Editor’s Summary

Although acute damage and age-related decline can lead to a loss of thymic and immune function, the adult mammalian thymus retains some limited regenerative capacity. Czarkwiani et al. found that juvenile axolotls can fully regenerate their thymuses after complete removal. Thymus regeneration was associated with restoration of morphological and transcriptional features. Whereas the key mammalian thymic transcription factor FOXN1 was dispensable for thymus regeneration, single-cell transcriptomics identified the growth factor midkine as a likely driver. Future studies in axolotls could inform new therapeutic approaches for promoting thymus regeneration. —Claire Olingy

Abstract

The thymus is the primary site of T cell development, central to the establishment of self-tolerance and adaptive immune function. In mammals, the thymus undergoes age-related involution, resulting in a global decline in immune function. The thymus has some regenerative ability that relies on pre-existing thymic remnants but is insufficient to prevent involution. Here, we show that the juvenile axolotl (Ambystoma mexicanum) is able to regenerate its thymus de novo after complete removal, constituting an exception among vertebrates. Using single-cell transcriptomics and genetic and transplantation approaches, we demonstrate that de novo thymus regeneration results in the restoration of morphology, cell-type diversity, and function. FOXN1, although it has a conserved role in thymus organogenesis, is dispensable for the initiation of thymic regeneration. In contrast, we identify midkine signaling as a possible early driver of de novo thymus regeneration. This study demonstrates an instance of organ-level regeneration of the lymphoid system, which could guide future clinical strategies seeking to promote thymus regrowth.

INTRODUCTION

In humans, the loss of thymic function through thymectomy, environmental challenges, or age-dependent involution is associated with increased mortality, inflammaging, and higher risk of cancer and autoimmune disease (1). This is largely due to a decline in the intrathymic naïve T cell pool, whose generation is orchestrated by the thymic stroma, particularly thymic epithelial cells (TECs) (2). Upon challenges that affect the TEC compartment, the thymus is capable of triggering an endogenous regenerative response by engaging resident epithelial progenitors with stem cell features (3–5). Yet, after age-related atrophy or thymectomy resulting from myasthenia gravis or tumor removal (1), this regenerative response is unable to overcome the loss of thymic tissue, highlighting the need for therapeutic interventions.

The restoration of thymic functionality has been achieved to a limited extent via strategies targeting the thymic epithelial microenvironment or hematopoietic progenitors, modulating hormones and metabolism, or through cellular therapies and bioengineering (6). In mice, the up-regulation of Foxn1, a key transcription factor for thymus development and organogenesis (7), either directly or via its upstream effector bone morphogenetic protein 4 (BMP4), can support activity of cortical TECs (cTECs) (8, 9). Further, a combination of growth hormone and metformin has been shown to restore thymic functional mass in humans (10). Nevertheless, such strategies only lead to delayed thymic involution, and examples of complete thymus regeneration have not yet been described among vertebrates.

Because of its remarkable regenerative abilities that extend to parts of the brain, eye, heart, and spinal cord, and even entire limbs, the axolotl (Ambystoma mexicanum) is a powerful model for regeneration studies (11). The axolotl has offered insights into the mechanisms of positional identity (12), cell plasticity (13, 14), and the molecular basis of complex regeneration (15–18). The regeneration of axolotl body parts relies on remnants of the missing structure, with the exception of lens tissue, which can regrow from dorsal pigmented epithelial cells during a short window during development (19). However, whether de novo regeneration can occur for an entire complex organ, in axolotls or any other vertebrate, is unknown.

Given the critical role of the thymus in the development and function of adaptive immunity and the importance of the immune system in regenerative processes, we aimed to address the regenerative potential of the thymus in juvenile axolotls. Here, we show that the axolotl can regenerate its maturing thymus de novo, a feat unparalleled among vertebrates. By generating single-cell transcriptional landscapes of the axolotl thymus and its regenerative stages, we dissect the heterogeneity of its cellular components, show their recapitulation and functionality after regeneration, and identify possible molecular drivers of this process. Through functional studies, we identify conserved features and molecular innovations associated with de novo thymus regeneration.

RESULTS

The juvenile axolotl reconstitutes morphological and molecular features of the thymus upon removal

The axolotl thymus is composed of three nodules located bilaterally on each side of the head at the base of the gills (fig. S1, A and B) (20). The nodules are surrounded by extracellular matrix and not connected to other major structures (movie S1), enabling the efficient removal of intact nodules (movie S2 and fig. S1, B and C). Complete thymus removal was also supported by in situ hybridization (ISH) for TECs (Tbata) and T cells (Rag1 and Cd8b) at 1 and 5 days postthymectomy (dpt) showing no signal compared to intact thymic samples (fig. S1D). We took advantage of tgSceI(Xla.KRT12:eGFP)ETNKA [referred to as Krt12:eGFP (enhanced green fluorescent protein)] transgenic juvenile axolotls (21), in which a subset of cells in the thymus is labeled, for thymectomy visualization (fig. S1C). Thymus morphogenesis is structurally complete in axolotls at 3 weeks postfertilization (fig. S1E), and thymectomies were conducted at 6 to 8 weeks posthatching. Although the mouse thymus continues to mature until 4 weeks postnatally, the axolotl thymuses used in this study are molecularly equivalent to a structurally complete mouse thymus at stages embryonic day 13.5 (E13.5) and older (fig. S1, E and F).

