Research Article

A Microfluidic Device for Temporally Controlled Gene Expression and Long-Term Fluorescent Imaging in Unperturbed Dividing Yeast Cells

  • Gilles Charvin mail,

    To whom correspondence should be addressed. E-mail:

    Affiliation: Center For Studies in Physics and Biology, The Rockefeller University, New York, New York, United States of America

  • Frederick R. Cross,

    Affiliation: The Rockefeller University, New York, New York, United States of America

  • Eric D. Siggia

    Affiliation: Center For Studies in Physics and Biology, The Rockefeller University, New York, New York, United States of America

  • Published: January 23, 2008
  • DOI: 10.1371/journal.pone.0001468
  • Published in PLOS ONE

Reader Comments (4)

Post a new comment on this article

Referee comments: Referee 1

Posted by PLoS_ONE_Group on 25 Jan 2008 at 13:34 GMT

Referee 1's review:

Review for:

A microfluidic device for temporally controlled gene expression and long-term fluorescent imaging in unperturbed dividing yeast cells

G.Charvin, F.R. Cross, E.D. Siggia

In this work, the authors present a combined approach involving fluorescence imaging of dividing yeast cells growing in a microfluidic device, followed by careful computational analysis of these data, in order to develop an approach to address cell-cycle based studies for yeast. In particular, they have analyzed the transient induction of the MET3 promoter using this device, and have carefully characterized the various rates relating to its production, maturation and decay, while comparing these to other commonly used promoters like GAL1, ACT1 etc. They also show specific examples of using this approach to control cell cycle dynamics by providing external inputs to control the G1/S or the mitotic exit transitions. Thus, the study presents an approach/ tool that would be interesting to a large community of scientists.

This reviewer would support the acceptance of this manuscript for publication in PLoS ONE, provided the authors take the following comments/ suggestions into account and address them.

1) The authors have developed a simple microfluidics-based experimental framework to perform long-term imaging of yeast cells by trapping them between a dialysis membrane and a thin PDMS layer. This trapping ensures that the yeast cells bud in the plane of focus, thus ensuring better imaging over time. However, many recent studies have turned to long-term fluorescence imaging of yeast cells (including references 8 and 10 of this paper) and many have included microfluidics based approaches that constrain the cell growth in the plane of focus while providing a system for fast-media switching as well as long-term imaging (these include references 9 and 10 in the paper). Thus, the author's claim that their setup is the first to integrate these features into one device (as mentioned on page 3, Introduction) is largely misleading. The authors have used a different way of constraining their cells from the ones mentioned in their references viz. a dialysis membrane and have used their setup to study transient single-cell gene expression; and those should be their claims.

2) For the reasons mentioned in point (1) above, I would also suggest that the authors consider rewording the title of the study to focus on their overall approach, rather than the device itself. (see for example the title of the study in reference 10)

3) Some questions about the device: (a) Does the 40 um PDMS layer not affect imaging with objectives that have small working distances? The authors seem to have used a high NA aperture (63X, 1.4 NA) objective without any visible issues with the images, but I am curious to know if they have faced this issue. (b) Have any other cell types other than yeast been studied in this device? It is conceivable that this device may not be usable for mammalian cells, since the PDMS surface may not have an appropriate rigidity or may not be easily amenable to coating with several different substrates. The authors could potentially address this point. (c) The authors mention that cells did not grow as well without the PDMS layer, possibly due to the pressure of the dialysis membrane, thus resulting in a need for a soft PDMS cushion. Other studies where cells were constrained similarly do not report such problems with cell growth on glass. Have the authors investigated the option of constraining the dialysis membrane such that it applies lower pressure on the cells, while still providing a deterrent for out-of-plane budding.

4) The study includes the use of a MATLAB-based segmentation algorithm which seems to perform appreciably for cells that are budding relatively close to each other, and are indeed touching each other. Could the authors provide further details about the same in their methods or supplementary section? The authors have referred to their own previous study where this algorithm was used, however, that immediate reference did not provide further details about the implementation either. Similarly, could they detail the approach taken to track the cells across time-points/ generations. The authors should also very explicitly contrast their methods of image analysis with the very detailed Methods study from Brent et al., recently published in Nature Methods (Gordon et al., 2007).

5) Minor comments:
a) There is a typo on Page 9, where the authors refer to the use of a 1X Met solution twice. I presume it should be 10X Met solution, as shown in Figure 4.
b) Could the authors label a '0' & '1' (corresponding to the budding state) on the right vertical axis of Fig 5a, just for completeness (as in Fig. 5b).
c) On page 10, they mention that Whi5 exit from the nucleus is too rapid in mother cells to allow accurate measurement. What are those time scales, and why can the imaging not be done fast enough? Is the 3 min measurement interval constrained because of the autofocusing algorithm?
d) Another typo on Page 16 (data acquisition): should be "beginning" not "begging"

N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements

RE: Referee comments: Referee 1

gcharvin replied to PLoS_ONE_Group on 26 Jan 2008 at 22:14 GMT

This is the answer to the questions and comments of the editor and end the referee :

We thank the editor and reviewer for the careful reading of the manuscript.
Here are specific answers to the questions/issues addressed by the reviewer and the editor :

Answer to the Editor's question :

The editor pointed out that the time we reported for cycloheximide to apparently shut off protein synthesis ( 21 minutes) is likely an overestimate. It has indeed been reported that protein synthesis stops within a couple of minutes following the addition of cycloheximide at the concentrations we used, and it was an oversight on our part to neglect this known fact.

