Brief Communication | Published:

In vitro whole-organ imaging: 4D quantification of growing mouse limb buds

Nature Methods volume 5, pages 609612 (2008) | Download Citation



Quantitative mapping of the normal tissue dynamics of an entire developing mammalian organ has not been achieved so far but is essential to understand developmental processes and to provide quantitative data for computational modeling. We developed a four-dimensional (4D) imaging technique that can be used to quantitatively image tissue movements and dynamic GFP expression domains in a growing transgenic mouse limb by time-lapse optical projection tomography (OPT).


In addition to detailed observations of cells of multicellular organisms, tissue-level data on the global behavior of such a system1 would be extremely valuable to researchers. Comprehensive imaging has been achieved by confocal approaches in models such as sea urchin2 and zebrafish3 because of their small size, transparency and well-established culture techniques. However, an accurate quantitative description of normal tissue movements and dynamic gene expression for a complete mammalian organ has so far not been generated, owing to both limitations in culture techniques and inherent restrictions in established 4D imaging technologies: confocal and multiphoton microscopy are ideal 4D fluorescence imaging approaches for cells and tissues but are not suited to imaging whole organs because of their limited imaging depth. By contrast, optical coherence tomography, ultrasonic biomicroscopy and magnetic resonance imaging can be used to image macroscopic samples but cannot detect fluorescent signals. Because of these restrictions, tissue movements have largely been deduced by traditional fate-mapping studies4 using carbon particles or fluorescence-labeling techniques in the mouse5 and chick embryo6, and 4D dynamics of gene expression have been deduced from static ex vivo specimens7. In contrast, OPT has successfully been used for whole-organ imaging of ex vivo specimens8. We therefore explored whether OPT could be developed and combined with a new organ culture technique to allow quantitative and spatially comprehensive 4D imaging of a growing mouse organ, the limb bud.

To this end, we had to overcome several problems. First, in existing organ culture techniques, the specimen either floats in medium in a rolling chamber or rests on a support membrane ('static culture'). Both of these arrangements are unsuitable for OPT imaging: the former because the raw images must be captured at precisely controlled angles around the specimen, and the latter because many of these carefully controlled views would be obstructed by the membrane. Second, static culture tends to produce abnormal growth shapes because of mechanical pressure from the support, whereas our goal was to quantitatively describe normal morphogenetic movements. To address these problems, we developed a new 'semi-static' culture apparatus, in which we pinned the abdomen of a mouse embryo to a minimal-support rotary arm using tungsten needles (Fig. 1). Culturing the abdomen instead of the whole embryo was necessary to allow imaging 360 degrees around the limb bud and also gave the growing tissues more access to nutrients and oxygen. The apparatus rotates the limb bud around a vertical axis, capturing datasets every 15 min (from a horizontal objective lens, precisely perpendicular to the axis of rotation) while the tissue grows in the medium. Third, high-quality tomographic imaging depends on optimizing both the position and the angle of the rotating limb bud with respect to the axis of rotation to maximize the effective focal depth9. We therefore devised a micromanipulator that contains 6 mechanical degrees of freedom so that the specimen can be accurately aligned while rotating in the imaging chamber (Fig. 1a–d). Fourth, to compensate for the fact that the limb bud is not placed at a medium-air interface10,11,12, we developed an oxygen delivery system by placing a layer of perfluorodecalin (which displays up to 100 times more affinity for oxygen than plasma) directly beneath the limb bud, under the medium (Fig. 1a and Supplementary Methods online). Oxygen is slowly released into the medium, such that after an initial equilibration period, it is maintained at 20–30% over about 20 h (Fig. 1e).

Figure 1: The 4D time-lapse OPT.
Figure 1

(a) The design features of the time-lapse OPT apparatus (for details see Supplementary Methods). Tissue to be cultured is pinned to the mount while it is in the dissecting dish and is protected during transport into the imaging chamber by the plastic capsule. Three liquids are placed into the imaging chamber: perfluorodecalin to supply oxygen, medium and mineral oil to prevent evaporation. (b) Tungsten needles are pinned through the trunk to minimize both the mechanical impact on the specimen and interference with tomographic imaging. Scale bar, 1 mm. (c,d) The diagrams of the apparatus illustrate how the mount engages with the conical support (c) and how it can be tilted to allow accurate positioning of the limb bud (d). (e) Change in oxygen levels (top) and relative proximodistal limb bud growth for three experiments (bottom). The dashed line is an estimate of in utero growth.