Unexpectedly, thymic-like structures reappeared after complete nodule removal (Fig. 1, A and B). Upon single- or double-sided thymectomy, regenerated thymic nodules were first detected at 7 days dpt and grew substantially between 14 and 35 dpt, at which point the nodules appeared morphologically similar to contralateral controls. Overall, regenerated thymic nodules were observed in ~60% of the thymectomised animals by 35 dpt (n = 167; fig. S2, A and B). The variability of regeneration further lends support to the lack of remaining rudiment, in contrast with the usual 100% regeneration observed in instances of regrowth from a retained scaffold (for example, in axolotl limb regeneration). Together, these observations suggest that the juvenile axolotl thymus is capable of de novo regeneration.

Fig. 1. The thymus regenerates de novo after complete thymectomy in the juvenile axolotl.

(A) Optical clearing, nuclear staining (SYTOX green), and light-sheet microscopy to visualize the regeneration progression of thymic nodules (white arrowheads) at 3 to 35 dpt in juvenile axolotls (n = 7 to 9 samples per time point). Scale bars, 500 μm. (B) H&E histological staining of sections corresponding to animals in (A) showing the deposition of extracellular matrix and the morphological restoration of the regenerating thymic nodules (black arrowheads) from 7 to 35 dpt. Scale bars, 500 μm. bv, blood vessel; ecm, extracellular matrix; g, ganglion; m, muscle; n, nerves; rbc, red blood cells. (C) Schematic of the thymectomy, nodule collection, and processing for scRNA-seq. (D) scRNA-seq UMAP projection of control and 60 dpt regenerated thymic nodules (nodules pooled from six animals; n = 2 biological replicates). (E) Identification of clusters and (F) annotation of cell types found in the axolotl thymus. (G) Expression of cell type markers in the axolotl thymus labeling different populations. (H) Differential density of regenerated and control cells on the UMAP embedding and cell type–specific differential abundance (dotted line indicates not significant). Color bars indicate the log number of cells per cluster in the control and regenerated samples. (I) HCR ISH on control and regenerated thymic sections showing cell types including Foxn1+ thymic epithelial cells, Cd3e+ lymphocytes, and Krt12+ epithelial cells. Dashed line square indicates position of higher magnification images. Scale bars, 50 and 10 μm in inset panels.

To determine whether this process is accompanied by restoration of cellularity and molecular signatures, we performed single-cell transcriptomic profiling on contralateral control and regenerated thymic nodule samples at 60 dpt (Fig. 1C). To align our control and regenerated thymic samples, we generated a custom axolotl genome reference and improved gene annotations with orthologs from six species—human, mouse, frog, zebrafish, chicken, and lizard. The Uniform Manifold Approximation and Projection (UMAP) embedding obtained after processing showed strong intermixing between the two replicates (fig. S2C) with cells from control and regenerated juvenile thymuses evenly spread, suggesting regeneration of all thymic populations (Fig. 1, D to F). Unsupervised Leiden clustering identified 14 distinct clusters (Fig. 1E) comprising blood (Ptprc+) populations including T lymphocytes (Cd3e, Bcl11b, Cd4, Lck, Cd8b, Myb, Tcf7, and Tox) (22, 23), B lymphocytes (Cd79b and Iglc) (23), macrophages (Apoe, Marco, and Siglec1) (23–25), other myeloid cells (Chit1, Azu1, cathelicidin-al, Epx.l, and Alox15b) (26–31), and an unknown population expressing Psalp1, lectin, Prf1, Sftpb.l, and Cpa1 (Fig. 1, F and G; fig. S2D; and data file S1). The thymic stroma comprised endothelium (Cdh5, Cd34, Egfl7, Vwf, Plk2, Mmrn1, and Plvap) (23, 30, 32); two mesenchymal populations (Col3a1), one expressing Acta2, Alpha1, Col3a1, Col5a2, Col1a2, Cygb, Dcn, Fbn1, Fn1, Col1a1, Dpt, Mfap4, Pdgfra, Pdgfrb, and Ptx3 (22, 23, 33, 34) and a presumably neural crest–derived mesenchymal population expressing Schwann cell markers (Mbp, Mpz, Foxd3, Plp1, Sox2, and Sox10) (35–37); and two thymic epithelial populations (Epcam+), one expressing TEC markers (Foxn1, Foxg1, Cd274, Cd83, Cldn4, Ifngr1, Itgb4, Tbata, Prss16, cathepsin, and Krt5) (22, 38, 39) and the other expressing keratins (Krt12, Krt17, and Krt18) as well as other TEC markers (Cpe and Relb) (23, 40). Both TEC populations exhibited antigen presentation–related transcripts, including Cd74, Hla-dra, MHC class 2 beta, and MHC class 1 A alpha (Fig. 1G and fig. S3A) (22, 23). Although we observed TEC heterogeneity, we were unable to identify traditionally defined cortical and medullar TECs (41), matching the lack of a histologically distinct medullar structure (Fig. 1B and figs. S1E and S3A) in the axolotl thymuses of juveniles, consistent with observations in adult animals (42).

