Bile acids induce neurite outgrowth comparable to RA
We conducted experiments using mouse NSC-34 cells, considered to be a highly stable and widely used murine motor neuron cell line model [12, 13], and SH-SY5Y cells, a human neuronal cell model [14] in which retinoic acid is used to induce neurite outgrowth and functional characteristics of motor neurons. Neurite outgrowth was measured after exposure to basal differentiation medium (DM), 1 uM retinoic acid (RA), and 20 uM of the bile acid deoxycholic acid (DCA) for 6 days. NeuronCyto II, a Matlab-based software package that was developed to address the “crossover” issues in quantifying neuronal morphology [16, 17] was used to measure neurite outgrowth. We found that 20 uM deoxycholic acid (DCA) could induce neurite outgrowth in NSC-34 cells that was comparable to the neurotrophic effects of 1 uM retinoic acid (RA) (Fig. 1A). Similar, but slightly less robust effects, were observed for CDCA at 20 uM (Fig. 2D and Fig. 3). In SH-SY5Y cells, 1 uM RA and 20 uM DCA induced a less robust neurite induction, although optimum culture conditions and seeding densities were not established (Fig. 1B).
Fxr Or Lxr Stimulation Much Less Neurotrophic Than Ra
To determine if a known bile acid receptor was part of the mechanism of stimulated neurite outgrowth, we then used several available chemical agonists and antagonists of FXR, LXR, and TGR5 to determine whether they could substitute for, or block, the effects of bile acids (Figs. 2 and 3). In addition to 1 uM RA, which served as the reference control, 20 uM DCA and 20 uM CDCA were used. The commonly used FXR agonist GW 4064 [18] at 1 uM and 10 uM, WAY 252623, a potent LXR agonist that has been reported to stimulate the proliferation of mouse neural progenitor cells [25] at 1 uM and 10 uM, the dual FXR/TGR5 ligand obeticholic acid (INT 747) [7] at 1 uM and 10 uM, and the TGR5 agonist RG-239 that is a derivative of the naturally occurring triterpenoid betulinic acid [20] at 1 uM and 10 uM were used. To block bile acid pathways, Z-Guggulsterone, an inhibitor of FXR transactivation [19] at 5 uM and LXRα antagonist GSK2033 [22] at 5 uM, were used. As described for Fig. 1, NSC-34 cells were incubated for 6 days following by imaging and processing of 40 fields per condition with NeuronCyto II.
RA at 1 uM (Figs. 2B and 3) had the expected effect of strongly inducing neurite outgrowth with a median neurite length of 25.3 µm in RA treated cells compared to 3.2 µm in DM treated control cells without the addition of RA (p < 0.001 using a two-tailed Mann Whitney test assuming a non-Gaussian distribution after post-hoc adjustment for multiple comparisons). DCA (Figs. 2C and 3) and CDCA (Figs. 2D and 3) at 20 uM both induced neurite outgrowth with DCA more potent (median length 38.4 µm) than CDCA (median length 20.5 µm). The median neurite length in DCA treated cells was significantly longer (P < 0.01 after post-hoc adjustment for multiple comparisons) than in RA treated cells, while CDCA median length was not different than in RA treated cells (20.5 µm vs. 25.3 µm). DCA thus appears to have more robust neurotrophic effects than RA.
The median length in cells treated with 1 uM GW4064 (Figs. 2E and 3), an FXR agonist, was 10.9 µm, significantly shorter than RA (P < 0.001 after post-hoc adjustment for multiple comparisons), while 10 uM was toxic and resulted in cell death (Fig. 2F). The LXR activator, WAY 252623, induced mean neurite lengths of 10.3 µm and 8.4 µm at 1 uM and 10 uM concentrations (Figs. 2G, 2H, and 3), respectively, also significantly lower than RA (P < 0.001 after post-hoc adjustment for multiple comparisons). FXR or LXR stimulation was much less neurotrophic than RA, DCA, or CDCA.
Tgr5 Agonism Is More Neurotrophic Than Ra
Neurite length induced by 1 uM of the FXR/TGR5 dual ligand obeticholic acid (INT 747) was 15.6 µm (Figs. 2I and 3), significantly lower than induced by RA (P < 0.001 after post-hoc adjustment for multiple comparisons), but at 10 uM was 22.3 µm (Figs. 2J and 3)and not statistically different than induced by RA. The dual specificity INT 747 is a much stronger ligand for FXR, thus the dose response may be due to increasing TGR5 agonist activity. Consistent with these results, the TGR5 agonist RG-239 at 1 uM induced a neurite length of 9.5 µm (Figs. 2K and 3), significantly lower than induced by RA (P < 0.001 after post-hoc adjustment for multiple comparisons), but at 10 uM the median neurite length was 37.0 µm (Figs. 2L and 3), almost identical to that induced by DCA. TGR5 agonism thus appeared to be comparable to DCA stimulation and stronger than RA.
Neither Fxr Or Lxr Inhibition Blocks Bile Acid-induced Neurite Growth
FXR and LXR antagonists were then tested to determine if they could block the effects of DCA. Neither 5 uM of FXR antagonist Z-Guggulsterone (Fig. 2M) nor 5 uM of pan-LXR antagonist GSK2033 (Figs. 2O) had any effect on median neurite length by themselves, nor had any effect on DCA induced (Fig. 2N and 2P) median neurite length (38.2 µm with Z-Guggulsterone and 42.0 µm with GSK2033 vs. 38.4 µm without either). Neither FXR nor LXRα antagonism thus could block bile acid induced neurite outgrowth. These data suggest that TGR5 is a mechanism through which bile acids mediate their neurotrophic effects in NSC-34 cells.
