Imaging dynamic processes: a centriole case study
Submitted by fdavid on 26 November, 2018.
1. Sir William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford, OX1 3RE
Alan did his undergraduate degree at the University of Cambridge and was a PhD. student with David Glover (Cambridge). He then worked as a Post-doctoral Fellow with Maurizio Gatti (Rome) and James Wakefield (Oxford) before joining the Raff laboratory. He runs the light microscopy part of the Dunn School Bioimaging Facility, which is part of Micron Oxford.
Super-resolution methods such as Structured Illumination Microscopy (SIM) and Single Molecule Localisation Microscopy (SMLM) have allowed many cellular structures to be imaged at increasing detail. However, these studies are incomplete without an understanding of how molecular components are dynamically incorporated. Here I use the centriole of Drosophila as a case study for how a biological question can drive the development of innovative ways of combining super-resolution techniques and image analysis to provide a dynamic view of subcellular processes. Specifically I describe how we combined 3D-SIM and SMLM to assess protein localisation; and two colour SIM (and Airyscan) with FRAP in living embryos to examine new molecule incorporation.
I would like to thank Jordan Raff, in whose laboratory this work was carried out. Lisa Gartenmann, Mustafa Aydogan, Thomas Steinacker, and Ian Dobbie who performed many of the experiments described. Members of the Raff laboratory and Micron Oxford for their comments. This research was funded by the Wellcome Trust Strategic Award (107457) to Micron Oxford, and a Wellcome Trust Senior Investigator Award to Jordan Raff (104575).
In the last few years, the super-resolution revolution has allowed many cellular structures to be imaged at increasing detail.1 However, biology is complicated. Sometimes these complexities are not revealed by simply developing a more powerful microscope. As biological structures and processes are seldom static, developing methodologies to image the dynamics of growth and change beyond the diffraction limit is likely to be the next challenge.
In this article I will use the centriole of Drosophila melanogaster as a case study for how a biological questions can drive the development of innovative ways of combining super-resolution techniques and image analysis to provide a dynamic of view of a subcellular process.
CENTRIOLES IN DROSOPHILIA
Centrioles are microtubule-based cylindrical structures that form two subcellular organelles in most eukaryotic cells. One of these structures is the cilium, a protrusion from the cell surface involved in movement or signalling; the second is the centrosome, the major microtubule organising centre in many cell types. Centriole dysfunction in humans can lead to congenital diseases such as microcephalies and ciliopathies, as well as other pathologies like obesity and cancer.2
Unlike other organelles, such as the endoplasmic reticulum and the Golgi, centrioles display a rigid structure.3 Their intricate 9-fold symmetry has long been revealed by electron microscopy (EM), but due to their small size the observation of these details using light microscopy has long eluded the field [Fig 1A]. In addition to their beautiful sub-resolution structure, centrioles are also interesting as their formation is a highly dynamic process, tightly coupled with the cell cycle. Each centriole will template the formation of a ‘daughter’ centriole, that grows at right angles to the ‘mother’ once per cell cycle, during G1/S phase.4
In the Raff laboratory we study centriole structure and formation in the fruit fly Drosophila melanogaster. Drosophila is a good biological system for a variety of reasons—comprehensively covered elsewhere5—but several of its tissues, such as the embryo, are particularly well suited for light imaging of centrioles, as described in more detail below.
Figure 1. (A) Cartoon of centriole growth. A younger ‘daughter’ centriole grows at right angles to the older ‘mother’ centriole (identified by the presence of a ring of Asl shown in red). Looking down the cylinder of the centriole, the core of the centriole contains a 9-fold radially symmetrical ‘cartwheel’ structure. The cartwheel is arranged in stacks along the length of the centriole. (B) Centriole duplication requires 5 essential proteins, which can be organised into a pathway. (C) 3D-SIM microscopy reveals the localisation of the core duplication proteins. Sas-6, Ana2 and Sas-4 are seen on both mother and daughter centrioles, whereas Asl is a ring around the mother only. Bar 1μm.
SPATIAL CONTEXT: COMBINING 3D-SIM, SMLM AND IMAGE ANALYSIS
Centriole duplication in flies requires 5 proteins: PLK4, Sas-6, Ana2, Sas-4 and Asl [Fig 1B].2 At standard resolution these 5 components appear as dots [Fig 1C]. Centrioles were one of the first organelles to be imaged in detail using super-resolution techniques allowing the centriolar localisation of some of these crucial proteins to be determined.6-9 3D-Structured Illumination Microscopy (3D-SIM) shows that Sas-6, Ana2 and Sas-4 localise to both the mother and growing daughter centrioles;10-11 Asl localises as a ring around the mother centriole12 and Sak localises as a single dot on the Asl ring at the base of the growing daughter [Fig 1C].