A transcriptomic comparison across the highly variable and the top cluster-specific differentially expressed genes revealed a high correlation between control and regenerated cells for each of the clusters (fig. S3, B and C). There were no significant differences in the number of cells for each of the identified cell types between the two conditions (Fig. 1H and data file S2). To rule out that this result is due to technical corrections, we created a dataset for the control thymus, annotated it, and projected the regenerated cells onto the control embedding. This resulted in excellent intermixing between control and regenerated cells across all populations (figs. S3 and S4 and data files S3 and S4), confirming that all cell populations in the axolotl thymus are reconstituted to a similar extent after de novo regeneration. Supporting this notion, the restoration of Foxn1+ TECs, Cd3e+ T cells, and Krt12:eGFP+ epithelial cells in the regenerated thymus was confirmed through hybridization chain reaction in situ hybridization (HCR-ISH; Fig. 1I). Collectively, these data provide a description of the cellular constituents of the juvenile axolotl thymus and show that its de novo regeneration results in morphological, cellular, and molecular restoration.

Recovery of thymus functionality upon de novo regeneration

We next asked whether the regenerated axolotl thymus is functional. First, we assessed its capacity to sustain thymopoiesis, which entails hematopoietic progenitor cell (HPC) (43) recruitment and maturation into functional T cells. Focusing on the single-cell RNA sequencing (scRNA-seq) T lymphocyte subset (Fig. 2A, colored cluster), we observed cells derived from the regenerated thymus across all T cell populations, consistent with reconstitution of thymopoiesis (Fig. 2B and fig. S5). Unsupervised clustering followed by visualization of the hierarchically ordered clusters on a force-directed layout revealed several similarities with mammalian thymopoiesis (Fig. 2, C and D; fig. S5; and data file S5). Key markers were similarly expressed in control and regenerated samples, including pre- and early thymic progenitor (ETP) markers (Flt3, Cd44, Gata1, Hoxa9, Itga2b.1, Cd34, Mcpt2, Gata2, and Mcpt1) (44–47) and double-negative (DN; Cd4−, Cd8−, Lyl1, Hhex, and Bcl11a) (45) and double-positive (DP; Cd4+ and Cd8+) T cell markers. Key developmental regulators such as Hes1, Rag1, and Tcf12 (45) were also similarly expressed in control and regenerated samples. We identified T cell receptor–α (TCR-α), TCR-β, and TCR-δ transcripts, indicative of αβ and δ T cells. Cluster 8 contained markers previously attributed to mammalian nonconventional lymphocytes (Tigit, Klrf1, and *Tcr-*δ) (Fig. 2, D and E, and fig. S5) (23, 44). We further validated our scRNA-seq analysis by conducting in situ hybridization against the ETP marker Flt3 and the broadly expressed (from ETPs to single positives; SPs) thymopoietic marker Runx1 (45). As above, we found similar scRNA-seq and spatial expression patterns in nonthymectomised and regenerated nodules (Fig. 2, F and G).

Fig. 2. Thymus functionality is restored after de novo regeneration.

(A) UMAP highlighting the T lymphocyte subset. (B and C) Force-directed layout colored by (B) condition (control/regenerated) or (C) Leiden cluster. (D) Row-standardized heatmaps displaying curated thymopoeisis markers in clusters split by condition: control (blue) and regenerated (orange). (E) UMAPs of imputed expression of the indicated lymphocyte markers. (F) UMAPs of imputed expression of Flt3 and Runx1 split by condition. Cells not found in the corresponding condition are colored in gray. (G) ISH of Flt3 and Runx1 in a control and regenerated thymus. (H) Schematic of experimental transplantation strategy. Scale bars, 100 μm. (I) Migration of fluorescent lymphocytes derived from the transplanted regenerated thymus into the amputated stump, blastema, and the digit stages (n = 5 limbs) during regeneration. Dashed line indicates amputation plane. Asterisks denote nonspecific autofluorescence. Scale bars, 1000 μm in 0 days postamputation (dpa) and 200 μm in other panels. (J and K) Immunohistochemistry showing representative images of colabeling of mCherry+ cells in the limb with T cell markers (J) CD8 and (K) CD3. Scale bars, 20 μm. (L) Thymus 1 year after transplantation; mCherry+ signal does not overlap with CD3+ signal. Scale bars, 20 μm. (M) Quantification of cells in (L) showing that no double-positive (mCherry+/CD3+) cells have been identified (n = 2 thymuses). (N) HCR-ISH showing Rag1+ and Flt3+ cells in regenerated transplanted nodules. Note that Flt3+ cells are mCherry−. Scale bars, 20 and 10 μm in insets.

To directly address the functionality of regenerated thymuses, we transplanted complete regenerated thymic nodules from reporter animals ubiquitously expressing fluorescent mCherry or GFP [tgSceI(CAGGs:Cherry)ETNKA or tgSceI (CAGGs:eGFP)ETNKA] into white axolotls (Fig. 2H and figs. S6 and S7). Starting at 3 days posttransplantation, fluorescent putative lymphocytes from the regenerated thymus could be observed populating host tissues including the blood stream (movie S3), spleen, and limbs (fig. S6). Furthermore, transplanted cells were recruited to regenerating limbs upon amputation (Fig. 2I), consistent with reports of T cell presence in limb blastemas (36). Recruited cells expressed lymphocyte markers CD8 (Fig. 2J) and CD3 (Fig. 2K). Inspection of the regenerated thymus 1 year posttransplantation revealed that the only cells remaining from the donor tissue were TECs (mCherry+, Foxn1+, and Tbata+), whereas its mature lymphocyte population was host derived (mCherry−, Rag1+, and CD3+) (Fig. 2, L to N, and fig. S7), indicating that the regenerated thymus can sustain host thymopoiesis posttransplantation. In agreement, host-derived ETPs (mCherry− and Flt3+) homed to the regenerated thymus (Fig. 2N). Thus, the juvenile axolotl thymus retains its key roles after de novo regeneration: HPC recruitment, thymopoiesis, and specification of migratory T cells.