We tested limb growth in vitro using our culture technique. Although gross changes in size have previously been reported for cultured limbs10, neither have accurate measurements of size and shape been reported, nor has the tissue been monitored by time-lapse imaging to determine the period of successful in vitro growth. Instead, previous studies have routinely used the organ as a 'living test tube' to assess the effects of ectopic expression of a given gene on possible downstream targets11.

By contrast, we wanted to measure the normal tissue dynamics of the whole organ. We took advantage of the isotropic three-dimensional (3D) resolution of OPT data to compare the shapes of cultured limb buds with carefully staged controls that we scanned by OPT immediately after dissection (Fig. 2a–f; for examples of older specimens, see Supplementary Methods). The growth diverged on average after 6 h of culture. After this time shape changes were still similar to those in utero (and ectoderm integrity routinely remained intact for over 24 h), but the proximodistal growth rate decreased (Fig. 1e). This is an intrinsic problem of whole-organ in vitro culture experiments and could be due to a lack of oxygen or a lack of waste removal for the internal mesenchyme, possibly leading to necrosis. Using our culture technique, however, normal in utero growth rate and 3D shape changes were possible for limb buds grown in vitro, and we attempted to map global dynamics within the 6-h window of normal morphogenetic movements.

Figure 2: Tracking growth for 6 h (E11.0–E11.25).
Figure 2

(af) Dorsal (ac) and anterior (df) views of a representative cultured limb bud (starting shape is green, end of culture is red) compared to carefully staged freshly dissected limbs (blue). (g,h) Imaging fluorescent landmarks on the ectoderm (g), allowed 3D tracking of these points over time (red arrows, h), and the extraction of velocity vector fields over the surface of the limb (green arrows) by interpolation using radial basis functions (Supplementary Methods). Asterisks highlight the two inverted vortices. (i) Overlap of data from two additional experiments with more extensive labeling (pink versus blue arrows). (j,k) Surface expansion was estimated by calculating the increase in area for triangulated regions and averaging the 8 experiments at defined sampling points. Scale bar, 200 μm.

Despite continued advances in understanding the molecular patterning events important for limb development13, very little is understood about how individual cell behaviors are mechanically coordinated into global tissue movements. Instead of recording individual cell activities (which is already possible using confocal microscopy), in this work we measured tissue movements of the whole organ; in other words, we obtained a geometric transform for the organ shape. We placed fluorescent microspheres on the ectoderm of early embryonic stage E11.5 hind limbs at the time when autopod formation is initiated to provide landmarks for accurate 3D tracking of ectodermal movements (Fig. 2g–i and Supplementary Movies 1, 2, 3 online). For a discussion of the evidence that microsphere movements reflect ectodermal tissue movements, see Supplementary Note online.

The discrete nature of the microspheres allowed accurate tomographic reconstruction of their 3D positions from 100 views per time point (every 3.6 degrees). It is known that development of the autopod involves expansion along the proximodistal and anteroposterior axes13, and we therefore expected to observe radially expanding tissue movements. In our experiments (n = 8), however, we observed a new feature. While the anteroposterior width of the autopod was expanding, the tissue in the most anterior and posterior regions, rather than simply moving distally (away from the body) or radially (away from the center of the bud), actually twisted as it grew, curving around two almost fixed points. A complete velocity vector field can be interpolated from the moving landmarks, highlighting the power of these data for computer modeling and also pinpointing the centers of the vortices (Fig. 2h). Two additional experiments with more extensive labeling (Fig. 2i) confirmed the interpolated velocity field.

We characterized the ectodermal growth pattern by estimating surface expansion rates. We wrote software to create a surface triangulation based on the fluorescent landmarks and to calculate the isotropic expansion rates of regions. The averaged results (for the same eight experiments as above) yielded a quantitative map of the locally varying expansion rates across the limb (Fig. 2j,k) and showed a surprising degree of spatial variation: expansion rates were very high near the apical ectodermal ridge (up to 80% increase in surface area over 6 h) and were dramatically lower on the central dorsal and ventral sides (as low as 20%). Notably, the ectoderm near the ridge did not have uniform expansion rates: growth rates were highest at the distal tip and gradually decreased toward the anterior and posterior ends of the ridge. As cell sizes and cell density do not change substantially during this 6-h period (Supplementary Note), and cell death rates are very low at this stage of development14, this suggests a cell cycle time of less than 7 h, whereas the slower dorsal, ventral and proximal regions have an estimated cell cycle time of more than 24 h.

