A Critical Role for the Regulated Wnt–Myc Pathway in Naive T Cell Survival

Hematopoietic progenitors, from the bone marrow, migrate into the thymus and undergo a well-regulated developmental program to produce T lymphocytes (1, 2). Once functionally mature, T lymphocytes emigrate from the thymus to populate the spleen and lymph nodes, where they wait for stimulation by their cognate Ag to mount a protective immune response (3, 4). Although the developmental and activation programs have been well characterized, the program that maintains peripheral naive T cells in the resting stage remains poorly understood. Recent studies from our team and others have implicated critical roles in regulation by mammalian target of rapamycin (mTOR) activity (5, 6) and by the FoxO and Foxp1 transcription factors (7–9).

The Wnt signaling pathway is an evolutionarily conserved pathway that regulates cell proliferation, differentiation, survival, migration, and polarity (10–13). Wnt stimulation releases β-catenin from a destruction complex scaffolded by adenomatous polyposis coli (Apc), thus allowing β-catenin to regulate its transcriptional targets by interacting with T cell factors, such as Tcf-1 (14–16). Mice lacking different components of the Wnt signaling pathway reveal a broad dysfunction in various stages of T cell development, including the generation of double negative CD4⁻CD8⁻ (DN) thymocytes and differentiation/survival of multiple functional T cell subsets in the periphery. Tcf1-deficient mice show an age-dependent reduction in thymocyte production and a corresponding loss of early thymic progenitors (17). Deletion of Ctnnb1 (the gene that encodes β-catenin) results in a developmental blockage at the DN3–DN4 stage (18). In activated T cells, ectopic expression of the β-catenin partner, Tcf1, stimulates differentiation to Th2 (19), whereas that of a Wnt signaling inhibitor, Dkk-1, abrogates it (20). Ectopically expressing a β-catenin mutant that evades Apc-mediated destruction also enhances the survival of T regulatory cells (21). A recent study suggests that heterozygous mutation of the Apc gene in the Apc^min/+ mice partially attenuates regulatory T cell function (22). Perhaps because of the difficulties in deleting Apc in mature naive T cells, the role for Wnt signaling in mature naive T cells in the periphery has not been investigated.

To address this gap, we used mice with exon 14 floxed Apc locus (23) and a CD4-Cre transgene to induce exon 14 deletion in the T cell lineage (24). Deleting exon 14 in Apc generates a truncated polypeptide that lacks most of the functional domains of Apc (25), including seven repeated sequences of 20 aa, each in the central region of the Apc protein. Because these repeats are critical for Apc binding to β-catenin, the key step in canonical Wnt signaling (23), the mutant cells will have constitutive activation of the Wnt pathway. The mutant also lacks the binding sites for EB1 and microtubules that are responsible for cell polarity and mitosis (26, 27). The truncated Apc is still capable of encoding a polypeptide that contains the oligomerization domain (28) and some of the armadillo repeats, which have been shown to interact with the Apc-stimulated guanine nucleotide exchange factor (Asef) (29). Thus, although the truncated Apc may still have a role in stabilization and motility of the actin cytoskeleton network through its interaction with Asef and Rac and Rho GTP binding proteins (30), the essential role for Apc in canonical Wnt signaling is completely inactivated. This tool provided us with a unique opportunity to investigate the role of Wnt signaling in naive T cell function. Surprisingly, we found that deletion of exon 14 of the Apc gene, using CD4-Cre, activated Wnt signaling without affecting T cell development. Our data revealed that inactivation of Apc resulted in a drastic loss of mature naive T cells in the periphery and severe T cell lymphopenia. This loss is due, at least in part, to overexpression of cMyc, as it is attenuated by deletion of the cMyc gene. Our data unveil an unexpected impact of Wnt signaling on the survival of naive T cells in the periphery.

CD45.1 C57BL/6 mice were obtained from Charles River Laboratories, through a contract with the National Cancer Institute. Mice with homozygous knockin of the floxed Apc (23) and transgenic mice expressing the Cre recombinase, under the control of either the proximal Lck promoter (31) or CD4 promoter (24), were obtained from The Jackson Laboratory. Mice with floxed cMyc locus (32) were kindly provided by Dr. De Alboran (National Center for Biotechnology CNB/CSIC, Madrid, Spain). All mice used in this study have been backcrossed to C57BL/6 background for ≥10 generations. These strains were maintained in our animal facilities under pathogen-free conditions. All experiments were performed in accordance with the guidelines of the Institutional Animal Care and Use Committees of the University of Michigan and the Children’s National Medical Center.