Molecular profiling reveals stepwise reconstitution of the thymic stroma

To analyze the cellular and molecular dynamics of de novo thymus regeneration, we performed scRNA-seq on an unthymectomized (control) region including nodules (equivalent to 5 dpt), a sham-thymectomized region at the same stage, and the same region at various time points (5, 10, 21, and 35 dpt) postthymectomy (Fig. 3A, figs. S8 and S9, and data file S5). All samples thus included a variety of cell types observed in the environment of the thymus (fig. S1A), including the skin, muscle, cartilage, connective tissue types, nerves, blood vessels, and the thymic nodules in all samples except for 5 dpt, at which point no nodule could be observed. The cluster analysis of only the stromal subset defined by the presence of epithelial cells (Epcam+) revealed TECs and ciliated (Tubb4b, Rsph3, Iqcd, and Iqub) (48, 49), skin (Krt4 and Krt5), secretory (Muc5ac and Muc5b), and myoepithelial (Cd109) cells (34) and other epithelial populations containing specific markers Il17b, Col17a1, Pou2f3, and Krt124; mesenchymal cells (Col3a1+ and Vim+) including cartilage (Epyc and Alpha1), fibroblasts (Pdgfra, Lum, Dpt, Ptx3, Fbn1, and Fn1), mural cells (Acta2, Rgs5, and Pdgfrb), two muscle populations (Mymk, Myog and Msc, and Myf5), neural crest–derived mesenchyme (Mbp, Mpz, Foxd3, and Sox10), and a Col3a1+Vim+alpha-globin+ population. Additional stromal cell types were detected including endothelium (Cdh5, Cd34, and Lyve1) (34) and an Epcam− population containing neural markers Nefm and Stmn2 (Fig. 3, B to D; figs. S8 and S9; and data files S6 and S7).

Fig. 3. Molecular profiling reveals stepwise regeneration of the thymic stroma.

(A) Schematic illustrating the axolotl regeneration time course samples. (B and C) UMAP embeddings showing cells colored by clusters (B) and annotated cell types (C) (samples pooled from six animals each; n = 2 biological replicas except for 21-dpt time point). (D) Dotplot indicating curated markers for the cell types. (E) UMAP embeddings as in (B) showing different time points (outline, epithelial cell types; arrowhead, TEC cluster). (F) Bar graph of cell-type proportions (%) in the TEC cluster at different time points. (G) Violin plots indicating the data-driven and literature-curated TEC scores across the time points profiled. (H) Histology (H&E staining) and ISH for Foxn1 in the uninjured thymus, 1, 3, and 5 dpt (blue arrows). Scale bars, 200 μm. (I) Line plots showing the expression trends of selected literature TEC genes. (J) Heatmap of row normalized −log10-corrected P values (Benjamini-Hochberg) from the binomial test on a curated list of enriched terms for each of the groups from fig. S8K.

To dissociate regenerative and wound-healing events, we performed a per cell-type differential gene expression analysis between unthymectomized versus sham samples and between sham versus 5-dpt samples (log2 fold change ≥ 1, adjusted P value < 0.05) (fig. S10 and data file S8). Reflecting a wound-healing signature, muscle (131 genes) followed by fibroblast (51 genes) exhibited the highest number of genes up-regulated in sham, whereas the skin population had the highest number (27 genes) up-regulated in the control. A similar analysis in 5 dpt versus sham thymectomy indicated an early regenerative response, with the maximum number of differentially expressed genes in the Col17a1+ epithelial population (92 up-regulated in sham and 6 in 5 dpt) (fig. S10 and data file S9). Most of the cell types in the thymus environment were present in all samples, showing that the tissue isolation procedure was consistent across different stages (Fig. 3E and fig. S13D). In contrast, the TEC cluster, the major structural and functional component of the thymus, was largely populated by cells in the control, sham, and 35 dpt samples, with very few cells present at 5, 10, and 21 dpt (Fig. 3, E and F, and fig. S13D). In a differential expression analysis, there were zero genes found in the TEC cluster comparison between sham and 5 dpt (data file S9). Therefore, although most of the cell types are represented throughout the regenerative time course, the thymus-specific TEC cluster showed almost no cells at 5 dpt and slowly recovered between 10 and 35 dpt.

TEC emergence is central to thymus development and function (50). Thus, we analyzed in detail the dynamics changes in this population across regeneration by generating TEC scores. TEC scores and nearly all common TEC markers dropped strongly at 5 dpt and increased thereafter as regeneration progressed (Fig. 3G). In contrast, no major changes were observed in scores generated from tissues not expected to participate in thymic regeneration (e.g., muscle score; fig. S9B). In vivo, cells expressing the key TEC factor Foxn1 (7) were observed in thymic nodules, disappeared upon thymectomy, re-appeared as a few individual scattered cells at 5 dpt (Fig. 3H), and were eventually found exclusively within the regenerating nodules. This is consistent with the de novo formation of TECs, a potentially key event for thymus regeneration.

