Introduction

Tadpoles of most anuran amphibians share a similarly structured olfactory periphery, consisting of the main olfactory epithelium in the principal nasal cavity (PC), a vomeronasal organ (VNO), as well as some minor additional epithelial surfaces (Jungblut et al. 2021; Weiss et al. 2021). This is well documented for all major groups of anurans: Archaeobatrachians (Benzekri and Reiss 2012), Mesobatrachians (Manzini and Schild 2010) and Neobatrachians (Jermakowicz et al. 2004; Jungblut et al. 2011, 2017; Nowack and Vences 2016; Quinzio and Reiss 2018).

The transition from an aquatic tadpole to a more or less water-independent adult frog requires the neuronal network associated with odor processing to restructure and rapidly adapt to the new habitat (Duellman and Trueb 1994; Wells 2007; Reiss and Eisthen 2008). During metamorphosis, the olfactory periphery transforms into a tripartite chamber system, consisting of the main olfactory epithelium in the PC, a middle cavity (MC) lined with non-sensory epithelium in most species, and the VNO (Helling 1938; Reiss and Eisthen 2008). While the larval system (PC and VNO) is specialized for the detection of waterborne odors, the sensory epithelium in the adult PC is associated with sensing volatile odors. In some anuran species, a small patch of sensory epithelium on the floor of the PC — the recessus olfactorius — is also tuned to detect waterborne odors (for reviews, see Reiss and Eisthen, 2008; Weiss et al., 2021).

In contrast to many other anurans, the pipid frog Xenopus laevis is secondarily aquatic. Adult Xenopus are fully water-dwelling (Wells 2007; Reiss and Eisthen 2008) and only occasionally move overland (Du Plessis 1966; Measey 2016). The special ecology of adult Xenopus is reflected in the presence of a specialized “water-nose” in the MC, which starts to form around premetamorphic stage 51 after Nieuwkoop and Faber (Föske 1934; Nieuwkoop and Faber 1994; Reiss and Burd 1997a, b; Hansen et al. 1998; Higgs and Burd 2001; Dittrich et al. 2016). The epithelium in the adult MC exhibits strong similarities with the larval PC. It contains both major types of olfactory receptor neurons (ORNs, ciliated and microvillous; Hansen et al. 1998), is responsive to common waterborne olfactory stimulants like amino acids (Sorensen and Caprio 1998; Syed et al. 2017), and expresses a similar set of olfactory receptors (Freitag et al. 1995, 1998; Amano and Gascuel 2012; Syed et al. 2013, 2017).

During metamorphosis, major remodeling occurs in the PC of larval Xenopus caused by massive cell death and replacement of ORNs (Hansen et al. 1998; Higgs and Burd 2001; Dittrich et al. 2016). The remodeled adult PC is eventually composed of only ciliated ORNs and expresses olfactory receptors putatively responsive to airborne odorants (Freitag et al. 1995; Mezler et al. 1999, 2001), thus assuming the role of the adult “air nose” (Föske 1934; Hansen et al. 1998; Syed et al. 2017). In contrast, the VNO does not seem to change significantly during metamorphosis in regard to its cellular composition or function (Hansen et al. 1998; Dittrich et al. 2016).

The functional and cellular segregation of the olfactory periphery also translates to the level of the olfactory bulb. Receptor neurons in the VNO project their axons towards their target structures, the glomeruli, in the accessory olfactory bulb (Reiss and Eisthen 2008; Jungblut et al. 2012). ORNs residing in the PC of larval Xenopus project their axons to glomeruli arranged in the ventral portion of the main olfactory bulb (vMOB) (Reiss and Burd 1997b; Weiss et al. 2020a). During metamorphosis, these axonal projections from the PC to the vMOB are replaced by ORN axons originating in the newly formed MC, the adult “water nose” (Key and Giorgi 1986; Hofmann and Meyer 1991; Franceschini et al. 1992; Reiss and Burd 1997b). Newly generated ORNs in the PC are projecting their axons towards the dorsal part of the main olfactory bulb (dMOB) during metamorphosis (Hofmann and Meyer 1991; Reiss and Burd 1997b; Gaudin and Gascuel 2005). Since it is solely innervated by ORNs residing in the adult “air-nose,” the dMOB is putatively associated with the processing of volatile odorants (Föske 1934; Reiss and Eisthen 2008; Weiss et al. 2021).