Five-week-old mice were anesthetized with isoflurane. A midline incision was used to expose the ribs. A total of 10 μl 1 mM CFSE (Life Technology) was injected into each lobe of the thymus. Mice were sacrificed at either 6 or 24 h after CFSE injection. The thymus, spleen, and lymph nodes were collected and analyzed by flow cytometry.

Mature CD24⁻ thymocytes were isolated by Dynal negative selection using biotinylated anti-CD24 mAb M1/69 (eBioscience). Cells were labeled with 10 μM CFSE and injected i.v. into CD45.1 recipients (1 × 10⁶ per mouse).

Tissues were lysed in a protein lysis buffer (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 0.5% NP-40) supplemented with a protease inhibitor mixture (Pierce). Cell lysates were separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes. The membranes were incubated with anti-cMyc (Abcam, AU016757, Ab39688, 1:2000) and anti-actin (Abcam ACTN05, Ab1801, 1:5000). Either anti-rabbit or anti-mouse IgG HRP-linked Ab (Santa Cruz) was used as a secondary Ab. A chemiluminescence kit (Life Technologies) was used to visualize blots.

Mice were sacrificed at 6–8 wk of age. Thymus and spleen tissues were homogenized to generate a single-cell suspension. Cells were stained at 4°C for 20 min in PBS with 2% FBS with the following Abs from BD Bioscience (1:200): CD4 (RM4-5), CD8 (SK-1), B220 (HIS24), CD44 (IM7), CD62L (MEL-14), CD24 (30-F1), CD69 (H1.2F3), CCR7 (4B12), β7 integrin (FIB205), β-catenin (15B8), δϒ TCR (UC7-13D5), and CD25 (PC61.5). Samples were analyzed by a BD LSR II Flow Cytometer.

For the colocalization experiment, splenocytes were stained for CD3 and permeabilized using the BD Cytofix/Cytoperm Kit followed by anti–β-catenin (15B8) staining overnight. DRAQ5 (eBioscience) was added (1:10,000) before analysis. Samples were analyzed by Amnis Imaging Flow Cytometry. Analysis was carried out using IDEAS software. Calculations were made by following the IDEAS software wizard.

CD45.1⁺ B6 mice were lethally irradiated [11 Gy delivered in two installments, 4 h apart, as reported (33)] and reconstituted with 2.5 × 10⁶ CD45.1⁺ bone marrow cells in conjunction with equal numbers of either Apc^fl/flCD4-Cre⁺ (cKO) or Apc^fl/flCD4-Cre⁻ (Ctrl) bone marrow cells. Reconstitution was confirmed at 8 wk from peripheral blood staining of CD45.1 and CD45.2. Bone marrow chimeras were analyzed for T cell numbers and phenotypes at 10 wk post transplantation.

Data were analyzed using two-tailed unpaired Student t tests. All statistics were performed using GraphPad Prism, version 5 (GraphPad Software). The p values were as follows: *p < 0.05, **p < 0.01, ***p < 0.001.