Finally, we explored whether this approach could be suitable for imaging dynamic changes in spatial gene expression patterns over time within the mesenchyme. Owing to photon scattering in biological tissue, confocal microscopy, OPT and other optical imaging modalities are limited by the balance between imaging depth and imaging resolution, so a major technical question for this approach was whether biologically useful 3D images could be obtained. We chose a transgenic mouse line in which GFP is under control of the Scx promoter15. Scx gene activity is dynamically patterned across the mesenchyme during the process of defining which cells will form tendons and digital condensations15.

With our time-lapse OPT technique, we observed the Scx-GFP expression pattern changing over time. At the beginning of the culture, the GFP signal could be seen on the dorsal and ventral sides of the autopod, restricted to the medial regions where tendon specification is starting (that is, away from the anterior, posterior or distal edges; Fig. 3). During 19 h of culture, the fluorescent domain increased in size and changed shape (Fig. 3b,c,e,f,h,i and Supplementary Movies 4, 5, 6 online). Although limb growth was reduced after the initial 6 h, the molecular patterning process continued up to 19 h, in line with previous studies11. At the tissue level, the fluorescence intensity was a graded continuum (Fig. 3a–c), which is due to both different expression levels and different ratios of expressing and non-expressing cells. However, to illustrate the changing 3D shape of this dynamic domain, we chose a single threshold isosurface value to create the surface rendering, which captured the spatiotemporal dynamics of the process. In particular, the emergence of digits 4 and 3 (Fig. 3b,e,h) preceded that of digit 2. The process appears to involve gene activation in the digital domains (Fig. 4), but cell movement could also be involved as our images do not display sufficient resolution to distinguish between the two processes.

Figure 3: The 4D monitoring of a dynamic gene expression pattern (Scx-GFP) in limb buds.
Figure 3

(ac) Raw projection images from one angle at the indicated time points during limb bud culture. (di) 3D rendering of the gene expression pattern at the same time points, from 2 different angles: a dorsal view (df) and a distal end-on view (gi). Scale bars, 200 μm.

Figure 4: Comparison of physical sections with OPT sections.
Figure 4

(ae) Images of different positions along the proximodistal axis (top row is most distal, bottom row is most proximal) of physical paraffin sections (a,b,d) in which the transgenic GFP signal has been imaged directly in a freshly dissected embryo at E11.5 (a is a close-up of boxed region b), and at E12.5 (d), or virtual OPT sections (c,e) at roughly equivalent proximodistal positions to the adjacent paraffin sections, at 0 h (c) and 19 h (e) of in vitro culture. Scale bars, 50 μm (a) and 200 μm (b).

The resolution of all optical imaging techniques is limited by intrinsic properties of the tissue as well as sample shape and size. To determine how well the tomographically reconstructed images represented the true fluorescence pattern, we cut paraffin wax sections from equivalent-aged Scx-GFP transgenic mouse embryos and captured fluorescence for E11.5 and E12.5 embryos, which correspond to the start and end of our time-lapse experiments (Fig. 4). As expected, the resolution of our reconstructions was considerably lower than that of directly imaged sections. However, despite the intrinsic scattering caused by the thick, living tissue, the pattern of fluorescence recorded by OPT corresponded well to the domains of expression observed with physical sections, even though certain features, such as the absence of expression in the center of some physical sections, are not clear from the OPT data. As our method monitors tissue dynamics in live samples, our results nevertheless present an important new tool in the ongoing endeavor to map vertebrate organogenesis. The technique is not restricted to the study of limb development. We also performed 4D OPT to image the developing eyes and brain of an E9.5 embryo, using a transgenic Pax6-GFP mouse line (Supplementary Note).

The ability to monitor growth and gene expression in three dimensions over time is a technical step forward, which we believe will be valuable for fully understanding organogenesis and will serve as a quantitative basis for computational modeling of organ development.

Note: Supplementary information is available on the Nature Methods website.


  1. 1.

    & Cell 130, 784–795 (2007).

  2. 2.

    , , & Dev. Biol. 299, 543–550 (2006).

  3. 3.

    et al. Development 134, 1011–1022 (2007).

  4. 4.

    & Nat. Cell Biol. 1, E103–E109 (1999).

  5. 5.

    , & J. Exp. Zool. 249, 50–54 (1989).

  6. 6.

    et al. Development 124, 1909–1918 (1997).

  7. 7.

    , , & Evol. Dev. 7, 18–28 (2005).

  8. 8.

    et al. Science 296, 541–545 (2002).

  9. 9.