Functionally, the ORN axonal projections to the vMOB of Xenopus tadpoles have been shown to be segregated into two major odor processing streams: projections to laterally located glomeruli use a second messenger cascade dependent on phospholipase C and diacylglycerol (DAG), while projections to medially located glomeruli are dependent on the adenylate cyclase and use cyclic adenosine monophosphate (cAMP) as a second messenger (Gliem et al. 2013; Sansone et al. 2014). A substantial portion of the lateral glomeruli are responsive to amino acids and putatively innervated by microvillous ORNs (Gliem et al. 2013). Medially located glomeruli are putatively connected to ciliated ORNs and mainly respond to alcohols, aldehydes, and amines instead (Gliem et al. 2013). While there is evidence that the anatomy of the glomeruli in the vMOB remains stable during metamorphosis (Gaudin and Gascuel 2005; Weiss et al. 2020a), it is still unknown whether the innervation shift from the larval PC to the adult MC affects its function and/or its behavioral output.

Odor information transferred to the glomeruli in both parts of the MOB is subsequently relayed to the postsynaptic cells, the projection neurons (Imamura et al. 2020). In amphibians, these neurons feature one or more apical dendrites that terminate in densely branched tufts within glomeruli, and an axon projecting to higher brain centers (Jiang and Holley 1992b; Dryer and Graziadei 1994; Nezlin et al. 2003; Imamura et al. 2020). It has been shown in other vertebrates that the morphology of projection neurons varies based on their function or their association with distinct olfactory subsystems (Nagayama et al. 2014; Imamura et al. 2020; Braubach and Croll 2021). In amphibians, such analyses are lacking so far.

In the present work, we analyze the neuronal network restructuring of the MOB during metamorphosis on an anatomical, functional, and behavioral level. We show that ORN axonal projections from the larval PC to the vMOB are progressively replaced by ORN axons originating in the newly formed MC. Despite these massive reorganization processes, the functional segregation into two separate odorant processing streams in the vMOB remains stable. Similarly, the remodeling does not change behavioral responses to amino acids, implying that connections to higher brain centers remain intact during the process. Axons from newly develo** ORNs in the remodeled PC project to the dMOB, forming an unpaired structure around the interhemispheric midline. We show that the dMOB network follows a wiring logic distinct from the vMOB, with a progressively higher number of ORN axons projecting to glomeruli in the contralateral hemisphere. Additionally, the populations of postsynaptic projection neurons of the vMOB and the dMOB differ morphologically, revealing a higher prevalence of multi-tufted neurons connected to multiple glomeruli associated with the dMOB. These features point towards a higher degree of integration between the hemispheres and across several glomeruli in the dMOB, possibly adaptive to the processing of volatile odorants. The processing of waterborne odorants in the postmetamorphic vMOB, on the other hand, seems to mirror the larval system.

Material and methods

Animals and tissue preparation

All animals used in this study were wild type or albino Xenopus laevis (both sexes), kept and bred at the University of Giessen at a water temperature of 20 °C in water tanks with constant water circulation. Developmental stages were defined according to Nieuwkoop and Faber (Nieuwkoop and Faber 1994). Before experimental procedures, the animals were anesthetized using 0.02% MS-222 (ethyl 3-aminobenzoate methanesulfonate; TCI Germany) in tap water. For tissue preparation, anesthetized animals were killed by severing the transition between the brainstem and spinal cord. For lower staged animals, a tissue block containing the noses and the rostral part of the telencephalon was removed. For higher staged animals, the olfactory nerve was cut close to the noses and the entire brain was taken out of the cartilage. All animal procedures were performed in accordance with the guidelines for Laboratory animal research of the Institutional Care and Use Committee of the University of Giessen (V 54 – 19 c 20 15 h 01 GI 15/7 Nr. G 2/2019; 649_M; V 54 – 19 c 20 15 h 02 GI 15/7 kTV 7/2018).