A previous study suggested that deletion of Apc using Lck-Cre disturbs thymocyte development (34). To study the impact of Wnt signaling in mature naive T cells, we used CD4-Cre to delete Apc and used a more commonly known Apc conditional allele that has loxP sites inserted into introns 13 and 14 of the endogenous Apc gene (Apc^fl) (23). These mice were labeled as cKO, whereas their Cre⁻ littermates were used as a control and labeled as Ctrl. On the basis of the known specificity of the Cre promoter, Apc^fl/fl;CD4-Cre⁺ mice should have a frameshift mutation at codon 580 and a truncated Apc polypeptide beginning at the CD4⁻CD8⁻CD25⁺CD44⁻ (DN3) stage. To confirm Apc deletion among thymocytes, we FACS sorted DN thymocytes based on their expression of CD4, CD8, CD44, and CD25. Using PCR that yielded differently sized products from floxed (fl) and deleted (del) Apc alleles, we determined the kinetics of Apc deletion in developing DN thymocytes. As expected, no deletion was observed in DN1–4 of the Ctrl mice. In the cKO thymocytes, nearly equal amounts of fl and del alleles were observed in DN3. By DN4, most products were derived from the del allele (Fig. 1A). Thus, the Apc deletion began at the DN3 stage and was largely completed at the CD4⁻CD8⁻CD25⁻CD44⁻ DN4 stage in the cKO mice. Because the primary function of the Apc protein is to mediate the destruction of β-catenin, we used intracellular accumulation of the protein as the primary readout for functional Apc inactivation throughout thymocyte development. As shown in Fig. 1B, despite a nearly 50% deletion of the Apc gene, no increase of intracellular β-catenin was observed in the DN3 stage. Likewise, although the Apc deletion was nearly complete in the DN4 stage, only a small subset showed functional inactivation of the Apc protein. However, significant accumulation of β-catenin was observed from the bulk of double positive CD4⁺CD8⁺ (DP), single positive CD4⁺CD8⁻ (CD4 SP), and single positive CD4⁻CD8⁺ (CD8 SP) thymocytes. The delayed functional inactivation of the Apc protein in relation to the gene deletion is likely due to residual Apc protein and/or mRNA. To investigate whether β-catenin accumulation is adequate for its translocation into the nucleus, we stained for CD3 and β-catenin in the splenocytes from Ctrl and cKO mice. Using Aminis Imaging flow cytometry, we detected the presence of β-catenin in the nucleus, which was also stained with a nuclear stain, DRAQ5. As seen in Fig. 1C, cKO clearly results in an accumulation of β-catenin that in >50% of cases colocalizes in the nucleus, thus demonstrating that without APC, β-catenin will stabilize and translocate into the nucleus, thereby activating Wnt signaling. Regardless, the robust activation of Wnt signaling from the DP stage allowed us to assess the function of Wnt signaling in T cell development and function starting at the DP stage.

As shown in Fig. 1D, Ctrl and cKO thymi had comparable cellularity. Distribution of CD4 and CD8 markers among thymocytes revealed no gross abnormality in thymocyte development, as the percentages (Fig. 1E) and numbers (Fig. 1F) of DP, CD4 SP, and CD8 SP thymocytes were similar. As expected from the largely normal β-catenin levels in cKO DN thymocytes, the frequency (Fig. 1G) and number (Fig. 1H) of DN cells were also comparable throughout stages 1–4. Thus, the deletion of Apc that began at the DN3 stage did not affect thymocyte development. However, we did observe an increase in apoptosis of Apc-deficient thymocytes (Supplemental Fig. 1A), although apoptosis was more pronounced in peripheral T cells (Supplemental Fig. 1B).

The differences observed between our results and those in a previous study (34) were not entirely due to a subtle difference between the Lck-Cre used in the previous study and the CD4-Cre used in this study. In our hands, both CD4-Cre and Lck-cre had largely normal thymocyte development. (Supplemental Fig. 2). However, the modest but statistically significant reduction in cellularity in Lck-Cre mice suggests that even minor differences in timing of gene deletion can affect the outcomes. One potential phenotypic difference between Lck-Cre and CD4-Cre is in the expression of CD44, a known Wnt target gene. In the periphery, both CD8 and CD4 T cells from Lck-Cre and CD4-Cre upregulate CD44 at much higher levels than in the Ctrl mice. However, thymocytes show a different pattern, in which SP CD4 and CD8 thymocytes in CD4-Cre mice express much higher levels of CD44, whereas Lck-Cre levels of CD44, although higher, appear much more like those of Ctrl mice. In addition, we observed a loss of thymic cellularity and an apparent increase of DN thymocytes in older CD4-Cre–induced cKO mice that became moribund (Supplemental Fig. 1C, 1D). This raised the intriguing possibility that abnormal thymocyte development may be a secondary effect of overall immune activation in the mice, as reported by Martin et al. (35). Overall, our data demonstrate that Wnt signaling starting at DN4 did not directly disturb normal T cell development in healthy young mice. This normal T cell development allowed us to study the function of Wnt signaling after T cell maturation.