While imaging using 3D-SIM allowed us to examine the localisation of these proteins in more detail, it was still not possible to determine their exact localisation within the intricate structure of the centriole that can be observed by EM. For example, crystal structure analysis of Sas-6 shows that it is the main component of the cartwheel, forming the core of the centriole.13,14 However, 3D-SIM still did not allow us to resolve Sas-6 from a dot (albeit at a more precise localisation). In addition, imaging by 3D-SIM did not provide details on the orientation of each protein within the centriole structure (namely the orientation of the N and C termini). Once again Sas-6 is a good example. We know from the crystal structure that its N-terminus should be at the core of the centriole while the C-terminus should be pointing outwards, but we were not able to confirm this by imaging techniques.13,14 It was therefore necessary to develop a methodology to examine this with more precision than that allowed by 3D-SIM.
Attempts using higher resolution light microscopy techniques like direct stochastical optical reconstruction microscopy (dSTORM), a type of SMLM, were only partially successful. Due to the cylindrical shape of centrioles, being able to precisely determine the localisation of a protein within it requires imaging precisely down the centre of the cylinder. Measurements of any off axis centrioles would inevitably reduce the accuracy. To overcome this limitation we developed a new methodology, SIM-SMLM, where 3D-SIM was used to assess the orientation of centrioles, and SMLM allowed protein imaging at higher precision.15 We took advantage of the equipment and expertise of Micron Oxford [micron.ox.ac.uk].
As shown in Fig 2, Asl localises to a ring around the mother centriole, making it an ideal marker for centriole orientation. We first imaged centrioles using 3D-SIM, looking at Asl localisation (labelled with a primary antibody and Alexa 594 secondary). A ring Gaussian was fitted to the maximum intensity projection for each centriole imaged, producing a localisation model. We then measured the major and minor axis of the resulting model. When observed ‘end-on’ the centriole is a perfectly circular structure, so we can conclude that centrioles where this ratio is greater than ~1.2 were significantly misoriented, and so were discarded from our analysis. This allowed us to be sure that only near-perfectly oriented centrioles were selected for further SMLM imaging [Fig 2A].
We tagged the proteins of interest with GFP and labelled this with anti-GFP nanobodies. This made the sample amenable to SMLM microscopy and reduced the “linkage” error compared with standard antibodies, due to the smaller size of nanobodies (5nm, as opposed to a primary-secondary antibody conjugate that can reach 20nm, leading to a shorter distance between protein and fluorophore). These were then imaged on the SMLM light path of our microscope.16
One issue with combining these two techniques is that the 3D-SIM and SMLM images are captured on two different cameras, which results in two images that are not perfectly aligned. Manual alignment of the two images was not ideal as, at this level of resolution, even the most minor aberrations in the two light paths can create local deformations (i.e. the misalignment problem cannot be easily solved by simply moving one of the images across until all points align). To overcome this issue, and to then measure the subsequent SMLM localisations, we developed a python script (Centroid Origin Optimising Localiser – github.com/MicronOxford/cool). Briefly, we used the centre of the 3D-SIM Asl rings to provide us with a starting point regarding the localisation of centriole of interest. The script detected the closest SMLM localisation ‘cloud’ to this central point and identified it as being the same centriole as that imaged in 3D-SIM. The script then draws a radius of 90nm around the SIM central point to identify all SMLM localisations within that radius. The script then moves the centre of the radius to the localisation that includes most SMLM localisations in that cloud. It then draws a new radius and determines again whether the localisation of the centriole centre matches the SMLM localisations of that particular cloud. This iterative process allowed us to identify precisely the centre of the SMLM cloud corresponding to the initial 3D-SIM ring imaged. This analysis was done using a weighted approach, as some of the SMLM localisations are more precise than others [Fig 2B].15
Once the SMLM ‘cloud’ centre is identified, a ring Gaussian is fitted to all the localisations and the radius of that ring is measured, providing the precise localisation of the protein terminus of interest. This approach allowed us to determine the localisation of our proteins of interest within the centriole structure to the unprecedented level of 4-5 nm precision [Fig 2C].
Figure 2. (A) Assessment of centriole orientation using ellipse analysis of Asl SIM images. Only centrioles with a fitted ellipse with a major to minor axis of less than 1.2 are orientated and considered for further analysis. Bar 200nm. (B) SMLM imaging of GFP-tagged centriole proteins. Script identifies centre of SMLM localisations for each centriole and measures radius of centriole signal. (C) Localisation of N and C-termini of centriole proteins shown in green, with Asl SIM image in red. Bar 400nm. Figure adapted from Gartemann et al. 2017 with permission.