Tracings of olfactory receptor neurons via fluorophore-coupled WGA and electroporation

For bulk loadings of olfactory projections from the nose to the MOB, we filled the nasal cavities with droplets of approx. 3 µl of fluorophore-coupled wheat germ agglutinin (WGA Alexa Fluor 488 or 594 conjugate, Thermo Fisher) diluted at a concentration of 10 mg/ml in saline Frog Ringer’s (in mM: 98 NaCl, 2 KCl, 1 CaCl2, 2 MgCl2, 5 glucose, 5 Na-pyruvate, 10 Hepes, pH 7.8). For bulk electroporations, we placed dried dye crystals of fluorophore-coupled dextrans (Alexa Fluor dextran 488, 594, Cascade Blue dextran, Thermo Fisher or Cal520 dextran conjugate, AAT Bioquest, 10 kDa, 3 mM in Frog Ringer’s) in the nostrils and applied six electric square pulses using two platinum electrodes (15 V, 25 ms duration at 2 Hz with alternating polarity) to each nostril (for detailed protocol see Weiss et al. 2018). After the procedures, animals were left to recover for at least 24 h before tissue preparation and imaging.

Sparse cell electroporation in the olfactory epithelium and the olfactory bulb

We sparsely electroporated ORNs in the nasal epithelia of the PC/MC and projection neurons in the vMOB/dMOB using micropipettes pulled from borosilicate glass capillaries (Warner instruments, resistance 10–15 MΩ) filled with fluorophore-coupled dextrans (Alexa Fluor dextran 488 and 594; 3 mM in Frog Ringer’s). The dye-filled micropipettes were mounted on the headstage of a single cell electroporator (Axoporator 800A; Axon Instruments, Molecular Devices) equipped with a wire electrode and approached cells in the olfactory epithelium (PC/MC) or the MOB using a micromanipulator. A 500-ms train of square voltage pulses (50 V, single pulse duration 300 μs at 200–300 Hz) was applied.

For ORN labeling, the animals were first anesthetized and sparse cell electroporation was repeated at multiple locations using Alexa Fluor 488 dextran in the MC and Alexa Fluor 594 dextran in the PC to trace their respective projections during metamorphosis. After the procedure, animals were left to recover for at least 24 h prior to image acquisition (Hassenklöver and Manzini 2014). Sparse labeling of projection neurons in the MOB was conducted in an excised tissue block containing the olfactory system. We fixed the bulb and the caudal portion of the olfactory nerve under a platinum grid strung with nylon threads, approached the micropipette to the cell layer containing MOB projection neurons, and applied the voltage pulse trains as described above (detailed protocol in Weiss et al. 2018).

Image acquisition and processing of morphological images

Images were acquired as multi-color virtual image stacks with a z-resolution of 1–3 µm using multiphoton microscopy (Nikon A1R-MP) at an excitation wavelength of 780 nm. We used ImageJ (Schindelin et al. 2012; RRID:SCR_003070) to adjust the brightness and contrast of the image stacks and applied a median filter to remove pigmentation-derived artifacts in some images. Separate images were stitched together where necessary (Preibisch et al. 2009). For thresholding analyses conducted on the image-stacks, we eliminated tissue-derived autofluorescence by subtracting the blue-wavelength color channel (when no blue-emitting dye was introduced into the tissue). Images are presented as maximum intensity projections along the z-axis or in 3D using the 3D-viewer implemented in ImageJ.

