bands. The purity before (Fig. 5 A) and after Fig. 5 B) gel extraction improved considerably.
 Figure 3: Gel electrophoresis of gradient PCR (UCP HiFidelity PCR Kit) of fragment F</b>. The
                    products were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA
                    Ladder was used. The specific gel bands can not be clearly differentiated from the DNA smears.

Figure 3: Gel electrophoresis of gradient PCR (UCP HiFidelity PCR Kit) of fragment F. The products were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. The specific gel bands can not be clearly differentiated from the DNA smears.

Figure 4: Gel electrophoresis of gradient PCR (UCP HiFidelity PCR Kit) of fragment C (Fig. 4A) and
                    fragment F (AllTaq PCR Master Mix Kit) (Fig. 4B)</b>. The products were separated by electrophoresis
                    in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 900
                    bp for fragment C (GFP - Terminator fragment) and ca. 1300 bp for fragment F. In both gel images,
                    specific gel bands are observable, accompanied by non-specific bands.

Figure 4: Gel electrophoresis of gradient PCR (UCP HiFidelity PCR Kit) of fragment C (Fig. 4A) and fragment F (AllTaq PCR Master Mix Kit) (Fig. 4B). The products were separated by electrophoresis in 1 % agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 900 bp for fragment C (GFP - Terminator fragment) and ca. 1300 bp for fragment F. In both gel images, specific gel bands are observable, accompanied by non-specific bands.

Figure 5: Gel electrophoresis of PCR of fragment C and fragment F before (Fig. 5A) and after (Fig. 5B) gel extraction </b>. These products were separated by electrophoresis in 1% agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 900 bp for fragment C and ca. 1300 bp for fragment F. In both gel images, specific gel bands are observable. The intensity of non-specific bands decreases considerably after gel extraction.

Figure 5: Gel electrophoresis of PCR of fragment C and fragment F before (Fig. 5A) and after (Fig. 5B) gel extraction . These products were separated by electrophoresis in 1% agarose gels. As a marker, the BenchTop 1kb DNA Ladder was used. Desired band sizes: ca. 900 bp for fragment C and ca. 1300 bp for fragment F. In both gel images, specific gel bands are observable. The intensity of non-specific bands decreases considerably after gel extraction.

The last step to confirm successful cloning was sequencing. We completely sequenced the DNA of our composite parts with the Sanger method. Here, we realized that the sequences contained a PstI restriction site, which is incompatible with this year’s standards BioBrick RFC[10] and Type IIS RFC[1000]. Initially, we had excluded this restriction site in the design, but we accidentally ordered the unmodified sequences. By ordering the correct DNA sequences that contain a silent mutation in the leucine codon of the PstI restriction site, and by sticking to our optimized cloning procedures, we were able to correct this mistake efficiently (Fig. 6).
Figure 6: Alignment of Sanger sequencing results of the correct version of GA1 with the
                     incorrect reference sequence (SnapGene). Primers utilized: Microsynth standard primers M13-40
                      (forward) and  M13-r (reverse), and custom-made primers FAST_mid_FW and FAST_mid_RV.
                       The reference sequence is incorrect, as it contains the illegal restriction site PstI.
                       The sequence of GA1 aligns perfectly, except with the PstI restriction site.

Figure 6: Alignment of Sanger sequencing results of the correct version of GA1 with the incorrect reference sequence (SnapGene). Primers utilized: Microsynth standard primers M13-40 (forward) and M13-r (reverse), and custom-made primers FAST_mid_FW and FAST_mid_RV. The reference sequence is incorrect, as it contains the illegal restriction site PstI. The sequence of GA1 aligns perfectly, except with the PstI restriction site.

