Construction of rAAV vectors expressing human HGF
In this study, we tested whether an HGF-expressing rAAV vector could facilitate the regeneration of motor neurons and alleviate disease progression of ALS when delivered to the subarachnoid space through IT injection. AAV used in this study was designed to express two isoforms of human HGF—HGF723 (or dHGF) and HGF728 (or cHGF)—as in the case of our bodies [6]. In order to simultaneously express two isoforms, we used the gDNA-cDNA-hybrid sequence similar to that described by Cho et al. (Additional file 1: Figure S1a). This chimeric sequence was inserted into pAAV-MCS containing multiple cloning sites, resulting in pAAV-HGF.
To produce rAAV vectors, equal amounts of three plasmids (pAAV-HGF, rep/cap-expressing plasmid, and helper plasmid) were co-transfected into 1 × 106 HEK293T cells, initially producing AAV serotype 2 [4]. Seventy-two hours later, the supernatant was collected and viral titer was determined as described previously [29]. To determine whether rAAV vectors could indeed produce RNAs of two HGF isoforms, 1.6 × 105 C2C12 cells were transduced with 1.23 × 108 GC of rAAV, and total RNAs were isolated 48 h later followed by RT-PCR. A control virus—rAAV2-C lacking the HGF sequence—did not produce any HGF RNA, whereas two HGF RNA species, each for HGF723 and HGF728, were readily detectable in cells transduced with rAAV2-HGF (Additional file 1: Figure S1b).
To test whether HGF proteins were indeed produced from rAAV vectors, 8 × 104 C2C12 cells were transduced with 5 × 1013 GC of rAAV, and total proteins were isolated 48 h later followed by ELISA specific to human HGF proteins. rAAV2-HGF produced 159.13 ± 23.92 ng/mg of HGF, while no HGF was detectable in cells transduced with a control vector. (Additional file 1: Figure S1c).
Comparison of four AAV serotypes
Different subtypes of AAV have been shown to possess different tropism. To select an appropriate AAV serotype that could be used for gene transfer to the spinal cord, four serotypes of AAV (rAAV-1, − 2, − 5, and − 6) were prepared as described above, and 4.12 × 108 GC of each AAV were injected into the LSC. Four weeks later, total proteins were extracted from the LSC, and the expression level of HGF was compared using ELISA. As summarized in Fig. 1a, mice injected with rAAV1-HGF produced the highest level of HGF protein (621.89 ± 112.98 pg/mg), and other serotypes generated 7- to 17-fold lower amounts, 86.25 ± 7.59 pg/mg for rAAV6, 35.54 ± 4.87 pg/mg for rAAV2, and 80.15 ± 17.94 pg/mg for rAAV5 (Fig. 1a).
To determine the tissue distribution of transgene expression from different AAV vectors inside the LSC, different subtypes of AAV expressing GFP were used because it was technically difficult to analyze human HGF by IHC. Four weeks after IT injection of AAV vectors, tissue samples were taken and GFP-expressing cells were measured by IHC. rAAV1-GFP transduced the largest area of the ventral horn, while rAAV2 or rAAV6 expressed GFP primarily on the surface of white matter (Fig. 1b). Taken together, AAV1 appeared to be the most effective serotype for IT gene delivery, and thus was used for further experiments.
Kinetics of HGF expression from rAAV1-HGF
To determine how the HGF expression level changed over time, 5 × 109 GC of rAAV1-HGF were injected into the LSC of 2-month-old C57BL/6 mice, and total proteins were isolated, followed by ELISA. One week after injection, 1.98 ± 0.95 ng/mg of hHGF were detectable, and the level was the highest after two weeks (2.44 ± 1.37 ng/mg), thereafter gradually decreasing over time until 16 weeks after injection (0.95 ± 0.58 ng/mg) (Fig. 1c). A control virus lacking the HGF sequence did not produce any HGF protein. While hHGF was not detectable in the motor cortex and serum, 170.17 ± 155.95 pg/mg were observed in the TA (Additional file 1: Figure S1d).
To investigate whether the HGF proteins expressed from rAAV1-HGF could increase the phosphorylation of Met, a receptor for HGF, 5 × 109 GC of rAAV1-HGF were intrathecally injected into C57BL/6 mice at P60, Met phosphorylation levels were measured 7 days later using IHC. As shown in Fig. 1d, SMNs were labeled with NeuN (red) and ChAT (green). In mice injected with rAAV1-C, 37.32 ± 0.02% of SMNs were co-stained with p-Met. The number of p-Met-positive SMNs was increased to 58.09 ± 0.01% in the rAAV1-HGF group (Fig. 2e). These results indicated that HGF proteins expressed from rAAV1-HGF could indeed augment the phosphorylation of Met.