DYNAMICS BEYOND THE DIFFRACTION LIMIT
Increasing resolution is, of course, important, and as shown in the previous section it is possible to improve on standard resolution techniques if required by the biological question. However, as most biological processes are dynamic, being able to observe them as they happen at super-resolution can be vital. That is the case for centriole duplication. In the previous section we were able to determine with unprecedented resolution where the centriolar components localise. However, we still do not know how these components are dynamically incorporated into the structure as it is being built. To this end we applied two approaches: measuring the rate of centriole growth and then combining super-resolution with Fluorescence Recovery After Photobleaching (FRAP).
For this analysis we used the Drosophila embryo, a particularly well-suited system for dynamic imaging of centriole growth. In the first two hours of development the Drosophila embryo develops as a syncytium (meaning that the nuclei divide within a common cytoplasm), where nuclear cycles occur sequentially and synchronously.17 As described above, the cycle of centriole growth and duplication is tightly linked to the cell cycle, so synchronous cell cycles during the embryo development correspond to synchronous cycles of centriole growth and duplication. This allows for multiple events in the exact same stage of the cycle to be imaged simultaneously – providing many points for quantification in each experiment. In addition, at one-and-a-half hours of development the nuclei and associated centrioles move to the cortex of the embryo (closest to the coverslip) making them particularly amenable for good quality imaging [Fig 3A].17 Each cycle lasts 15-20 min (compared to many hours in cell culture), allowing for several ‘rounds’ of centriole duplication to be imaged within the time frame of an experiment. An additional, and particularly convenient feature of the Drosophila embryo is that the older mother centrioles orientate towards the cortex (towards the coverslip) allowing us to ‘look down’ the centriole cylinder [Fig 3A].10 If the mother centrioles are all orientated towards the coverslip, then it follows that the daughter centriole growth can be observed at right angles from the signal marking the mother (as illustrated in Figure 3B).
Figure 3. (A) Workflow of Drosophila embryo imaging. Embryos laid by female Drosophila are aged for 1½ to 2 hours and then placed on a microscope. At this time the embryos develop as a syncytium, with many cell cycles and hence centriole cycles happening in synchrony. The centrioles are also all located at the embryo cortex and so near to the coverslip, and oriented so that most of the centrioles orientated towards the cortex. Bar 5μm. (B) Cartoon of centriole length measurements. The centriole ‘capping’ protein Cep97 localises to the distal tip of both the mother and daughter centriole. Measuring the distance between Cep97 signal on the mother and daughter centriole can be used as a proxy for centriole growth. (C) Graph showing how centrioles grow over time. Figure adapted from Aydogan et al 2018. Originally published in Journal of Cell Biology. https://doi.org/10.1083/jcb.201801014.
In the first technique used to image the dynamic growth of centrioles, we took advantage of this arrangement. If we consider a centriole component that localises to the tip of the centriole, e.g. Cep97, at super-resolution it appears as ‘dot’ in the mother centriole. As the daughter centriole grows on the side of the mother, a second dot appears. Using a Zeiss Airyscan system followed by subsequent image analysis, we were able to measure the precise centre of the growing daughter centriole signal, and determine how the distance between this signal and that of the mother increases over time [Fig 3B]. This provided us with a proxy measurement of centriole growth [Fig 3C]. Whilst other super-resolution techniques might provide a higher resolution image, the Airyscan system requires much less light than other super resolution systems, and so allowed us to follow the process for longer, without considerable phototoxicity or bleaching of the fluorescent signal.
Whilst we were able to measure the rate of centriole growth, we still did not obtain information regarding how centriole component molecules were incorporated into the structure as it grew. We were interested in two questions: (1) Are any centriole components stably incorporated into the growing centriole; (2) Where along the length of the daughter centriole are the molecules being incorporated?
To address these two questions we turned to FRAP, where a high energy laser is used to bleach the fluorescent signal, and signal recovery is monitored over time as unbleached molecules are incorporated, or exchanged with bleached molecules, into the structure from the cytoplasm.
In order to address whether centriole components are stably incorporated into centrioles, or whether they are dynamically turning over, we only required imaging of two timepoints—immediately before bleaching, and a defined timepoint after bleaching. As bleaching over time was not a concern, we were able to take advantage of the higher resolution offered by an OMX 3D-SIM system. GFP was fused to the centriole protein of interest, and used to label both mother and daughter.