Fluorescence measurement with a plate reader

Once all of our constructs had been cloned into E. coli, the next step was to assess the functionality of our systems. For this, we measured the emitted fluorescence after growth in selective LB media containing three different manganese concentrations: 0 µM, 10 µM, and 20µM. We also compared our older incorrect constructs to our correctly modified constructs to analyze the effect of codon usage on protein expression.
Contrary to our expectations, there was no clear distinction between the fluorescence emitted by E. coli containing our constructs and E. coli containing an empty pUC19 plasmid. Moreover, the fluorescence intensity of E. coli containing manganese-inducible promoters (GA1 and GA2) did not increase with higher manganese concentrations (Fig. 7 A-C). Since the results were inconclusive, we assume that we only measured background noise generated by the auto-fluorescence of the cells.
Figure 7 Part 1 Figure 7 Part 2 Figure 7: Fluorescence intensity of E. coli containing different constructs after incubation in
                    media with different manganese concentrations</b>. The experiments were repeated three times, each
                    in triplicates. The mean values +/- standard deviation are shown. Constructs labelled as “old
                    sequence” still include the PstI restriction site while constructs labelled as “new sequence” do
                    not. (A) 0 µM manganese (II) chloride. (B) 10 µM manganese (II) chloride. (C) 20 µM manganese (II)
                    chloride.

Figure 7: Fluorescence intensity of E. coli containing different constructs after incubation in media with different manganese concentrations. The experiments were repeated three times, each in triplicates. The mean values +/- standard deviation are shown. Constructs labelled as “old sequence” still include the PstI restriction site while constructs labelled as “new sequence” do not. (A) 0 µM manganese (II) chloride. (B) 10 µM manganese (II) chloride. (C) 20 µM manganese (II) chloride.

Imaging by fluorescence microscopy

An alternative option for assessing fluorescence is by fluorescence microscopy. Here, we compared E. coli containing the empty pUC19 plasmid (negative control) and E. coli with the corrected GA2 construct. The cells were grown in selective LB media containing 10 µM manganese(II) chloride to induce our regulation system before they were labelled with the TF-Lime fluorogen. The fluorescence intensity observed in both cultures was very similar (Fig. 8). Again, background from the auto-fluorescence emitted by the cells is presumed.
Figure 8: Live cell imaging of cells containing pUC19 and GA2 (corrected version) under the fluorescence microscope. Laser settings: wavelengths 485 nm (absorbance) and 535 nm (emission). 1.5 x 630 zoom. (A) pUC19 control, 10 µM manganese(II) chloride. (B) GA2 (corrected version), 10 µM manganese(II) chloride.

Figure 8: Live cell imaging of cells containing pUC19 and GA2 (corrected version) under the fluorescence microscope. Laser settings: wavelengths 485 nm (absorbance) and 535 nm (emission). 1.5 x 630 zoom. (A) pUC19 control, 10 µM manganese(II) chloride. (B) GA2 (corrected version), 10 µM manganese(II) chloride.

Chromoprotein test

Since the results obtained from fluorescence experiments were inconclusive (Fig. 7 and Fig. 8), we decided to qualitatively test the expression of the blue chromoprotein of GA1 and GA3. For this, a sequenced clone for GA1 and GA3 (old and new versions) and a pUC19 clone were streaked on selective LB agar plates (100 µg/ml ampicillin sodium salt) containing three different manganese(II) chloride concentrations (0 µM, 10 µM, and 20 µM). Blue colonies were expected for colonies grown on manganese as opposed to colonies grown on the control plates without manganese. However, the plates did not reveal any blue coloring of the colonies growing on plates with manganese, regardless of the concentration (Fig. 9).
Figure 9: E. coli colonies containing the constructs with the chromoprotein gene (GA1 and GA3) grown on plates with different  manganese(II) chloride concentrations. The selective agar plates (100 µg/ml ampicillin sodium salt)  were prepared with three different manganese(II) chloride concentrations: 0 µM (A), 10 µM (B), or 20 µM (C)..

Figure 9: E. coli colonies containing the constructs with the chromoprotein gene (GA1 and GA3) grown on plates with different manganese(II) chloride concentrations. The selective agar plates (100 µg/ml ampicillin sodium salt) were prepared with three different manganese(II) chloride concentrations: 0 µM (A), 10 µM (B), or 20 µM (C).

The failure to produce fluorescence (Fig. 7 and Fig. 8) or a visible blue color (Fig. 9) with our GA1-4 constructs (old and corrected versions) made us take a closer look at our sequences again. We noticed the absence of a ribosome binding site (RBS) between our riboswitch sequence and the phytochelatin/chromoprotein gene. The mistake in our design was caused by the misconception that the riboswitch sequence available in the parts registry would include an RBS downstream of the riboswitch.

Stability test

We assessed the extent of transgenerational stability or toxicity of the constructs GA1, GA2, GA3, and GA4 in comparison to the pUC19 plasmid control in E.coli by nine transfers of clones through non-selective media with at least one overnight incubation at 37 °C in between (Fig. 10). The final transfer was performed to a selective purity plate which showed single colonies after overnight incubation at 37 °C (Fig. 11).