Effects of rAAV1-HGF on the sciatic nerve crush model
We first used the sciatic nerve crush mouse model to quickly check the effects of intrathecally delivered rAAV1-HGF on nerve damage in general [53]. Crush injury was introduced in 2-month-old C57BL/6 mice, and 5 × 109 GC of virus were intrathecally injected immediately. As shown in Fig. 2a-b, rAAV1-HGF significantly improved hindlimb strength and rotarod performance from 10 days after nerve crush, and these effects were maintained until the end of the experiment on day 28. These results suggested that rAAV1-HGF could promote functional recovery after sciatic nerve crush when delivered into the spinal cord.
It is well established that when the sciatic nerve is injured, Wallerian degeneration occurs and axons regenerate again as time goes by [42]. Therefore, we used IHC to examine whether rAAV1-HGF had effects on nerve regeneration in the areas of the nerve damage. Five days after nerve injury, the sciatic nerve was collected, and labeled with SCG10, a marker for the regenerating axon [43]. When the length of the regenerated sciatic nerve was measured, rAAV1-C showed a 2.11 ± 0.09 mm increase from the damaged site, while in the rAAV1-HGF group, it was 37.91% higher at 2.91 ± 0.27 mm (Fig. 2c-d).
The presynaptic terminal of the sciatic nerve forms NMJs with the postsynaptic end plate of the muscle to transmit the contractile signal to the muscle [53]. Therefore, TA muscle connected to the sciatic nerve was analyzed by IHC. In a normal state, the NMJ is pretzel-shaped and has well-preserved integrity. After nerve injury, however, the shape of the NMJ becomes abnormal and the degree of integrity decreases (Fig. 2e-f). Compared with the sham group to which the nerve injury was not introduced, the proportion of fully innervated NMJs in the rAAV1-C group was reduced 2.34-fold, and that of abnormally shaped NMJs was increased 7.75-fold. On the other hand, in the rAAV1-HGF group, the proportion of fully innervated NMJs increased by 69.7%, and that of denervated NMJs decreased 3.83-fold in comparison to the rAAV1-C group (Fig. 2g-h). The average α-BTX area was decreased by 41.75% in the rAAV1-C group compared to the sham group, but was increased by 56.76% in the rAAV1-HGF group. These results suggested that IT injection of rAAV1-HGF could promote the regeneration of neurons and re-establishment of NMJs possibly by acting on the cell body of motor neurons located in the spinal cord.
Effects of rAAV1-HGF in SOD1-G93A TG mouse model
Encouraged by the above results, the effects of rAAV1-HGF were also tested in the SOD1-G93A TG mouse model, the most commonly used animal model for ALS. First, we investigated whether HGF proteins expressed from rAAV1-HGF could also increase the phosphorylation of Met in SOD1-G93A TG mice. 5 × 109 GC of rAAV1-HGF were intrathecally injected to SOD1-G93A TG mice at P60, and the levels of Met phosphorylation were measured 40 days later using IHC (Additional file 1: Figure S2a). In non-TG mice, 35.24 ± 0.55% of SMNs were co-stained with p-Met. The number of p-Met-positive cells was lowered in TG mice injected with a control vector, rAAV1-C, to 29.98 ± 0.54%, but increased to 59.65 ± 1.35% in the rAAV1-HGF group (Additional file 1: Figure S2b). These findings suggested that the HGF proteins expressed from rAAV1-HGF could indeed increase the phosphorylation of Met in this mouse model.
To test whether rAAV1-HGF could exert any effect on disease progression of SOD1-G93A TG mice, body weight was measured and behavioral tests were performed once every week until P158 after injecting rAAV1-HGF into the LSC of TG mice at P60. IT delivery of rAAV1-HGF did not affect the rate of weight loss (Fig. 3e), but significant improvements were observed in rotarod, hanging wire, and grip strength tests (Fig. 3a-d). In addition, survival rate was increased from 20 to 66.67% at P150, and median survival was also increased by 6.53% from 145.5 days to 155 days (Fig. 3f). These results suggested that IT delivery of rAAV1-HGF might slow disease progression and improve survival rate in this mouse model.