We next performed pairwise differential analysis between every pair of time points and grouped the resulting genes into seven clusters, which we synthesized into molecular pathways and processes (Fig. 3I and data file S10). The TEC genes higher at 5 dpt contained several collagens and genes such as Sparc and Vim, which are associated with an EMT signature (cluster G; fig. S8K). Early TEC marker Foxg1 was part of cluster F (highest at 21 dpt), whereas TEC-associated genes Cd74 and Cd83 were prominent in clusters E and B (highest at 21 and 35 dpt, respectively; fig. S8K and data file S10), suggesting progressive TEC maturation. Key enriched terms suggested the existence of distinct phases of de novo thymus regeneration based on TEC population behavior: (i) an early phase (5 dpt, no nodules) comprising wound healing, EMT, epithelial cell migration, proliferation, and thymus development; followed by (ii) a phase (10 dpt, nodule rudiment) characterized by epithelial cell differentiation, polarity, MET, and morphogenesis; and (iii) a late phase (21 to 35 dpt, expanding nodules) involving cell migration, thymopoietic processes, pharyngeal system development, and angiogenesis (Fig. 3I). Collectively, these data resolve the cellular and molecular dynamics of de novo thymus regeneration.

FOXN1 is required for thymus organogenesis but not for the formation of the thymic rudiment during development or regeneration

FOXN1 is a critical determinant of mammalian thymic function whose expression has been shown to promote TEC identity (8) and regeneration (51). We thus asked whether FOXN1 could constitute a molecular driver of de novo thymus regeneration. We generated CRISPR-Cas9–mediated mutants targeting the fourth exon of Foxn1 (Fig. 4A and fig. S11). Consistent with mammalian models (7), F2 Foxn1−/− axolotls developed miniature thymic nodules ~100 times smaller than their wild-type (WT) counterparts (Fig. 4, B and C, and fig. S12A). Foxn1−/− axolotls displayed an impairment in mature, migratory T cell production, as identified by decreased expression of Trac in the spleen, whereas, as expected, B cell development remained unimpaired (Fig. 4D). Cd8 and Cd4 levels in the spleen are very low and not necessarily exclusive to T cells (52, 53), consistent with their lack of detectable down-regulation in Foxn1−/− animals (Fig. 4D).These miniature nodules regenerated in Foxn1−/− animals after thymectomy (Fig. 4E), and although significantly smaller than WT regenerated nodules, they were of comparable size to control nodules in unthymectomised Foxn1−/− siblings (Fig. 4F). In addition, there were no significant differences in the percentage of animals exhibiting regeneration (68% Foxn1−/− versus 71% WT) or in the number of regenerated nodules (Fig. 4G, movie S4, and fig. S12B). Consistent with the impairment in thymopoiesis observed in the Foxn1−/− thymus, the regenerated thymic nodules in knockout (KO) animals contained fewer CD3+ lymphocytes relative to WT controls (Fig. 4, H to J). These data suggest that FOXN1 has a conserved role in thymus organogenesis during development and regeneration in the axolotl, yet it is dispensable for initial formation of the thymic rudiment in both contexts.

Fig. 4. FOXN1 is necessary for correct thymic organogenesis but not initiation of the regenerative program.

(A) Axolotl Foxn1 gene structure with introns and exons, highlighting the region in exon 4 targeted for mutagenesis. The expected cutting site of Cas9 is indicated with a dashed line. (B) Whole-mount imaging of optically cleared Foxn1−/− mutants compared with controls. Scale bars, 1000 μm. (C) Quantification of thymic nodule volumes (n = 3 thymuses). Unpaired two-tailed Student’s t test (D) qRT-PCR quantification of T cell–specific marker Trac and B cell marker Igm expression in the spleens of mutant and control animals (n = 3 animals). Unpaired two-tailed Student’s t test. (E) Whole-mount imaging of optically cleared regenerated thymic nodules in control and mutant backgrounds. Scale bars, 1000 μm. (F) Quantification of volumes of developing and age-matched regenerated thymic nodules in Foxn1−/− and WT siblings (n = 9 or 10 animals of each genotype). Ordinary one-way ANOVA and Student’s t test. (G) Foxn1−/− axolotls regenerate numbers of nodules comparable to those of controls. Two-tailed Mann-Whitney test. (H and I) Immunofluorescence staining showing CD3+ cells in Foxn1+/+ and Foxn1−/− axolotls (H) developing and (I) regenerating (Reg) thymic nodules. Scale bars, 50 μm. (J) Quantification of overall proportion of lymphocytes relative to total cells in the developing Foxn1−/− compared with WT nodules (n = 5 thymuses). Unpaired two-tailed Student’s t test.