Figure 10: Some of the non-selective LB agar plates for the plasmid stability test of constructs GA1, GA2, GA3, GA4, and the empty pUC19 vector in E. coli. With a few exceptions for the pUC19 clone, all clones were transferred to the same plate each time, separated by a grid.

Figure 10: Some of the non-selective LB agar plates for the plasmid stability test of constructs GA1, GA2, GA3, GA4, and the empty pUC19 vector in E. coli. With a few exceptions for the pUC19 clone, all clones were transferred to the same plate each time, separated by a grid.

Figure 11: Final selective LB agar plate after nine transfers of E. coli containing GA1, GA2, GA3, and GA4.  Single colonies were picked from this purity plate for colony PCR.

Figure 11: Final selective LB agar plate after nine transfers of E. coli containing GA1, GA2, GA3, and GA4. Single colonies were picked from this purity plate for colony PCR.

Colony PCR was performed to evaluate the presence of the constructs after the transgenerational stability test (Fig. 12).
The targets of the PCRs were observed at the desired band size of GA1 (ca. 1800 bp), GA2 (ca. 1200 bp), GA3 (ca. 1600 bp) and GA4 (ca. 1000 bp) on the gel (Fig. 12). Since the bands for samples GA1 and GA3 were very faint (indicated by red arrows), and smearing and inconsistency occurred in the positive controls, a PCR amplification error was discussed. Therefore, a repetition of this PCR is advisable. Nonetheless, it can be cautiously surmised that the presence of the plasmids without expression of our constructs is tolerated for nine transfers and therefore presumably does not constitute a heavy burden for the bacteria. As the proteins encoded in the constructs cannot be expressed, no further assumptions can be made concerning their stability or toxicity.
Figure 12: Gel electrophoresis of colony PCR of stability experiment samples with controls. Samples for E. coli containing the constructs GA1, 2, 3, and 4 were taken from the final selective plate, while the pUC19 colony that underwent similar transfers was chosen from the penultimate non-selective plate. Positive controls for GA1, GA2, GA3, GA4  and pUC19 were picked from their respective pre-existing purity plates. Results of the colony PCR and subsequent separation via  electrophoresis in 1% agarose gels are shown in this figure. As a marker, the BenchTop 1kb DNA Ladder was used.

Figure 12: Gel electrophoresis of colony PCR of stability experiment samples with controls. Samples for E. coli containing the constructs GA1, 2, 3, and 4 were taken from the final selective plate, while the pUC19 colony that underwent similar transfers was chosen from the penultimate non-selective plate. Positive controls for GA1, GA2, GA3, GA4 and pUC19 were picked from their respective pre-existing purity plates. Results of the colony PCR and subsequent separation via electrophoresis in 1% agarose gels are shown in this figure. As a marker, the BenchTop 1kb DNA Ladder was used.

mRNA experiment

With only two days left for wet lab work after noticing the missing RBS in our constructs, we designed a Reverse Transcriptase (RT)-PCR experiment to investigate mRNA synthesis and validate our promoters, which function independently of an RBS.
We expected an amplicon of ca. 320 bp and a difference in signal intensity between growth with 0 µM and 10 µM manganese(II) chloride for constructs with manganese-inducible promoters. However, the negative controls had the same band sizes and different manganese concentrations did not produce different signal intensities (Fig. 13). Although the band sizes of the negative controls seem to be slightly larger, this could be explained by a variable gel morphology. Interestingly, all negative controls have similar bands to the AllTaq PCR Master Mix negative control, which did not contain template cDNA. In summary, the experiment was inconclusive. In order to gain more conclusive results, optimization of the PCR is required.

Figure 13: Gel electrophoresis of cDNA amplification via PCR. The desired band size is ca. 320 bp. Test subjects include the old GA1-4 clones and the corrected GA2 and GA3 clones. Negative controls include a pUC19 clone, the reverse transcription (RT) reaction mix, and the AllTaq PCR master mix without template cDNA. As a marker, the BenchTop 1kb DNA Ladder was used.