Effects of rAAV1-HGF on SMNs and NMJs in the SOD1-G93A TG mouse model
We tested whether IT injection of rAAV1-HGF could delay degeneration of SMNs in SOD1-G93A TG mice. 5 × 109 GC of rAAV1-HGF were injected into the LSCs of SOD1-G93A TG mice at P60, and the LSCs were collected 40 days later (P100) followed by IHC. When compared to the non-TG group, the rAAV1-C group showed the number of SMNs in the ventral horn reduced 2.07-fold, the ratio of SMNs to total cells decreased by 58.65%, and the diameter of SMNs diminished by 39.83% (Fig. 4a). Although rAAV1-HGF did not increase the diameter of SMNs, the number of SMNs was increased 2.36-fold and the ratio of SMNs to total cells was increased by 67.11%, compared to the rAAV1-C (Fig. 4b-d).
We also examined whether changes in SMNs by the rAAV1-HGF could lead to improvements in connected axon terminals and muscles. Similar to the experiments described above (Fig. 2e-h), the NMJ was observed and the weight of TA was measured. Compared to the non-TG group, in the rAAV1-C group, the proportion of fully innervated NMJs was decreased by 25.48%, the level of abnormal-shaped NMJs was increased 3.5-fold, and the weight of TA was decreased by 35.42%. When rAAV1-HGF was injected, however, the proportion of pretzel-shaped NMJs was increased by 32.58% (Fig. 4e-f), and the weight of TA, which was decreased by muscular atrophy in the rAAV1-C group, was increased by 38.42% (Fig. 4g). No visible change was observed in the degree of NMJ integrity in the rAAV1-HGF group. These results indicated that IT delivery of rAAV1-HGF could promote the protection of peripheral nerves and NMJs as well as SMNs.
Effects of rHGF on axonal outgrowth of CSMNs
The above data showed that IT delivery of rAAV1-HGF could improve motor functions and survival rates of SOD1-G93A TG mice, presumably by slowing down the degeneration of SMNs and restoring the morphology of NMJs. Since the interaction of HGF with lower motor neurons has been relatively well established compared to the interaction of HGF with upper motor neurons [8, 54, 56], we were interested in testing the possible involvement of upper motor neurons by using the motor cortical cultures system consisting of CSMNs and glial cells [18, 35].
The motor cortices were isolated from P3 non-TG or TG mice, and 2 × 104 cells were cultured on a 24-well plate. Three days later, cells were fixed, followed by IHC analysis. CSMNs were labeled with UCHL1 and Ctip2, and Fiji software was used to measure the axon length of CSMNs [58] (Fig. 5a).
Since CSMNs have previously been shown to produce HGF which may give experimental noise, the effects of endogenously expressed HGF were first tested using PHA665752, a chemical inhibitor of Met. When cells were treated with 1 μM of PHA665752 for 3 days, the axon length of CSMNs was reduced 3.42-fold in the non-TG group and 2.33-fold in the TG group (Fig. 5b). This result indicated that endogenously expressed HGF could positively affect the axonal outgrowth of CSMNs.
When 100 ng/ml of rHGF were added to the culture for 3 days, however, the axon length of CSMNs was increased by 48.61% in the non-TG group, and by 124.73% in the TG group (Fig. 5c). When PHA665752 was co-treated with rHGF, the HGF-mediated increase of CSMNs’ axon length was inhibited by 127.7% in the non-TG group, and by 120.82% in the TG group (Fig. 5d). This is consistent with the data in Fig. 7a-b showing activation or inhibition of phosphorylation of the Met protein by HGF or PHA665752. These results suggested that activation of the HGF-Met signaling pathway might be involved in the promotion of axonal outgrowth of CSMNs.
Effects of inhibition of ERK, PI3K, and p38 on axonal outgrowth of CSMNs
Several signaling pathways have been shown to be turned on upon interaction between HGF and Met receptor. To identify the key pathway involved in the regulation of the HGF-mediated axonal outgrowth of CSMNs, 2 × 104 cells were treated with specific chemical inhibitors for ERK (U0126), PI3K (LY294002), p38 (SB203580), and JNK (SP600125). When cells were treated with 3 different concentrations of respective inhibitors, the axon length of CSMNs was reduced in a dose-dependent manner by all agents except for SP600125 (Fig. 6a-d). In all concentrations of chemical inhibitors used in this experiment, cytotoxic effects were not observed (Additional file 1: Figure S3a). These results suggested that ERK, PI3K, and p38 might play roles in the HGF-mediated axonal outgrowth of CSMNs.