MDK signaling regulates de novo thymus regeneration

To uncover potential factors underlying thymus regeneration, we performed an unbiased scRNA-seq ligand-receptor interaction analysis combined with a targeted analysis of BMP, extracellular signal–regulated kinase (ERK), and WNT, pathways known to be involved in mammalian thymus development, homeostasis (54, 55), and regeneration (56). Stromal subset cell-cell communication analysis with CellPhoneDB (57) identified 2462 significant interactions across all clusters and time points (Fig. 5, A and B; figs. S13 to S15; and data file S11), highlighting NOTCH, WNT, and midkine (MDK) within the top 5% and transforming growth factor–β and BMP signaling, albeit at lower levels (<0.25 and <0.1 mean interaction strength, respectively) (Fig. 5C and data file S11). Similar analysis per regeneration stage (figs. S14 and S15 and data file S10) identified the known thymic signaling interaction delta-like 4 (DLL4):NOTCH2 at all time points except 5 dpt, consistent with Notch signaling in TEC-mediated T cell commitment (figs. S14 and S15) (58). Although linked to early mouse cTEC development (data file S5) (22), but not previously explored in thymus development or regeneration, MDK stood out as a top interaction at 5 and 21 dpt yet was absent in the control sample (Fig. 5D, data file S11, and fig. S16, D and G), suggestive of a regeneration-specific signaling pathway. This was also interesting in light of recent findings implicating MDK in axolotl limb regeneration (59). At the cellular level, our analysis revealed MDK cross-talk between specific stromal compartments, with skin, fibroblasts, Msc+, and Col17a1+ populations as main players during early time points (Fig. 5D). In agreement, Mdk began to be expressed in the injured basal layer of the epidermis at 3 dpt and, notably, in single cells in the mesenchyme at 5 dpt (Fig. 5E), resembling our observation of Foxn1+ cells in the region where the new nodule formed (Fig. 3H). Throughout regeneration, Foxn1+ cells largely coexpressed Mdk and its receptor Ptprz1 in the Col17a1+ population (Fig. 5D and fig. S15D). The Col17a1+ population shared several features with the epithelial-mesenchymal hybrid phenotype, recently described in human TECs (3, 60). To further investigate this population of putative progenitors, we analyzed the spatial expression of key markers from the Col17a1+ population, including Col17a1, p63 (an epithelial stem cell marker), and Krt17 (injured basal epidermis marker), which were localized in the domain shared by Mdk and Foxn1 at 5 dpt (fig. S16A), suggesting that cells in these domains could be contributing to de novo TEC formation.

Fig. 5. MDK signaling is necessary for de novo thymus regeneration.

(A) UMAPs showing the cell types in the stromal subset of the 5-, 10-, 21-, 35-, and 60-dpt control and regenerated thymus. (B) Heatmap showing the log number of significant interactions between cell types containing ligands on the x axis and those containing receptors on the y axis. (C) Dotplot indicating the top 5% significant (P value < 0.05) ligand-receptor signaling interactions between relevant pairs of cell types across all time points. On the y axis, the first gene indicates the ligand, and the second indicates the receptor. On the x axis, the first entry indicates the source cell type (containing the ligand), and the second indicates the target cell type (containing the receptor). (D) Communication plots showing the MDK:PTPRZ1 interaction across time points. (E) HCR-ISH showing expression of Mdk and Ptprz1 at 1, 3, and 5 dpt. Arrowheads denote examples of coexpression of ligand and receptor seen both in basal epidermis and thymectomized area at 5 dpt. Scale bars, 50 μm. (F) Schematic of functional assessment of role of signaling pathways in thymus regeneration. (G to K) Effects of perturbation of different signaling pathways using small-molecule inhibitors for the duration of thymus regeneration. Animals treated with (G) C59 (WNT pathway inhibitor; n = 23 thymectomized areas), (H) U0126 (ERK/MAPK pathway inhibitor; n = 21), (I) DMD (BMP pathway inhibitor; n = 39), and (J) iMDK (MDK pathway inhibitor; n = 23) for 21 days or (K) iMDK for 10 days (n = 33) (P values from bootstrapping test). (L) Representative image of iMDK treatment phenotype (arrowheads indicate thymic nodules) visualized in Krt12:eGFP animals. Scale bars, 500 μm.

To test the requirement of MDK and selected candidates [WNT, ERK/mitogen-activated protein kinase (MAPK), and BMP] during de novo thymus regeneration, we performed loss-of-function analysis with inhibitors previously used and characterized in the axolotl [inhibitor of MDK (iMDK), C59, U0126, and dorsomorphin dihydrochloride (DMD), respectively] (59, 61–63). We injected Krt12:eGFP transgenic juvenile animals interperitoneally with the aforementioned inhibitors every 2 days throughout the time course of regeneration (Fig. 5F and fig. S16B). The functionality of the inhibitors was validated using quantitative reverse transcription polymerase chain reaction (qRT-PCR) and Western blot analysis to determine direct effects on downstream targets (fig. S16, H and I). Whereas no significant effects were observed upon WNT or ERK inhibition (Fig. 5, G and H), the abrogation of BMP signaling and MDK inhibition throughout the regenerative process led to reductions in both the number of animals able to regenerate and the regenerating nodules per animal (Fig. 5, I to L, and fig. S16C). Mdk expression was readily up-regulated upon injury, peaking at 1 to 3 dpt (Fig. 5E and fig. S16D), preceding the appearance of Foxn1+ cells (Fig. 3H and fig. S16F). Further, its receptor Ptprz1 peaked in expression at 5 dpt (Fig. 5E and fig. S16D), suggesting a model whereby early Mdk up-regulation is critical for de novo TEC induction and thymus regeneration. Consistent with this, iMDK treatment exclusively during the first 10 dpt resulted in substantially impaired regeneration (Fig. 5K). Collectively, these data uncover a requirement for BMP and MDK signaling during de novo thymus regeneration, identifying MDK as a previously undescribed driver of this process.

Last, we asked whether MDK signaling could be relevant to thymus formation in mammalian contexts. The analysis of top ligand-receptor interactions (figs. S17 and S18) using mouse and human scRNA-seq datasets (23, 64) uncovered MDK signaling involvement in cell type–specific stromal communication at key stages in both species (at 13 weeks in humans and from E13.5 onward in mice). Understanding whether MDK signaling is involved in and can be leveraged for mammalian thymic regeneration would present possible avenues for therapeutic strategies.