Figure 13: Gel electrophoresis of cDNA amplification via PCR . The desired band size is ca. 320 bp. Test subjects include the old GA1-4 clones and the corrected GA2 and GA3 clones. Negative controls include a pUC19 clone, the reverse transcription (RT) reaction mix, and the AllTaq PCR master mix without template cDNA. As a marker, the BenchTop 1kb DNA Ladder was used.

Outlook

Ribosome binding site (RBS)
Having to include an RBS in our constructs in order to have them expressed, we evaluated two choices. First, ordering corrected DNA fragments and repeating PCRs and Gibson Assemblies, and second, appending an RBS either downstream of the riboswitch fragment or upstream of the FAST2 fragment by PCR amplification with primers containing the RBS sequence. The second choice would also require Gibson Assembly consecutively. Either way, new primers for Gibson Assembly would have to be ordered, because the addition of the RBS changes the end of one fragment and therefore the necessary overhangs. As amplification with long primers to add novel features usually requires a notorious PCR optimization process, and both methods would require ordering DNA, we would favour the first option for which we already have well-established protocols. Upon successful adaptation of our construct, the experiments described in the following may be conducted.
Manganese homeostasis - Potential for system improvement
While researching manganese homeostasis in E. coli, we found that the bacteria reduce the expression of the manganese importer MntH in the presence of high manganese levels [1, 2]. Presumably, this would hamper the functionality of our biosensor, which can only detect intracellular manganese. One way to circumvent this issue would be to add the mntH gene downstream of a constitutive promoter in our plasmids. Consequently, we would uncouple the manganese importer from the manganese homeostasis regulatory network of E. coli.
Moreover, it was shown that manganese import is increased under oxidative stress [2]. Strains lacking catalase/peroxidase (Hpx-) cannot degrade hydrogen peroxide and accumulate H2O2 when cultured in aerobic media [3]. Therefore, we could use a sensitive strain to assess whether the fluorescent signal is more reliable and dependent on manganese concentrations when cells react to oxidative stress. As a control and to quantify the effect of maintained manganese import, it would be necessary to culture the same strains under anaerobic conditions. Here, our FAST2-tagged system comes in handy, because it is able to provide a fluorescent signal under anaerobic conditions.
Phytochelatin
To assess whether the phytochelatin successfully binds manganese, we thought about three experiments:
1. Evaluation of the rescue of E. coli manganese tolerance by induced expression of phytochelatin.
In E. coli lacking the manganese exporter MntP, high manganese concentrations in the medium lead to reduced growth [5]. We presume that high manganese concentrations may be toxic due to the interference of manganese with intracellular biochemical processes and enzymes. Therefore, we hypothesize that the complexation of manganese by the phytochelatin may restore bacterial growth by preventing this interference.
The experiment (Fig. 1) would include ten E. coli cultures containing different plasmids: Constructs BBa_K3510003 (GA2) and BBa_K3510005 (GA4) with the phytochelatin and their control constructs with the chromoprotein BBa_K3510002 (GA1) and BBa_K3510004 (GA3). Additionally, the empty pUC19 plasmid serves as a negative control for regular cell growth. All cultures would be maintained at the same conditions with standard media or media including 0.5 mM Manganese(II) chloride. For 4 to 5 hours, cell growth would be assessed by measuring the OD600 every 30 min to 1 hour. With successful chelation of manganese(II) ions by the phytochelatin, we would expect higher OD600 values and hence growth for these cultures, while the chromoprotein controls should show the same (lower) OD600 as the pUC19 negative control.
Furthermore, the experiment should be performed with a mntP null mutated strain transformed with the vector containing the phytochelatin or an empty pUC19 vector as a control. Here we hope to show that phytochelatin expression rescues bacterial growth.

Figure 1: Experiment to determine cell growth in different manganese(II) chloride concentrations.

Figure 1: Experiment to determine cell growth in different manganese(II) chloride concentrations.