Despite normal T cell development, we observed a marked reduction of T cell numbers in the periphery. Thus, the frequencies of CD4⁺ or CD8⁺ T cells in the cKO mice were reduced by 10-fold or more in comparison with frequencies in the Ctrl mice (Fig. 2A). In addition to the loss of CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells, approximately one-third of the T cells in the cKO mice lost expression of CD4 and CD8 (Fig. 2B). This loss of CD4 and CD8, however, is not the result of a lineage switch to γδT cells, as cKO mice have similar numbers of γδ T cells in the thymus and the spleen (Supplemental Fig. 3A, 3B). The ontogeny of these cells remains to be determined. Moreover, in both CD4⁺ (Fig. 2C) and CD8⁺ (Fig. 2D) T cell compartments (Supplemental Fig. 4A, 4B), a profound loss of naive T cells was observed in the spleen. A similar reduction was observed in the lymph nodes (Supplemental Fig. 4C–E). Although >60% of the T cells in Ctrl mice exhibited the naive phenotype, CD44^loCD62L^hi, ∼5% or less of the cKO T cells are naïve. In fact, most of the cKO T cells displayed markers associated with effector memory T cells (CD44^hiCD62L^lo), although significant increases in the central memory compartment were also observed. In addition, expression of CD69 was also elevated in the cKO T cells (Fig. 2E).

The loss of T cells in the periphery could be due to defects in either T cell emigration or survival. To elucidate whether the rate of thymic output was reduced in the Apc-deficient thymocytes, we injected CFSE intrathymically and tracked recent thymic emigrants in the periphery. To avoid potential artifacts associated with peripheral labeling, we monitored the labeling of the thymus, excluding any samples that had <3% labeling of either CD4 SP or CD8 SP thymocytes. Moreover, because the efficacy of CFSE labeling of thymocytes is variable in different mice, we normalized the percentage of CFSE⁺ T cells in the lymph node by dividing them by the percentage of CFSE⁺ of either CD4 or CD8 SP thymocytes. At 6 and 24 h after CFSE injections, we detected comparable frequencies of CFSE⁺ T cells in the lymph nodes of cKO and Ctrl mice (Fig 3A, 3B). These data demonstrate that T cell emigration from the thymus was unaffected by the deletion of the Apc gene and suggest that the reduced accumulation of Apc-deficient T cells was due to loss of T cells after their emigration from the thymus.

To confirm this idea, we isolated CD24⁻ mature thymocytes from CD45.2⁺ cKO and Ctrl mice, labeled them in vitro with CFSE, and injected them i.v. into CD45.1⁺ congenic hosts and followed their persistence in peripheral blood and lymphoid organs. As shown in Fig. 4A, upper panel, similar numbers of CFSE⁺ Ctrl and cKO T cells were observed in the spleen of the congenic host at 6 h after transfer. By 24 h, the numbers of cKO T cells in the spleen were less than one-third of their Ctrl counterparts. The frequencies of cKO T cells in the pooled lymph nodes were reduced by >10-fold (Fig. 4B). The rapid loss of T cells was attributable to cell death, as there was a higher percentage of 7-aminoactinomycin D (7-AAD)⁺ and/or Annexin V⁺ T cells in the cKO mice (Fig. 4C, 4D).

To determine whether the loss of peripheral T cells in the cKO mice was cell intrinsic, we created bone marrow chimeras using cKO or Ctrl donor-type (CD45.2) cells mixed with an equal number of recipient-type (CD45.1) bone marrow cells. After 10 wk, we harvested the thymi and spleens to investigate the reconstitution of CD45.2 donor cells. As shown in Fig. 5A, the percentage of CD45.2⁺ cells in the thymus was unaffected by the deletion of Apc. Based on the distribution of CD4 and CD8 markers, the development of both Ctrl and cKO thymocytes was grossly normal, as all major subsets were present among CD45.2⁺ thymocytes (Fig. 5B). However, the percentage of SP thymocytes was reduced by ∼2-fold (Fig. 5B), perhaps owing to a subtle competitive disadvantage of the cKO thymocytes. Among SP thymocytes, the percentages of Qa-2^hi and CD24^lo thymocytes were comparable between cKO and Ctrl groups (Fig. 5C).