Functional calcium imaging and data processing

For functional calcium imaging in the vMOB, we loaded the ORNs with a calcium sensitive dextran coupled dye (Cal 520 dextran conjugate, 10 kDa, AAT Bioquest; 3 mM in Frog Ringer’s) via bulk electroporation (Weiss et al. 2018) as described above. After killing the animal, we cut out the tissue block containing the noses and the anterior part of the brain and removed the ventral palatial tissue as well as tissue around the sensory epithelia to facilitate odorant flow into the nasal cavities. The tissue block was positioned on the stage of the multiphoton microscope using a platinum grid strung with nylon threads. A perfusion manifold with silicone tubing outlet (Milli Manifold; ALA Scientific) connected to a gravity-fed multi-channel perfusion system (ALA-VM-8 Series; ALA Scientific) was positioned in front of the nasal cavity and a constant Ringer’s flow was established. Ringer’s was constantly removed from the recording chamber via a syringe needle connected to a peristaltic pump via silicone tubing. Fast volumetric recordings of the ORN axon terminals in the glomeruli of the vMOB were made using the resonant scanning mode of a multiphoton microscope (780 nm excitation wavelength). We measured time series of 3D virtual image stacks (lateral dimensions: 509 × 509 µm, 512 × 512 pixel; axial dimensions: 180–300 µm, inter-plane distance 4–7 µm) at 0.5–1 Hz per image stack (Offner et al. 2020). Stimuli were applied with a duration of 5 s and an inter-stimulus interval of 60 s and repeated at least twice. The stimuli used were as follows: amino acid mixture: L-valine, L-leucine, L-isoleucine, L-methionine, glycine, L-serine, L-threonine, L-cysteine, L-arginine, L-lysine, L-histidine, L-tryptophan, L-phenylalanine, L-alanine, L-proline (100 µM in Frog Ringer’s solution); amine mixture: 2-phenylethylamine, tyramine, butylamine, cyclohexylamine, hexylamine, 3-methylbutylamine, N,N-dimethylethylamine, 2-methylbutylamine, 1-formylpiperidine, 2-methylpiperidine, N-ethylcyclohexylamine, 1-ethylpiperidine, piperidine (100 µM in Frog Ringer’s solution); bile acid mixture: taurocholic acid, cholic acid, glycholic acid, deoxycholic acid (100 µM in Frog Ringer’s solution); odorant mixture: mixture of amino acids (excluding proline and alanine), amines, bile acids, alcohols and aldehydes (positive control, 50 µM in Frog Ringer solution) (Gliem et al. 2013); Frog Ringer’s solution (negative control); forskolin, a direct stimulant of the adenylate-cyclase and the cAMP second messenger pathway. Forskolin was dissolved in DMSO (stock of 10 mM) and used at a final concentration of 100 μM. All chemicals were purchased from Sigma-Aldrich. The mixtures were stored as frozen aliquots and diluted to their final concentration shortly before the experiments.

Data evaluation was done using Python. A piecewise-rigid motion correction algorithm was applied to remove motion artifacts (Pnevmatikakis and Giovannucci 2017). We used the CaImAn toolkit for calcium imaging data to generate denoised and deconvolved templates of the 3D image stacks (Pnevmatikakis et al. Full size image

In the vMOB, the axons coming in from the left and right olfactory nerve project to two vMOB projection fields that are spatially separated at the interhemispheric midline (highlighted by the white dotted lines in the left images, Fig. 5a). The general outline of the glomerular projections in the vMOB remains constant throughout metamorphosis and only slightly increases in overall volume. Contrastingly, the first axonal projections reach the dMOB around stage 50 (Fig. 5a, top right) and project towards the interhemispheric midline. Until the onset of prometamorphosis, the incoming fibers from both olfactory nerves have formed a single projection area around the midline, which grows until the end of metamorphic development. (Fig. 5a, right). During prometamorphosis and metamorphosis proper, the hemispheres of the dMOB projections are not clearly separated anymore since axons cross the midline from both sides, frequently innervating glomeruli on the contralateral side. This feature of the dMOB becomes even more apparent when looking at sparse cell labeling (inset, Fig. 5a). Among the ORN axons crossing to the contralateral side, we found some axons only connecting to the ipsilateral or contralateral side, but some axons bifurcating and innervating glomeruli on both sides of the midline.

To quantify the positions of incoming ORN axons from the left (magenta, Fig. 5) and right epithelia (cyan), we counted pixels containing fluorescence signal from the ORN axons of the two epithelia along the left–right axis of the vMOB (Fig. 5b) and dMOB (Fig. 5b’). In the vMOB, the pixel count curves have symmetrical peaks on both sides of the interhemispheric midline (at approx. 80% of the distance to the edges; asterisks, Fig. 5b), while in immediate proximity to the midline, no pixels containing fluorescence signal could be counted. This pattern did not change from earlier stages (50–56; n = 4 animals) to later stages (57–66; n = 7) during metamorphosis. In the dMOB (Fig. 5b’), on the other hand, the symmetrical peaks were located closer to the midline, at approx. 15% (asterisks) of the distance to the edges in animal of stages 50–56 (n = 7) with some overlap around the midline (area under both curves). In the later metamorphic stages (57–66; n = 9), the two peaks are almost on the midline (5% of the distance to the edges; asterisks), with even more substantial overlap.