DISCUSSION

The possibility of thymus regeneration in salamanders was hinted at over 50 years ago (65), yet its features and whether this could occur de novo remained unclear. Here, we demonstrate that the juvenile axolotl thymus undergoes regeneration upon removal, with restoration of morphology, cell type diversity, and functionality. Although our histological, morphological, and molecular analysis of the region postthymectomy suggests the complete removal of the thymus organ, further supported by the variability in regeneration outcomes, we cannot, with complete certainty, exclude that there were no single thymic cells remaining after surgery. Nevertheless, organ removal was sufficiently extensive to merit the term “complete,” and thus, we refer to the regeneration of the thymus as de novo.

Through our single-cell resolution atlas of the axolotl thymus, we define the molecular traits of this organ and offer insights into axolotl thymopoiesis of direct relevance to understanding of adaptive immunity evolution. We found that several thymic features and functions are conserved between axolotls and mammals and identified points of divergence, including a lack of clear cortico-medullar junctions at the spatial and cellular levels, the implications of which are as yet unknown. The relative low number of epithelial cells collected in contrast with lymphoid cells represents a limitation of scRNA-seq collection, which could have prevented the identification of molecular differences between TEC subpopulations. This should be further investigated in the future, possibly with scRNA-seq and spatial coexpression analysis. Although direct cross-species comparisons are limited by differences in developmental mode, evolutionary distance, and completeness of RNA-seq datasets, we compared the axolotl and mouse thymus at the most molecularly and morphologically similar stages. From a developmental point of view, the mouse thymus is still immature at E13.5, which is why this stage can be used for re-aggregate thymic organ cultures and fetal thymic organ cultures. After E13.5, the thymus progressively becomes structurally complete and functionally mature (22, 66, 67). The mouse thymus continues to mature in terms of cell types and gene expression until and including after birth. The axolotl thymus completes organogenesis between 3 and 6 weeks posthatching, after which it continues to grow in size but does not appear to change its architecture (fig. S3B). These considerations, together with our in-depth molecular characterisations, guided our comparative approach. Future work should address to what extent thymic populations continue to mature in axolotls and refine the staging equivalence across species. Last, whether thymus involution takes place in this species of heightened regenerative abilities remains an important open question.

We identified critical events during thymus regeneration, most notably the appearance of de novo TEC cells, which we suggest act as a nucleation center for nodule formation. In agreement, this population expresses MDK and its receptor, placing it centrally in the stromal cross-talk required for thymus regeneration. The origin of this population is still outstanding; however, we expect this could involve de- or transdifferentiation events, which often underlie progenitor formation in salamanders (14, 68, 69), a hypothesis that should be addressed by future research.

When contrasting mammalian (remnant-dependent) and axolotl (de novo) thymus regeneration, we identified both potentially conserved (e.g., BMP necessity) and divergent (e.g., lack of WNT and ERK involvement) molecular requirements. We uncovered MDK as a driver of de novo thymus regeneration. Its specific function during early regeneration stages suggest that this dependency may be linked to the de novo process. Although its function in mammalian contexts is unknown, its presence during key stages of mouse and human embryogenesis suggests that it could potentially be leveraged in developing regenerative therapies. Thus, the insights hereby provided could be relevant to clinical cases involving full thymectomy, such as myasthenia gravis, myoma and pediatric heart surgery, and age-related thymus involution, although this requires direct investigation.

Together, our study uncovers de novo regeneration of a lymphoid organ within a vertebrate. Understanding this process’s molecular and cellular underpinnings may serve as a platform to identify potential therapeutic targets for improving thymic function in aging or immunocompromised patients.

MATERIALS AND METHODS

Study design

The study aimed to thoroughly investigate the process of de novo thymus regeneration in juvenile axolotls from a morphological, cellular, and molecular standpoint. To do so, we surgically removed the thymus in transgenic axolotls and used whole-mount clearing and light-sheet microscopy, histology, and scRNA-seq to characterize the restoration of the organ. We transplanted regenerated thymuses from fluorescent animals to white animals to determine the ability of the restored organ to home new hematopoietic progenitors and form mature T cells capable of migrating throughout the body. scRNA-seq analysis of the regenerative time course of the thymus was used to confirm de novo TEC generation and identify putative cell types that could be involved in their restoration. To evaluate the role of Foxn1, a gene crucial for mammalian thymus development and TEC maintenance, during axolotl regeneration, we generated axolotl Foxn1 KOs via CRISPR-Cas9. To determine which signaling pathways may play a role in de novo thymus regeneration, we used cell-cell communication analysis leveraging scRNA-seq dataset and signaling pathway inhibitors to quantify their impact on regeneration. Animal numbers per group are indicated in the respective figure legends.

Animal husbandry

Animal procedures and husbandry were performed following the ARRIVE guidelines and in compliance with the current German Animal Welfare Act and legislation from the State of Saxony, Germany, under the licenses TVV 21/2019 and TVV 47/2024 issued by the animal welfare authorities. Axolotls (Ambystoma mexicanum) used in this study came from the axolotl facility at TUD-CRTD Centre for Regenerative Therapies Dresden (Germany). Leucistic (d/d) strain axolotls and transgenic d/d animals (as specified) were used in all experiments.