In sharp contrast to a roughly normal thymocyte development, severe cKO T cell loss was observed in the spleens of chimera mice. Thus, although the frequency of CD45.2 donor leukocytes was unaffected by Apc deletion (Fig. 6A, 6B), >20-fold reduction was observed in the frequency of T cells within the CD45.2 splenocytes (Fig. 6C, 6D). In addition to loss of T cell cellularity, the remaining T cells exhibited a marked difference in cell surface markers. Even though the majority of the Ctrl T cells displayed CD44^loCD62L^hi markers of naive T cells, most of the cKO T cells had either central or effector memory markers (Fig. 6E, 6F). Loss of naive T cell markers and acquisition of memory/effector T cell markers in cKO T cells were reminiscent of what was observed in the Ctrl and cKO mice. In contrast, cellularity of the recipient-type T cells was not affected in a trans fashion by the genotype of the donor-type cells in the recipients (Supplemental Fig. 3C, 3D). Because the cKO T cells have developed and emigrated into a normal environment, both the loss of cellularity and the acquisition of activation markers following Apc deletion are cell intrinsic.

cMyc is a β-catenin target gene that causes T cell apoptosis through Fas/FasL interaction, death ligand TNF, and TNF-related apoptosis-inducing ligand (36–38). As the first step to investigate its potential contribution to lymphopenia in cKO mice, we evaluated cMyc overexpression in CD24⁻ thymocytes from Ctrl and cKO mice by quantitative PCR. As shown in Fig. 7A, the cMyc transcript was doubled in cKO mice. However, cMyc protein was not obviously overexpressed among the mature thymocytes. In contrast, peripheral cKO T cells had drastically higher levels of cMyc than did Ctrl T cells (Fig 7B). We therefore generated the cMyc^fl/+Apc^fl/fl, CD4-Cre⁺ (Myc^+/−cKO) mice to diminish the impact of transcriptional activation of cMyc by Wnt signaling. Heterozygous deletion of cMyc resulted in a substantial reduction of the cMyc protein when compared with cKO T cells (Fig. 7B). Nevertheless, heterozygous deletion of cMyc partially restored the numbers of cKO T cells in the periphery (Fig. 7C). An ∼3-fold increase of both CD4 and CD8 T cells was also observed in cMyc^+/− cKO mice in comparison with cMyc^+/+cKO mice (Fig. 7D). To further demonstrate that overexpression of cMyc caused apoptosis of Apc^−/− T cells, we adoptively transferred either Apc^−/− or Apc^−/−cMyc^+/− mature thymocytes into wild-type mice and followed the survival of the T cells in the host at 6 h postinjection. As shown in Fig. 7E, Apc^−/−cMyc^+/− T cells exhibited reduced apoptosis and therefore increased survival in the wild-type hosts. These data demonstrate that poor survival of Apc^−/− T cells can be rescued by downregulation of cMyc expression. In addition to an overall increase in T cell numbers, the percentage of naive CD4 and CD8 T cells increased by ≥3-fold, whereas that of naive CD8 T cells increased by 50% (Fig. 7F, 7G). A corresponding reduction of central and effector memory T cells was also observed in CD4 T cells, although the accumulation of effector memory CD8 T cells was less affected by cMyc deletion (Fig. 7F, 7G).

Our data in Fig. 7B demonstrated that heterozygous deletion of cMyc failed to reduce cMyc protein levels in cKO mice to wild-type levels. To determine whether the partial effect of a heterozygous cMyc deletion was due to an insufficient reduction in cMyc protein, we generated mice with a homozygously floxed cMyc locus. Deletion of cMyc had no effect on thymocyte development (Fig. 8A) and T cell cellularity in the periphery (Fig. 8b), although a reduction of CD44 levels was observed on both CD4 and CD8 T cells (Fig. 8c). Combinational deletion of both Apc and cMyc did not affect thymocyte development (Fig. 8D). However, a homozygous deletion of cMyc resulted in a further rescue of T cell survival defects in the cKO mice, although it did not completely restore T cell cellularity and survival defects (Fig. 8E, 8F).