Rna-seq Identified Bile Acid Specific Genes
We sought to characterize the early initiating molecular events in neurite induction by bile acids using RNA-seq analysis [24] of NSC-34 cells. To identify early effects on gene expression induced by either RA or the bile acids DCA and CDCA, RNA was isolated from NSC-34 cells treated for 48 hours, a time frame before significant morphological effects can be seen and presumably when early transcriptional responses are occurring to enable subsequent neurite outgrowth. Principal component analysis grouped DM (control) separately from DCA and CDCA (Figure S1) and log2 fold changes for transcripts plotted versus the mean of normalized counts over all genes showed more dispersion with CDCA than with RA or DCA (Figure S2).
Treatment with 1 uM RA identified 112 genes (Table S1) whose change in expression level relative to control basal neurite induction media lacking RA met a nominal (P < 0.05) level of statistical significance. Three genes achieved a statistically significant adjusted p-value threshold, including Cyp26b1, which is the form of CYP26 that is expressed in brain and for which RA is not only a specific substrate but also a potent inducer of expression [26]. RA is also known to induce the other two genes, DKK3 [27] and CTSB [28].
Treatment with 20 uM DCA identified 44 genes whose change in expression level relative to control basal neurite induction media lacking DCA or RA met a nominal (P < 0.05) level of statistical significance (Table S2). Treatment with 20 uM CDCA identified 94 genes whose change in expression level relative to control basal neurite induction media lacking CDCA or RA met a nominal (P < 0.05) level of statistical significance (Table S3). Comparing these three datasets identified genes whose differential expression was similar in 2 or more of the treatment conditions. For RA and either DCA or CDCA, 9 genes were found, whose direction of differential expression was the same direction and the magnitude of differential expression highly similar (Table 1). For DCA and CDCA genes, 4 genes were found whose differential expression was in the same direction and magnitude highly similar (Table 2).
Table 1
Genes whose changes in expression were similar in RA and DCA or CDCA treated NCE34 cells.
LIGAND
|
GENE
|
log2 Fold Change
|
p-value
|
BASELINE*
|
2 DAYS **
|
RA
|
Ikbip
|
5.54
|
0.0023
|
0.18
|
8.36
|
DCA
|
Ikbip
|
4.92
|
0.0173
|
0.18
|
5.40
|
CDCA
|
Ikbip
|
5.13
|
0.0304
|
0.17
|
5.89
|
RA
|
Kif2c
|
-3.93
|
0.0278
|
7.80
|
0.51
|
DCA
|
Kif2c
|
-3.98
|
0.0239
|
7.96
|
0.50
|
RA
|
Lgals1
|
-1.46
|
0.0029
|
36.16
|
13.17
|
DCA
|
Lgals1
|
-0.93
|
0.0471
|
36.93
|
19.34
|
RA
|
Mvb12a
|
2.61
|
0.0331
|
1.50
|
9.13
|
DCA
|
Mvb12a
|
2.59
|
0.0299
|
1.54
|
9.31
|
RA
|
Sdf4
|
2.38
|
0.0058
|
4.52
|
23.53
|
DCA
|
Sdf4
|
2.78
|
0.0001
|
3.70
|
25.37
|
RA
|
Ubald2
|
-6.08
|
0.0004
|
12.59
|
0.19
|
DCA
|
Ubald2
|
-6.13
|
0.0004
|
12.85
|
0.18
|
RA
|
Adnp
|
5.97
|
0.00061
|
0.18
|
11.12
|
DCA
|
Adnp
|
5.61
|
0.0015
|
0.18
|
8.81
|
RA
|
Marcksl1
|
1.25
|
0.0161
|
12.16
|
28.97
|
CDCA
|
Marcksl1
|
1.65
|
0.0437
|
11.19
|
35.11
|
RA
|
Mki67
|
-0.90
|
0.0137
|
59.13
|
31.76
|
CDCA
|
Mki67
|
-1.80
|
0.0125
|
56.19
|
16.09
|
* Expression of the control group back-calculated from baseMean and log2FoldChange. |
** Expression of the comparison group back–calculated from baseMean and log2FoldChange. |
Table 2
Genes whose changes in expression were similar in DCA and CDCA treated NCE34 cells.
LIGAND
|
GENE
|
log2 Fold Change
|
p-value
|
BASELINE*
|
2 DAYS**
|
DCA
|
Ikbip
|
4.92
|
0.0173
|
0.18
|
5.40
|
CDCA
|
Ikbip
|
5.13
|
0.0304
|
0.17
|
5.89
|
DCA
|
Gas2l3
|
-1.98
|
0.0344
|
12.07
|
3.06
|
CDCA
|
Gas2l3
|
-3.42
|
0.0349
|
11.72
|
1.09
|
DCA
|
Rps12
|
0.63
|
0.0462
|
59.65
|
92.62
|
CDCA
|
Rps12
|
1.66
|
0.0108
|
55.62
|
175.96
|
DCA
|
Tprgl
|
3.82
|
0.0354
|
0.50
|
7.00
|
CDCA
|
Tprgl
|
4.39
|
0.0276
|
0.45
|
9.49
|
* Expression of the control group back-calculated from baseMean and log2FoldChange. |
** Expression of the comparison group back–calculated from baseMean and log2FoldChange. |