2. Investigation of phytochelatin expressing E. coli and their presumed withdrawal of manganese from the environment.
We hypothesise that with the complexation of manganese by the phytochelatin, the bacteria withdraw the ion from the surrounding media, therefore decreasing the extracellular manganese concentration. With the complexation of manganese, we further assume that the MntP exporter of the ion will not be upregulated via a feedback mechanism if we can show in the preceding experiment that phytochelatin rescues bacterial growth in mntP null mutated strains. As control, we will employ our chromoprotein control constructs which should turn a visible blue color in the presence of manganese without chelating the ion. Since a detoxification will not occur in non-phytochelatin expressing controls, we expect a feedback loop which will increase the MntP manganese exporter and decrease the genome-encoded MntH importer function. Hence, in these controls manganese concentrations in the bacteria's environment should remain stable.
To evaluate the influence of phytochelatin expression on manganese clearance we will employ a chamber with two reservoirs divided by a manganese ion permeable membrane. In manganese(II) chloride-containing culture medium, E. coli expressing phytochelatin is grown in one chamber, while the other side harbours the chromoprotein encoding control bacteria (Fig. 2). Due to the presence of manganese in the environment of both cultures in the beginning of the experiment, we expect the chromoprotein to be expressed in the control strain, providing a blue signal, while phytochelatin expression should be upregulated in the other strain. Here, the expression of phytochelatin presumably serves for complexation of manganese.
Since the permeable membrane ensures equal manganese concentrations on both sides of the membrane, we hypothesise that the general manganese concentration decreases by the amount which is cleared by the phytochelatin expressing strain. Furthermore, once the concentration falls below the limit necessary for the expression of the chromoprotein in the control strain, the blue color should vanish.
As a control, the same experimental set-up will be followed using the chromoprotein control and a pUC19 empty vector transformed E. coli, respectively. Provided the experiment works as expected, the manganese concentration should not fall below the detection limit of the chromoprotein strain, thus enabling continuous expression of the blue chromoprotein.

Figure 2: Investigation of phytochelatin-expressing E. coli and their presumed withdrawal of manganese from the environment.

Figure 2: Investigation of phytochelatin-expressing E. coli and their presumed withdrawal of manganese from the environment.

3. Measurement of intracellular manganese accumulation in phytochelatin expressing bacteria using mass spectrometry.
We expect manganese to accumulate in bacterial cells which express phytochelatin. A precise method to assess intracellular manganese is mass spectrometry. Therefore we propose the following experiment designed in accordance with a protocol by Martin, Waters et. al (2015). Phytochelatin or chromoprotein encoding E. coli cultures are cultured for 2.5 hours with or without 0.5 mM manganese(II) chloride. They are harvested via centrifugation and washing of the cells with Tris-HCl/EDTA and subsequently resuspended in Tris-HCl. Next, the cells are lysed, and samples are prepared for mass spectrometry to the amount of intracellular manganese (Fig. 3). For the analysis, the values are normalized to total protein concentrations in the lysate.


Figure 3: Determination total intracellular manganese by mass spectrometry.

Figure 3: Determination total intracellular manganese by mass spectrometry.

Side project Vitamin B12:
Because successful cloning was not achieved for the two constructs of our side project, we suggest doing it in a two-step Gibson Assembly (Fig. 4). Here, the fragments are prepared in the same manner, but first the assembly of two inserts with the vector is approached. After verification of the correct assembly of these three fragments by colony PCR and restriction digest, the construct is linearized again by PCR. Then the second Gibson Assembly follows by inserting the third and last fragment into the vector. As with all other Gibson Assemblies, we then would check for correct assembly and sequence the whole insert to make sure it does not contain any mutations.

Figure 4: Step-wise Gibson Assembly.

Figure 4: Step-wise Gibson Assembly.

References

  1. Patzer SI, Hantke K. Dual repression by Fe(2+)-Fur and Mn(2+)-MntR of the mntH gene, encoding an NRAMP-like Mn(2+) transporter in Escherichia coli. J Bacteriol 2001; 183(16):4806–13.
  2. Anjem A, Varghese S, Imlay JA. Manganese import is a key element of the OxyR response to hydrogen peroxide in Escherichia coli. Mol Microbiol 2009; 72(4):844–58.
  3. DSeaver LC, Imlay JA. Hydrogen peroxide fluxes and compartmentalization inside growing Escherichia coli. J Bacteriol 2001; 183(24):7182–9.
  4. Dambach M, Sandoval M, Updegrove TB, Anantharaman V, Aravind L, Waters LS et al. The ubiquitous yybP-ykoY riboswitch is a manganese-responsive regulatory element. Mol Cell 2015; 57(6):1099–109.
  5. Martin JE, Waters LS, Storz G, Imlay JA. The Escherichia coli small protein MntS and exporter MntP optimize the intracellular concentration of manganese. PLoS Genet 2015; 11(3):e1004977.