Previous studies demonstrated that β-catenin overexpression produces multiple developmental defects in thymocyte development (39–43). The most dramatic phenotype was reported in mice with a deletion of the Apc locus, with largely depleted mature T cells and immature DP thymocytes (34). Surprisingly, with the use of a different Cre system (CD4-Cre rather than Lck-Cre) and floxed Apc allele, we observed no changes in either thymic cellularity or percentages and numbers of DN, DP, and CD4/8 SP thymocytes. We took advantage of the largely normal T cell development in mice with CD4-Cre–driven deletion of the Apc exon 14 to show an unappreciated role for regulated Wnt signaling in survival of naive T cells, including a massive loss of cellularity and a lack of the naive T cell phenotype. Because CD44 is a target for Wnt signaling (44), it is of interest to consider whether loss of the naive T cell phenotype is a reflection of increased Wnt signaling. It should be recognized that Wnt signaling may directly upregulate CD44 expression, so the “true” naive T cell phenotype may be masked. We consider it unlikely, as the overwhelming majority of CD44⁺ T cells also lost CD62L expression. The concurrent loss of CD62L, massive loss of cellularity, and high levels of apoptosis are more consistent with activation-induced cell death.

Because the phenotype of the Apc deletion is significantly attenuated by simultaneous deletion of the cMyc gene, at least part of the T cell loss is due to overactivation of cMyc. The homozygous deletion of cMyc in T cells was insufficient to restore T cell cellularity, so additional signaling pathways downstream of Wnt signaling may be involved. In this context, it is of interest to note that Wnt signaling is upstream of the mTOR pathway (12), and studies by us and others have established a critical role for mTOR in survival of naive T cells (5, 6). Additional studies are needed to establish whether mTOR and cMyc constitute two parallel pathways for defective T cell survival caused by unregulated Wnt signaling.

It is of note that unlike the high rates and early onset of thymic lymphoma reported by another group (45, 46), we observed only rare cases of late-onset thymic lymphoma (data not shown). However, our data are consistent with lack of lymphoma in the Apc^min/+ mice (47). Additional studies are needed to reconcile the differences. Because the previously reported lymphoma requires the Rag recombinase that peaks at the DN3 stage (46), and because the β-catenin activation was not detectable at DN3 in our model, we suspect that the lack of concurrent Rag and β-catenin activities explains the lower risk of thymic lymphoma in our model.

A previous report showed a dramatic arrest of thymocytes at the DN4 stage by an Apc deletion (34). The developmental block of thymocytes was also noted using retroviral vector overexpressing stabilized β-catenin (39, 40). However, this phenotype was not recapitulated when the interaction between β-catenin and Apc was inactivated by deletion of the Apc binding domain on β-catenin (48). Our data clearly demonstrated that deletion of the Apc gene, starting at the DN3 stage, had no discernible impact on thymocyte development, as revealed by the comparable cellularity and a normal distribution of the CD3, CD4, CD8, CD44, CD25, CD24, and Qa2 markers. One likely explanation is the kinetics of functional inactivation of Apc. The previous study found β-catenin increased in most of the cells from the DN3 stage and on, whereas our study observed an increase in only a small fraction of DN4 and all of the DP and SP thymocytes. The delayed functional inactivation of Apc in our model allowed normal thymocyte development. The reason for the delayed activation of Wnt signaling is unknown, as both the Cre-driver and the Apc^fl differed in these studies. Because the CD4-Cre is activated later than the Lck-Cre in DN (49), we crossed the Apc^fl allele with the Lck-Cre. We also found largely normal thymocyte development and severe lymphopenia when the Lck-Cre was used. Although subtle differences exist between Lck-Cre– and CD4-Cre–induced cKO, these differences are not sufficient to account for the dramatic discrepancy between our study and the earlier work (34). One explanation could be due to the use of different Apc^fl alleles rather than different Cre drivers. Finally, because activation of T cells in the periphery can cause loss of thymocytes (35), the possibility that activation of T cells in the periphery may account for a part of the previously described phenotype (34) remains to be excluded.

Regardless of how the differences are reconciled, the lack of significant impact of T cell development allows us to reveal a critical role for regulated Wnt signaling in T cell survival. Given the significant role of lymphopenia in autoimmune diseases (50, 51), the missing link between T cell survival and the Wnt–cMyc pathway may be important for our understanding of the pathogenesis of autoimmune diseases.

We thank Drs. Eric Fearon, Yuan Zhu, Philip King, and Beth Moore for helpful discussions throughout the study and Dr. Kaoru Sakabe for critical reading of the manuscript. Part of the study was performed at the University of Michigan, where the authors were previously employed.

The authors have no financial conflicts of interest.