Animal procedures

tgSceI(Xla.KRT12:eGFP)ETNKA transgenic axolotls (21) or white d/d axolotls between 2 and 5 cm in length (6 to 8 weeks at start of experiments) from snout to tail tip were used for most experiments unless specified otherwise. At this stage, the axolotl thymus is both well-formed (three nodules distinctly separated and at final location) and accessible for surgical removal due to lack of skull bone thickening and skin translucency. The juvenile axolotls (of both sexes) were anesthetized in 0.01% benzocaine (Sigma-Aldrich) before each procedure, including thymectomies, inhibitor injections, and live imaging. Thymectomies were performed surgically using dissection scissors and forceps and a tungsten needle. First, a skin flap was cut from the base of the third gill toward the otic capsule and peeled back to reveal the three thymic nodules. The nodules were then removed manually using forceps, and the tungsten needle and the skin flap were used to cover the wound site. Axolotls were then allowed to regenerate at 20°C.

Inhibitor delivery was performed by intraperitoneal injections (amount adjusted according to size; 3 μl was used as a starting point in 2- to 2.5-cm animals) every 2 days throughout the regeneration procedure (unless specified otherwise) at concentrations used in previous studies. Specifically, C59, which is an inhibitor of porcupine necessary for WNT activity, was used at 10 μM (Abcam, ab142216). DMD (Tocris, 3093), which selectively inhibits BMP-type receptors, was used at a concentration of 2 μM. An inhibitor of phosphatidylinositol 3-kinase, which potently inhibits the MDK growth factor, iMDK, was used at 10 μM (Tocris, 5126). Last, a 50 μM concentration was used for the MAPK kinase inhibitor U0126 against the ERK pathway (Sigma-Aldrich, 109511-58-2). Dimethyl sulfoxide (DMSO; 1%) was used as a control in all experiments. Transplantation experiments were conducted using white or KRT12:-eGFP axolotls as hosts and either CAGGS:EGFP [tgSceI(CAGGs:eGFP)ETNKA] (70) or CAGGS:mCherry [tgSceI(CAGGs:Cherry)ETNKA] (13) transgenic animals ubiquitously expressing fluorescent reporters as donors. Control or regenerated fluorescent thymic nodules were transplanted into thymectomized white hosts and allowed to integrate, and then the limbs were amputated at the upper arm level to follow the contribution of fluorescent cells derived from the transplanted thymus into the regenerating limb. For all experiments on fixed samples, tissues were collected and fixed overnight at 4°C in 4% paraformaldehyde.

CRISPR KO of Foxn1

The Foxn1 gene was located in the axolotl genome [assembly v3.0.0 (BioNano)], and introns, exons, and the coding sequence were mapped with the use of the axolotl-omics site (transcriptome version Am_3.4). Two different guide RNAs (gRNAs) targeting exon 3 (FoxN1-KO-g1) or 4 (FoxN1-KO-g5) were designed with the Broad Institute CRISPick tool (71) and aligned against the axolotl genome to ensure the uniqueness of the sequences. The gRNAs were cloned into the DR274 plasmid (Addgene, 42250) (72) and synthesized using the MEGA ShortScript T7 Kit (Invitrogen). Further, gRNAs were prepared for injection by mixing 5 μg of gRNA and 5 μg of Cas9 in 10 μl of ribonuclease (RNase)–free water (final volume) and warmed up at 37°C for 5 min before injection. A total of 5 nl of each gRNA was injected in single-cell stage axolotl eggs, with a minimum of 50 eggs injected from two different batches. Animals were allowed to grow at 20°C for 3 weeks, when tail clips were performed for genotyping and characterization of the generated mutations. Genomic DNA was prepared from tail clips using the Mouse Direct PCR Kit (Biotool), and the target regions in exon 3 or 4 were amplified by PCR. For exon 3, we used the primer pair FoxN1_E3-scr_F and FoxN1_E3-scr_R, whereas for exon 4, we used FoxN1_E4-scr_F and FoxN1_E4-scr_R. Amplicons were sequenced by Illumina short-read, paired-end sequencing, and mutations were analyzed with the CRISPResso2 tool (73). On average, 5.2% (n = 38) of animals injected with FoxN1-KO-g1 and 38.2% (n = 136) of animals injected with FoxN1-KO-g5 incorporated a diversity of frameshift mutations with more than 50% of penetration. These frameshift mutations yielded early stop codons at amino acids 68 to 78, upstream of the Forkhead domain, which spans amino acids 282 to 378 according to the InterPro database (74). Only mutants derived from the FoxN1-KO-g5 gRNA injections (targeting exon 4) successfully inherited mutations to the F1 generation. FoxN1-KO-g5–derived mosaic animals with three different mutant alleles (5-, 7-, and 8-nt deletions) were raised, and F1 heterozygous (WT) or KO animals were bred for further characterization. We used the standardized transgenic animal nomenclature recommended by the international community of salamander researchers.

scRNA-seq methods summary

The details of scRNA-seq sample collection, scRNA-seq alignment and gene annotation, scRNA-seq quality control and filtering, scRNA-seq data processing and analysis, control-regenerated thymus analysis, control thymus analysis and staging the regenerated thymus with the control, control-regenerated T lymphocyte analysis, scRNA-seq quality control of thymic regenerating time course, scRNA-seq analysis of thymic regenerating time course, scRNA-seq data processing for ligand-receptor analysis, and scRNA-seq ligand-receptor analysis can be found in Supplementary Materials and Methods.

Antibody staining on cryosections

Whole head or thymic area cryosections of 10-μm thickness were rehydrated and permeabilized by using phosphate-buffered saline (PBS) with 0.3% Triton X-100 and then blocked for 1 hour with PBS with 0.3% Triton X-100 and 10% heat-inactivated normal goat serum (Sigma-Aldrich). Slides were