We are using total cell area to quantify growth in Figure 2, so it is quite possible that area increases for some reason unrelated to protein synthesis (e.g. due to an increase in vacuole size) even though protein synthesis has stopped. Taken together with prior work we are sure that protein production has ceased by 21 minutes (and almost surely earlier, since we know that the delivery time for small molecules is under a minute based on the fluorescein washout experiment described in the text) under our conditions. Fortunately, this ill-understood lag in cessation of area increase causes no problem for determination of the protein fluorescence maturation rate: we model it as a first order process, so the starting time is irrelevant. Since we fit the average pixel intensity, the cessation of area growth has to occur before our measurements become meaningful, but the first-order rate can be determined from any point thereafter. For exponential growth, no matter at what time we begin the fit, the same decay rate will be found. We included a remark in the text to make this clear.

Answers to the Reviewer's questions

1- We regret inadvertently causing any impression that we were failing to sufficiently credit prior work. The introduction section has been rewritten in order to provide more details about prior work and compare and contrast our device to previous ones. Previous papers eg [8,9,10] do not demonstrate rapid media change. Ref [8] (Brent group) did not use microfluidics; they attached their cells with ConA to 96 well plates. Ref [9] studied pherormone response over several hours and did not need rapid media exchange, and Ref [10] (Hasty group) uses microfluidics but does not change media. Certainly long term imaging of confined yeast cells is an old method, and our first 3 references were to papers from 1981-1996. To avoid misunderstandings and discussions about what was or could have been done before, we modify the text to say refs 8-10 did not demonstrate rapid media exchange and explore its application to long term cell cycle studies; this is definitively true, unless there are other literature references that we have missed. If this is the case, we request that the editor or reviewer tell us; we certainly want to give all credit where it is due.

2- We propose to leave the title unchanged, since it reads “A microfluidic...”,
we do not say 'novel' and the rest of the title suggests the cell biological context/applications of the device which is surely our intended audience; workers with interest in these applications are likely to be our primary readers.

a) PDMS layer and working distance
The working distance of our 63x NA1.4 objective is 100 microns (when used with a ~170 microns coverslip), as defined by Leica. So, practically, using a 40 microns thick PDMS layer allowed us to perfectly focus on the cells (placed on the top the PDMS layer). The highest focus point was usually close to the top of the membrane.

A sentence has been in added in the Methods section to point out this potential issue.

b) Usability of the device with other cell types
Besides S. cerevisiae, we have found that E. coli cells succesfully grow in the device. However, no trial has been made with mammalian cells. As pointed by the reviewer, it could be of interest to keep larger non-adherent eukaryotic cells under physical constraint for multi cell-cycle time-lapse assays. We currently don't know which cell type -if any- could handle the mechanical stress imposed by our device.
A short paragraph has been added to the manuscript describing potential applications to other organisms.

c) The importance of the PDMS membrane
We have observed that the PDMS layer is necessary with the dialysis membrane to ensure that cells grow well in the device, probably because PDMS is soft enough (as opposed to glass) to cushion the cells. Yeast will grow on glass when confined under an agar gel pad, which is easily deformable. The only prior work we know of that used a dialysis membrane for confinement is Ref [11] where there was also a PDMS layer between the cells (E. coli) and the coverslip. The PDMS in [11] had groves in it where the cells grew (which facilitated imaging) and so there was probably less pressure of the membrane on the cells. However this refinement is not necessary for yeast and indeed would be cumbersome, and we have also shown that E.coli will grow in our cell. Other flow cells confine the cells in a narrow cavity between PDMS surfaces.

4- MATLAB software and image analysis methods
We have considerably expanded the data analysis section in supplementary materials in order to provide more details about :
the cell segmentation based on phase contrast images
the tracking of cells across time point using a dedicated graphical user interface
the quantification of fluorescence signals

a) figure 4 was modified for consistency with the text : 1xMet wass used to block the cells in this experiment, because the experiment was done before we realized that complete shutoff required a bit more Met. Since we are only looking for sufficient shutoff to attain a block (which obviously occurs with 1X Met based on the data in the figure), this has no effect on the interpretation of the experiment.
b) The figure was modified in accordance with the useful suggestions of the reviewer.
c) According to our measurements, WHI5 exit in mother cells occurs on average about 1 minute after cytokinesis. Since the measurement interval we used is 3 minutes, we can get only an estimate of mean and almost no idea of real variability – almost all the observed variability will necessarily be experimental error. However, using a shorter interval is possible at the expense of acquiring less fields of cells per time-point or by suppressing the autofocusing routine (about 10s per field of cells). The latter option is only suitable for very short term movies, since thermal drifts move the focal plane. Also, in our previous work (Bean et al., 2006) we found that higher-frequency imaging than every 3 min became seriously deleterious to long-term growth of the cells, presumably due to cumulative or sporadic photodamage, and therefore we have avoided shorter intervals despite the necessary loss of temporal resolution.
d) Done.