We further quantified the percent share of innervation overlap relative to the total projections in animals of different stages for the vMOB (Fig. 5c, white dots) and dMOB (Fig. 5c, black dots). The average overlap of the right and left ORN axons in the vMOB amounts to 2.2 ± 3.7% (animals of all examined stages; n = 11, Fig. 5c). The percentage of overlap in the dMOB increases linearly throughout metamorphosis (premetamorphosis: 10 ± 7%, n = 5; prometamorphosis: 32 ± 14%, n = 6; metamorphosis proper: 40 ± 12%, n = 5, Fig. 5c). In addition, our results show that a single glomerular structure in the dMOB can be composed of axon terminals from both sides (magenta and cyan arrowheads, Fig. 5d).

The populations of projection neurons in the vMOB and the dMOB are distinct

The connection between peripheral receptor neurons and the glomeruli differs between the vMOB and the dMOB. To understand how these projections from the periphery connect to the postsynaptic projection neurons, we labelled and reconstructed single projection neurons in the vMOB and dMOB of postmetamorphic Xenopus laevis (stage 66) (Fig. 6).

Fig. 6
figure 6

Different morphology of projection neurons in the vMOB and dMOB of postmetamorphic Xenopus. ORN axonal projections from the olfactory epithelia (both PC and MC) were traced via electroporation of fluorophore-coupled dextrans (left: Cascade Blue dextran, magenta; right: Alexa Fluor 594 dextran, cyan), and projection neurons (yellow) with Alexa Fluor 488 dextran via sparse cell electroporation in the vMOB (a) and dMOB (b). White dotted lines indicate the ORN projections, tufts (white dotted squares in a, b) are shown in a higher magnification in (a’) (vMOB), (b’–b’’’) (dMOB). Reconstructions of representative neurons in the vMOB (c) and dMOB (d) are shown. Asterisks indicate the projection neuron somata, filled white arrowheads primary tufted dendritic branches and empty arrowheads primary basal neurites without tufted terminals. (e) Each radial axis on the radar chart represents a morphological descriptor of the projection neurons. The means of the descriptors of projection neurons in the vMOB (yellow; n = 18) and the dMOB (orange; n = 16) are shown. Dorsal projection neurons have a higher number of tufts, a longer average distance between soma and tufts and more primary tufted dendrites. Neurons in the vMOB have a significantly higher tuft volume. The number of primary basal neurites is similar between the groups. (f) The stacked barplots depict the percent share of uni-tufted (grey) and multi-tufted projection neurons (yellow, vMOB; orange, dMOB). A higher number of uni-tufted projection neurons was found in the vMOB (> 50%), while > 90% of neurons in the dMOB terminated in at least two tufts, maximally in five tufts. Boxplots: Average tuft volumes (above) and average distance between tufts and somata (below) are compared between uni-tufted neurons (grey, left, n = 11), bi-tufted neurons (yellow/orange, middle, n = 12) and neurons with three or more tufts (dark yellow/dark orange, right, n = 11). White triangles depict the means. Tuft volume decreases with a higher number of tufts, while tuft-soma distance increases. The groups contain neurons from both the vMOB and the dMOB. *p < 0.05, **p < 0.01, ***p < 0.005. A anterior, dMOB dorsal main olfactory bulb, L lateral, M medial, ON olfactory nerve, P posterior, POST postmetamorphosis, vMOB ventral main olfactory bulb, PRE premetamorphosis, PRO prometamorphosis, MET metamorphosis proper, vMOB ventral main olfactory bulb

Full size image

Projection neurons in the vMOB (n = 18 neurons, Fig. 6a, c) and dMOB (n = 16, Fig. 6b, d) share some general morphological features. Both groups have one or multiple primary dendrites (filled arrowheads in Fig. 6c, d) originating from the soma (white asterisks in Fig. 6) and terminating in highly branched dendritic tufts connecting to one or multiple glomeruli (Fig. 6a’, b’–b’’’). A representative projection neuron with a single tuft in the right vMOB is shown in Fig. 6a, and a multi-tufted neuron in the dMOB in Fig. 6b. While the tufts of neurons in the vMOB get synaptic input only from ipsilateral ORN projections (cyan projections in Fig. 6a), multiple tufts of a single projection neuron in the dMOB often project to both hemispheres, possibly integrating synaptic ipsilateral and contralateral ORN input (cyan and magenta in Fig. 6b). In addition to the dendrites receiving input from the dendritic tufts, most cells also have secondary dendritic branches with blunt endings and basal neurites (empty arrowheads, Fig. 6). Figure 6 c and d show reconstructions of representative projection neurons in the vMOB and dMOB that connect to a different number of glomeruli (white dotted circles).