Once daughter centriole growth had reached approximately half of its total length, the GFP signal was bleached. We then waited for five minutes and imaged again to see whether the signal had recovered. If the protein of interest is being stably incorporated into the structures then you would expect the mother centriole not to have recovered its signal (since it is not growing) whilst you would expect the daughter centriole to recover some signal (since the structure is still growing).
If the component is turning over then you should expect the signal to recover in both mother and daughter. Using this approach we showed that Sas-6 and Sas-4 are stably incorporated into centriole structures (only new incorporation in the daughter) while Ana2 turns over (signal recovered in both mother and daughter) [Fig 4A].
Figure 4. (A) Fluorescence Recovery After Photobleaching (FRAP) of centriole proteins. Postbleach image of Sas-6-GFP and Sas-4-GFP have new GFP signal in only one centriole. This represents new molecules being incorporated into the daughter centriole as it grows. Postbleaching image of Ana2-GFP has signal in both daughter and mother centriole suggesting that Ana2 is not stably incorporated into the centrioles, but is turning over dynamically. Bar 200nm. (B) FRAP of Sas-6-GFP protein to assess where new cartwheel molecules are incorporated into the growing daughter centriole. After the daughter centriole has grown to approximately half its final length the GFP signal is bleached. After one minute of recovery, the position of the new daughter Sas-6 GFP signal is measured. Three possibilities were envisaged (i) either the first recovery was closer to the mother centriole than before bleaching (ii) roughly the same or (iii) more distant than before bleaching. The results show that the first recovery point is closer to the mother than prior to beaching strongly suggesting that new molecules are incorporated at the base of the daughter centriole as it grows. Figure adapted from Aydogan et al 2018. Originally published in Journal of Cell Biology. https://doi.org/10.1083/jcb.201801014 and adapted from Conduit et al 2015 with permission.
To address the second question, namely where along the length of the daughter new molecules are being incorporated during growth, we performed similar experiments using a Zeiss Airyscan system. This particular application required a higher level of channel alignment, rather than just high spatial resolution. As the Airyscan is a point scanning system with a single detector, each colour can be sequentially captured, per line, onto the same detector ensuring close alignment between green and red channels.
We performed a similar FRAP experiment to the one described above. However, rather than waiting five minutes, we captured the first image after one minute recovery period. We then measured the distance between the centre of the mother signal (in the red channel) and the centre of the very first signal detected in the daughter in the green channel.
In a field of view only a proportion of the centrioles is bleached using this procedure, and the unbleached centrioles provide an important control. As is the case for the bleached centrioles, we also measure the distance between the mother signal and the centre of the GFP signal in the daughter centriole. The difference in the distance in these two cases is crucial.
If new molecules are being incorporated at the base of the daughter centriole (closest to the mother) then we should expect the distance in the bleached centrioles to be shorter than in the control centrioles. If new molecules are incorporated at the tip of the centriole then we should expect the distance to be longer in the bleached centrioles. If molecules are incorporated along the entire structure then we should expect the distance to be roughly the same. Our experiments showed that Sas-6 is incorporated at the base of the daughter centriole (closest to the mother), a result that overturned a long-standing assumption that the centriole cartwheel grows from the distal end [Fig 4B].10
With the development of increasingly sophisticated microscopes, the temptation is often to apply the newest technique to our biology of interest. However, as I have illustrated with these examples, the choice of microscopy method, and how they are combined and developed, should be driven by the biological question at hand.
Furthermore, the development of image analysis methods can be integral to the imaging, not only allowing one to overcome technical obstacles but also to detect subtle biological effects. In the first section I showed how combining two super-resolution techniques was necessary to assess centriole orientation enabling a high precision in localisation. In the second example I showed how dynamic information was essential to understanding the question of centriole growth, and that combining super resolution techniques with FRAP was more important than necessarily using the highest resolution microscope in our repertoire.
Biological structures and processes are by their nature dynamic. Observing these dynamic processes at super-resolution may turn out to be as informative as achieving increasingly higher resolutions. The approaches described here can be applied to the study of other dynamic organelles such as mitochondria, ER, Golgi or the kinetochore. However, the leading principle should always be that the biological question drives the use of the technique.
1. L. Schermelleh, R. Heintzmann, and H. Leonhardt, “A guide to super-resolution fluorescence microscopy.,” J. Cell Biol., vol. 190, no. 2, pp. 165–175, Jul. 2010.
2. P. T. Conduit, A. Wainman, and J. W. Raff, “Centrosome function and assembly in animal cells.,” Nat. Rev. Mol. Cell Biol., vol. 16, no. 10, pp. 611–624, Oct. 2015.