Effects of rHGF on ERK phosphorylation
We tested whether HGF could activate ERK, PI3K, and p38 in the cortical culture. 3.2 × 106 cells were treated with 100 ng/ml of rHGF, and 5 days later, total proteins were extracted, followed by Western blot analysis. As shown in Fig. 7a, the level of phosphorylated ERK was increased by treatment with rHGF, while the levels of PI3K and p38 were not affected. The effect of rHGF was inhibited when cells were co-treated with 1 μM of PHA665752 or 10 μM of U0126 (Fig. 7b-c), suggesting that the HGF-Met pathway is involved in the phosphorylation of ERK in this cortical culture. We tested whether inhibition of ERK could reduce the HGF-mediated increase of the axon length of CSMNs. As shown in Fig. 7d, treatment with U0126 decreased the axon length of CSMNs 2.41-fold in the non-TG group and 2.03-fold in the TG group, indicating that HGF could promote axonal outgrowth of CSMNs by specifically up-regulating phosphorylation of ERK.
To test whether the above in vitro results were reproducible in vivo, 5 × 109 GC of viral vectors were intrathecally injected into SOD1-G93A TG mice at P60, and the LSC was analyzed for ERK. Compared with non-TG mice, the level of phosphorylated ERK was increased 2.04-fold in TG mice injected with a control vector, and was further enhanced 5.54-fold when injected with AAV vectors expressing HGF (Fig. 7e-f). There was no difference in phosphorylation of other signaling molecules of the HGF-Met pathway, such as STAT3, cJUN, and GSK3β (Additional file 1: Figure S4a). These results indicated that ERK might indeed be an important factor of the HGF-mediated regeneration of motor neurons.
Effects of HGF-mediated ERK phosphorylation on levels of ROS
To get an overall picture of the effects of HGF on gene expression profile in CSMNs, cortical cells were treated with 100 ng/ml of rHGF, and total RNAs were extracted followed by microarray assay (Additional file 1: Figure S5a). Among the 116 genes whose expression levels were changed more than 1.5-fold, 52 belonged to six major categories defined for fALS by Taylor et al., and 37 (71.15%) were involved in the control of protein quality and RNA metabolism [49]. Therefore, we tested whether the HGF-Met-ERK signaling pathway could reduce hSOD1 protein aggregation and/or oxidative stress. Cortical cells were treated with rHGF, followed by IHC to examine the distribution of mutant hSOD1. When compared to the non-TG group containing no hSOD1 aggregates, the proportion of CSMNs in the TG group with hSOD1 aggregates was sharply increased to 0.4 ± 0.01 as shown in Fig. 8a-b. When treated with rHGF, however, this proportion was reduced to 0.06 ± 0.02, while it was increased to 0.3 ± 0.06 and 0.33 ± 0.04 by the addition of inhibitors for Met (PHA665752) or ERK (U0126), respectively (Fig. 8a).
Since the aggregated form of mutant hSOD1 could increase oxidative stress, it was also tested whether the rHGF-mediated reduction of protein aggregation could alleviate oxidative stress by measuring hydrogen peroxide, peroxynitrite, hydroxyl radicals, nitric oxide, and peroxy radical in the cortical culture. Compared to the control group treated with SFM, treatment with rHGF decreased the level of oxidative stress by 34.22%, which was comparable to the effect (32.11%) of N-acetyl-L-cysteine (NAC), a well-known scavenger of oxygen free radical (Fig. 8c). Such HGF-mediated decrease of oxidative stress was inhibited by 15.51 and 18.81% when rHGF was co-treated with PHA665752 and U0126, respectively (Fig. 8c).
To test whether the axonal outgrowth of CSMNs could be promoted by alleviation of oxidative stress, the effect of rHGF was tested in the presence of pyocyanin, a ROS inducer. As shown in Fig. 8d, the axon length of CSMNs was increased by 40.12% when treated with NAC. The effect of rHGF was almost 3-fold greater at 118.83%, which was inhibited by 32.1% in the presence of pyocyanin. Taken together, these results indicated that HGF-mediated induction of ERK phosphorylation plays an important role(s) in promoting the axonal outgrowth of CSMNs by mitigating protein aggregation and oxidative stress.