    Annu. Rev. Biomed. Eng. 6, 209–228 (2004).

  10. 10.

    & Nature 361, 68–71 (1993).

  11. 11.

    , , & Nature 401, 598–602 (1999).

  12. 12.

    , , , & J. Cell Biol. 130, 997–1003 (1995).

  13. 13.

    Dev. Cell 4, 449–458 (2003).

  14. 14.

    , & Dev. Dyn. 235, 2521–2537 (2006).

  15. 15.

    , , , & Dev. Dyn. 236, 1677–1682 (2007).

Download references


We thank M. Logan (MRC National Institute for Medical Research) for providing the Scleraxis-GFP line, the Zeller lab for their time and helpful suggestions on the in vitro mouse limb bud culturing technique, L. Hay for technical support in the design and building of the 4D time-lapse OPT scanner, D.A. Kleinjan and V. van Heyningen for Pax6-GFP embryos, C. DeAngelis for performing the Fgf-8 in-situ hybridization, B. Pryce for sectioning the Scx-GFP limbs, J. Swoger for help with 3D reconstructions, and the Edinburgh Mouse Atlas Project (EMAP) for software to support the computational analysis. This project was supported as part of the EU Integrated Project grant 'Molecular Imaging' (to M.J.B.), by a research grant from the Human Frontier Science Program (to C.H.W.), by a grant from the Spanish Ministry of Science and Education (BFU2006-10978/BMC; to J.S.-E. and M.T.), by MRC and Centre for Genomic Regulation (to J.C. and J.S.) and by ICREA (to J.S.).

Author information


  1. Medical Research Council (MRC) Human Genetics Unit, Crewe Road South, Edinburgh EH4 2XU, UK.

    • Marit J Boot
    • , James Cotterell
    •  & James Sharpe
  2. European Molecular Biology Laboratory–Centre for Genomic Regulation (EMBL-CRG) Systems Biology Unit, Universitat Pompeu Fabra, Dr. Aiguader 88, Barcelona 08003, Spain.

    • C Henrik Westerberg
    • , James Cotterell
    •  & James Sharpe
  3. Centro Nacional de Investigaciones Cardiovasculares, Melchor Fernández Almagro, 3, Madrid 28029, Spain.

    • Juanjo Sanz-Ezquerro
    •  & Miguel Torres
  4. Shriners Hospital, Research, 3101 SW Sam Jackson Park Road, Portland, Oregon 97239, USA.

    • Ronen Schweitzer
  5. Institució Catalana de Recerca i Estudis Avançats (ICREA), Passeig Lluis Companys 23, Barcelona 08010, Spain.

    • James Sharpe


  1. Search for Marit J Boot in:

  2. Search for C Henrik Westerberg in:

  3. Search for Juanjo Sanz-Ezquerro in:

  4. Search for James Cotterell in:

  5. Search for Ronen Schweitzer in:

  6. Search for Miguel Torres in:

  7. Search for James Sharpe in:

Corresponding author

Correspondence to James Sharpe.

Supplementary information

PDF files

  1. 1.

    Supplementary Text and Figures

    Supplementary Note, Supplementary Methods


  1. 1.

    Supplementary Movie 1

    Raw data of microsphere landmarks at 0 hours. Movie showing the raw data obtained from multiple angles during an OPT scan at the first time-point for one of the fluorescent landmark-tracking experiments.

  2. 2.

    Supplementary Movie 2

    Timelapse of microsphere landmarks over 6 hours. Timelapse movie from one angle of the same limb shown in Supplementary Movie 1.

  3. 3.

    Supplementary Movie 3

    Annotated timelapse of microsphere landmarks over 6 hours. Annotated timelapse movie from one angle of the same limb shown in Supplementary Movie 1.

  4. 4.

    Supplementary Movie 4

    Raw data of Scx-GFP expression pattern at 0 hours. Movie showing the raw data obtained from multiple angles during an OPT scan at the first time-point for one of the Scx-GFP imaging experiments.

  5. 5.

    Supplementary Movie 5

    Raw data of Scx-GFP expression pattern after 19 hours. Movie showing the raw data obtained after 19 hours in culture for the same Scx-GFP imaging experiment shown in Supplementary Movie 4.

  6. 6.

    Supplementary Movie 6

    4D timelapse of dynamic gene expression. Timelapse movie showing how the 3D domain of Scx-GFP expression changes over. (Same limb as shown in Supplementary Movie 4). All expression above a certain threshold level has been highlighted by the green isosurface.

About this article

Publication history





Further reading