Anatomy · the glass tardigrade

The X-ray tomograph

We turned an open nano-CT of one adult Hypsibius exemplaris into the glass tardigrade. We named its numbered organ labels, traced its body, measured the head scan's real voxel size and placed that scan inside the animal. The scan (270 nm per voxel, the head rescanned at higher resolution) and its hand-traced organs were published openly in 2019.

152 µmthe animal

270 nmwhole-body voxels

0.1065 µmhead voxels, measured by us

17structures placed

The data

We started from the image stacks and tracings ("label fields") of both scans, released under CC BY by Vladimir Gross, Georg Mayer, Franz Pfeiffer and colleagues with Gross et al. 2019. The tracings mark each voxel with a number, not a name.

Naming the labels

We measured every label and matched it to the organ volumes in the paper's Table 1. Label 1 is 1,763 µm³, against the pharynx's published 1,771. Label 3 is 13,812, against the gut's 13,706. Two labels share one colour in the file's palette, and together they are the Malpighian tubules, within 1% of the published volume. Structures that aren't in the table (buccal tube, stylets) were named by position and shape, and are marked "probable".

A long slice through the tardigrade CT, raw below and with coloured organ labels above — One slice down the length of the animal. Below: the raw X-ray. Above: the same slice with the published labels, coloured once we knew what they were.

Tracing the body

We traced the body, which had no label of its own. The cuticle is a bright rim in the CT, so we thresholded it, closed small gaps, and filled the interior. We kept a voxel only when at least two of the three slicing directions agreed it was inside, because the animal lay against a thin film and a curved holder that closed off false pockets between its legs. A final opening with a 1.6 µm ball removed sheets left along the film. The body comes out at 137,068 µm³, against the paper's 137,415 (0.3% under).

Cross-sections of the CT with a red body outline; a curved film touches a leg — Cross-sections near the third legs. The dark arc under the body is the mounting film that closed a false pocket beside the leg.

Shaded views of the belly before and after the fix — The belly, from below. Bottom: before, with sheets between the legs. Top: after.

The head scan, and a size that didn't add up

We measured the head scan's voxel size from its organs, which brings in the brain and salivary glands (they are only in the head scan). At 0.1065 µm per voxel, two independent organs match the paper's table exactly: brain 1,389 µm³ against 1,393, salivary glands 2,274 against 2,274. The paper states 200 nm, but at that size the organs come out far too large; the released stack was probably resampled.

To place the head scan inside the body we matched the two X-ray volumes directly. We tried all 24 ways a cube can be turned, and slid each over the whole body with a fast Fourier correlation of edge-enhanced images. One orientation won clearly (correlation 0.34 against 0.25 for the next). Afterwards 87% of the head scan's pharynx lies inside the whole-body pharynx, and the ganglion under the pharynx lands 86% inside the first swelling of the whole-body nerve cord, which identifies it as trunk ganglion 1 (the nervous system). Matching organ shapes alone failed, because the pharynx is nearly a sphere and the fit has no preferred orientation.

Two matching CT slices of the head with organ outlines on both — Left: the whole-body scan. Right: the head scan, moved into the same frame. The outlines are the head scan's organs, landing on the same tissue in both.

From voxels to a viewer

Each label became a surface with marching cubes, smoothed and simplified, then packed into one 3D file. In the viewer, the cut plane samples the real CT, the 109 storage cells are individually clickable with a histogram of their volumes, and scale spheres beside the animal are each drawn at a published size.

The glass tardigrade with storage cells coloured by size and a histogram — Storage cells coloured by volume. The scan's 137 cells sit between a hatchling's 72 and an adult's 281 (Quiroga-Artigas & Moriel-Carretero 2024).

What's uncertain

Structures marked "probable" are our identifications, not the original authors'. The head scan's three unnamed paired structures are now read as connectives of the nervous system, two of them confidently and one only probably (the nervous system).
The head voxel size is inferred from volumes, not stated anywhere.
This is one animal, fixed and dried for scanning, which shrinks tissue.

Get it

The meshes (GLB, STL, OBJ), a single label volume with the head scan merged in (NRRD, TIFF), the body-masked CT and a table of every structure are on Resources.

How this is built

Data: Gross et al. 2019's whole-body nano-CT (644 slices, 270 nm voxels) and head scan (728 slices), each with its label fields, and the organ volumes in the paper's Table 1. All CC BY 4.0, from figshare.
Built with: Our Python script build_meshes.py (NumPy, SciPy, scikit-image, trimesh). It names the labels by volume, traces the body by threshold, closing, filling and a two-of-three vote, and registers the head scan by FFT cross-correlation over all 24 orientations. Marching cubes turns each label into a surface, which is smoothed, simplified and packed into one GLB file. three.js draws it in the viewer.
Checked: By measurement against the paper's volumes: pharynx, gut and Malpighian tubules within 1%, the traced body 0.3% under, and brain and salivary glands matching at our 0.1065 µm head voxel. After registration, 87% of the head scan's pharynx lies inside the whole-body pharynx. The nervous-system measurements are recorded as tests in the graph; the rest is not yet.
Limits: Labels marked "probable" or "unidentified" are our calls. The head voxel size is inferred from volumes. One animal, fixed and dried, which shrinks tissue.
More: How the whole project is built. The figures below come from the figure library, each with its own credit and licence.

Inside the animal, from the papers

Figure 1 from Giovannini et al. 2022: Paramacrobiotus spatialis. — Paramacrobiotus spatialis. Figure 1, Giovannini et al. 2022 · CC BY 4.0

Figure 4—figure supplement 3 from Hering et al. 2016: Immunofluorescence labelling of cytotardin in the tardigrade Hypsibius dujardini with focus on the epidermis. — Immunofluorescence labelling of cytotardin in the tardigrade Hypsibius dujardini with focus on the epidermis. Figure 4—figure supplement 3, Hering et al. 2016 · CC BY 4.0

Fig. 3 from Mapalo et al. 2024: Habitus of Aerobius dactylus gen. — Habitus of Aerobius dactylus gen. Fig. 3, Mapalo et al. 2024 · CC BY 4.0

Figure 2 from Mayer et al. 2013: Organisation of the nervous system in the tardigrade Macrobiotus cf. — Organisation of the nervous system in the tardigrade Macrobiotus cf. Figure 2, Mayer et al. 2013 · CC BY 4.0

Figure 6 from Mayer et al. 2013: Organisation of the peripheral nervous system in the tardigrade Macrobiotus cf. — Organisation of the peripheral nervous system in the tardigrade Macrobiotus cf. Figure 6, Mayer et al. 2013 · CC BY 2.0

Fig 1 from Czerneková et al. 2018: Storage cells (SC) of R. coronifer. (A) Tun, SEM. Bar = 30 μm. (B) Active animal, LM. Bar = 20 μm. (C) Storage cells, SEM. Bar = 4 μm. (D-G) Ultrastructure of SC of… — Storage cells (SC) of R. coronifer. (A) Tun, SEM. Bar = 30 μm. (B) Active animal, LM. Bar = 20 μm. (C) Storage cells, SEM. Bar = 4 μm. (D-G) Ultrastructure of… Fig 1, Czerneková et al. 2018 · CC BY 4.0