3. P. Guichard, V. Hachet, N. Majubu, A. Neves, D. Demurtas, N. Olieric, I. Fluckiger, A. Yamada, K. Kihara, Y. Nishida, S. Moriya, M. O. Steinmetz, Y. Hongoh, and P. Gönczy, “Native architecture of the centriole proximal region reveals features underlying its 9-fold radial symmetry.,” Curr. Biol., vol. 23, no. 17, pp. 1620–1628, Sep. 2013.
4. E. A. Nigg and A. J. Holland, “Once and only once: mechanisms of centriole duplication and their deregulation in disease.,” Nat. Rev. Mol. Cell Biol., vol. 19, no. 5, pp. 297–312, May 2018.
5. E. Bier, “Drosophila, the golden bug, emerges as a tool for human genetics,” Nat Rev Genet, vol. 6, no. 1, pp. 9–23, Jan. 2005.
6. J. Fu and D. M. Glover, “Structured illumination of the interface between centriole and peri-centriolar material.,” Open Biol, vol. 2, no. 8, pp. 120104–120104, Aug. 2012.
7. K. F. Sonnen, L. Schermelleh, H. Leonhardt, and E. A. Nigg, “3D-structured illumination microscopy provides novel insight into architecture of human centrosomes.,” Biol Open, vol. 1, no. 10, pp. 965–976, Oct. 2012.
8. S. Lawo, M. Hasegan, G. D. Gupta, and L. Pelletier, “Subdiffraction imaging of centrosomes reveals higher-order organizational features of pericentriolar material.,” Nat. Cell Biol., vol. 14, no. 11, pp. 1148–1158, Nov. 2012.
9. V. Mennella, B. Keszthelyi, K. L. McDonald, B. Chhun, F. Kan, G. C. Rogers, B. Huang, and D. A. Agard, “Subdiffraction-resolution fluorescence microscopy reveals a domain of the centrosome critical for pericentriolar material organization.,” Nat. Cell Biol., vol. 14, no. 11, pp. 1159–1168, Nov. 2012.
10. M. G. Aydogan, A. Wainman, S. Saurya, T. L. Steinacker, A. Caballe, Z. A. Novak, J. Baumbach, N. Muschalik, and J. W. Raff, “A homeostatic clock sets daughter centriole size in flies.,” J. Cell Biol., vol. 217, no. 4, pp. 1233–1248, Apr. 2018.
11. P. T. Conduit, A. Wainman, Z. A. Novak, T. T. Weil, and J. W. Raff, “Re-examining the role of Drosophila Sas-4 in centrosome assembly using two-colour-3D-SIM FRAP.,” Elife, vol. 4, p. 15915, Nov. 2015.
12. Z. A. Novak, P. T. Conduit, A. Wainman, and J. W. Raff, “Asterless licenses daughter centrioles to duplicate for the first time in Drosophila embryos.,” Curr. Biol., vol. 24, no. 11, pp. 1276–1282, Jun. 2014.
13. M. van Breugel, M. Hirono, A. Andreeva, H.-A. Yanagisawa, S. Yamaguchi, Y. Nakazawa, N. Morgner, M. Petrovich, I.-O. Ebong, C. V. Robinson, C. M. Johnson, D. Veprintsev, and B. Zuber, “Structures of SAS-6 suggest its organization in centrioles.,” Science, vol. 331, no. 6021, pp. 1196–1199, Mar. 2011.
14. D. Kitagawa, I. Vakonakis, N. Olieric, M. Hilbert, D. Keller, V. Olieric, M. Bortfeld, M. C. Erat, I. Fluckiger, P. Gönczy, and M. O. Steinmetz, “Structural basis of the 9-fold symmetry of centrioles.,” Cell, vol. 144, no. 3, pp. 364–375, Feb. 2011.
15. L. Gartenmann, A. Wainman, M. Qurashi, R. Kaufmann, S. Schubert, J. W. Raff, and I. M. Dobbie, “A combined 3D-SIM/SMLM approach allows centriole proteins to be localized with a precision of ∼4-5 nm.,” Curr. Biol., vol. 27, no. 19, pp. R1054–R1055, Oct. 2017.
16. E. Johnson, E. Seiradake, E. Y. Jones, I. Davis, K. Grünewald, and R. Kaufmann, “Correlative in-resin super-resolution and electron microscopy using standard fluorescent proteins.,” Sci Rep, vol. 5, p. 9583, Mar. 2015.
17. V. E. Foe and B. M. Alberts, “Studies of nuclear and cytoplasmic behaviour during the five mitotic cycles that precede gastrulation in Drosophila embryogenesis.,” J. Cell. Sci., vol. 61, pp. 31–70, May 1983.