From the reconstructions, we extracted several descriptors for the projection neurons (Supplementary video 1) and compared the measurements of neurons labeled in the vMOB and the dMOB (Fig. 6e). The tuft-soma distance along the dendrites (for cells with multiple tufts, the distance was averaged) was significantly longer in the dMOB cells (323 ± 96 µm) compared to the vMOB cells (230 ± 70 µm; p = 0.0046, Fig. 6e). Projection neurons in the vMOB and dMOB connected to an average of 1.7 ± 0.8 and 2.6 ± 1.1 dendritic tufts (p = 0.043), respectively, with an average tuft volume of 7679 ± 8208 µm3 and 2435 ± 5410 µm3 (p = 0.0003). We classified neurite branches along two axes into primary (originating in the soma) or secondary (branching off from primary or other secondary branches; Supplementary video 1) as well as dendritic/tufted (connected to the tufts) or basal neurites (not connected to the tufts). Neurons in the vMOB had 1.2 ± 0.5 primary-tufted dendrites and 1.7 ± 1 primary-basal neurites. The dMOB neurons exhibited significantly more of the former (1.9 ± 0.7; p = 0.0003) and a similar number of the latter (1.3 ± 1.4, Fig. 6e). The count of secondary-tufted dendrites was similar among the two groups (vMOB 3.9 ± 2.1, dMOB 3.8 ± 3.4) while the dMOB cells had more secondary-basal branches (7.1 ± 2.7) compared to the vMOB (4.4 ± 2.7; p = 0.004).

The most apparent difference between the two neuronal populations is the share of uni- and multi-tufted cell morphologies (Fig. 6f, barplots). Of neurons labeled in the vMOB, 55.6% are uni-tufted (gray, Fig. 6f), while only a single uni-tufted cell has been found in the dMOB (6.25%). The bi- and tri-tufted neurons both make up 22.2% of the vMOB population (yellow in Fig. 6f). In the dMOB, the projection neurons maximally had five tufts. The biggest share of the population was bi-tufted and tri-tufted cells with 50% and 31.25%, respectively (orange in Fig. 6f).

In the next step of the analysis, we pooled the data based on the number of tufts (one, n = 11; two, n = 12; three or more, n = 11; boxplots Fig. 6f), combining both neurons from the vMOB and the dMOB. Uni-tufted neurons had the greatest tuft volume (12,711 ± 8921 µm3, Fig. 6f, middle), significantly larger than the average tuft volume of the multi-tufted cells (bi-tufted: 2246 ± 2911 µm3, p = 0.001; three or more tufts: 946 ± 600 µm3, p = 0.0002). Inversely, the mean dendritic distance between tuft and soma increased with a higher number of tufts, measuring 211 ± 60 µm in uni-tufted, 281 ± 93 µm in bi-tufted, and 330 ± 93 µm in cells with three or more tufts (Fig. 6f, below).

As an additional descriptor for all multi-tufted cells, we measured the average spatial distance between two tufts of a cell as an estimate of spatial span of the neuron. Tufts of bi-tufted neurons were on average 146 ± 93 µm apart, with a maximum distance of 325 µm, while we measured 128 ± 74 µm for neurons with three or more tufts. Comparing all multi-tufted projection neurons in the vMOB (n = 8) and dMOB (n = 15) showed that the inter-tuft distance is significantly bigger in the dMOB (178 ± 73 µm; p = 0.0003) compared to the vMOB (62 ± 32 µm). Inversely, the volume of multi-tufted neurons in the vMOB was bigger than in the dMOB (2500 ± 2671 µm3; 1157 ± 1836 µm3; p = 0.008).

In summary, the projection neurons in the dMOB develop multiple small dendritic tufts, which receive input from spatially distant glomeruli, while at least half of the vMOB neurons have a single larger tuft receiving input from only one glomerulus.