Reversal of Streptozotocin-Induced Diabetes in Mice by Cellular Transduction With Recombinant Pancreatic Transcription Factor Pancreatic Duodenal Homeobox-1 : A Novel Protein Transduction Domain–Based Therapy

To express recombinant rat Pdx1-His₆ (hereafter designated rPdx1), full-length rat PDX1 cDNA was amplified by PCR and subcloned using NdeI and XhoI sites in pET28b (Novagen, Madison, WI). To express PTD–green fluorescent protein (PTD-GFP)-His₆ (hereafter designated PTD-GFP), the PTD-GFP plasmid containing the coding sequence for the 11-residue (YGRKKRRQRRR) PTD of HIV-1 TAT positioned at the NH₂-terminus of green fluorescent protein was constructed by PCR and cloned into a pT7/CT-TOPO expression plasmid (Invitrogen, Carlsbad, CA). To prepare the 16-residue PTD-deletion mutant of rPdx1-His₆ (rPdx1-Δmut), we constructed a mutant rat PDX1 expression plasmid missing PTD residues 188–203 (RHIKIWFQNRRMKWKK) within the rPdx1 homeodomain using PCR amplification with appropriate primers containing sequences before and after the PTD of PDX1 cDNA. The two PCR products were subsequently ligated to generate the PTD-deletion mutant. After confirmation of the cDNA sequence, the resulting mutant PDX1 cDNA subcloned into the NdeI and XhoI sites of pET28b (Novagen). After growth at 37°C to an optical density (OD₆₀₀) of 0.8, plasmid-containing BL21 (DE3) cells were incubated at room temperature for another 18 h in the presence of 0.5 mmol/l (final concentration) isopropyl β-d-1-thiogalactopyranoside. Bacteria were lysed by pulse sonication in buffer A (20 mmol/l Tris/HCI pH 8.0, 500 mmol/l NaCl, and 0.1% Triton X-100) containing 5 mmol/l imidazole and proteinase inhibitors (Roche Diagnostics, Basel, Switzerland). After centrifugation, the cell-free supernatant was applied to a column of Ni-nitrilotriacetate agarose (Invitrogen) and washed with several volumes of buffer A containing 25 mmol/l imidazole. The protein was eluted by buffer A containing 250 mmol/l imidazole. The purity of the eluted rPdx1, PTD-GFP, and rPdx1-Δmut fusion proteins were characterized by SDS-PAGE/Coomassie Blue staining following dialysis against PBS.

We constructed, produced, and titered lentiviral vectors (LV) containing the mouse PDX1 cDNA as previously described (27). An LV containing 950 bp (−940 to 10) of human NeuroD/Beta2 promoter (28) cloned by PCR and ligated into GFP was prepared.

Rat liver epithelial stem cells (WB cells) (29,30) at 70% confluence were treated with purified rPdx1 or rPdx1-Δmut at a 1-μm concentration for various time points, and the cells were washed three times with PBS, harvested in lysis buffer (150 mmol/l NaCl, 50 mmol/l Tris-HCl, pH 7.5, 500 μmol/l EDTA, 1.0% Triton X-100, and 1% sodium deoxycholate) containing a protease inhibitor cocktail (Roche Diagnostics). For Western blotting (27,31), antibodies against rPdx1 (rabbit serum, 1:1,000 dilution, made in our lab using purified rPdx1 as immunogen), his-tag (1:2000; Invitrogen), or actin (1:2000; DAKO, Carpinteria, CA) were used to detect rPdx1, rPdx1-Δmut, actin, or PTD-GFP fusion protein in cell lysates (50 μg total protein/lane) after SDS/PAGE separation.

WB cells (70% confluence) were first transduced by LV-pNeuroD-GFP at a multiplicity of infection of 20 as previously described (27). Transduced WB cells were then incubated in the absence or presence of 1 μm rPdx1 protein. pNeuroD-GFP–containing WB cells were transduced with LV-Pdx1, which served as a positive control for cell-made Pdx1 protein. Cells were harvested 72 h posttreatment, fixed in 1% formaldehyde for 10 min, washed with 1% BSA in PBS, and resuspended. Cytospin slides were prepared with the treated cells and covered with mounting medium containing DAPI.

BALB/c mice (8–10 weeks-old, University of Florida, pathology mouse colony) were injected with 50 mg/kg body wt streptozotocin (Sigma-Aldrich, St. Louis, MO) for 5 consecutive days to induce diabetes (27,31). Animals with fasting blood glucose levels for two consecutive readings of >300 mg/dl received protein treatment. The diabetic mice were intraperitoneally injected with either rPdx1 or PTD-GFP protein (0.1 mg · day⁻¹ · mouse⁻¹) for 10 consecutive days. Fasting blood glucose levels were measured regularly using a glucometer after the mice were fasted for 6 h. The sequential experimental events of the animal studies are summarized in Fig. 3A.

The normal, rPdx1-, or PTD-GFP–treated mice (n ≥ 4 for each group) were fasted for 6 h following intraperitoneal injection with glucose (1 mg/g body weight), and blood glucose levels were measured at 5, 15, 30, 60, and 120 min postinjection. Subtotal pancreatectomy (>90%) (n = 4) was performed at day 30 posttreatment under general anesthesia. The mice were killed at two time points (days 14 or 40 posttreatment), and organs and blood were collected for evaluation of histology, gene expression, serum, and tissue insulin levels.

Mice were injected intraperitoneally with rPdx1 (0.1 or 1 mg). Blood samples were drawn at 0.25, 0.5, 1, 2, 6, and 24 h. Tissues from normal and rPdx1-treated mice were harvested at 1 or 24 h, fixed in 10% formalin, and embedded in paraffin for Pdx1 immunohistochemistry using rabbit anti-Pdx1 antibody (1:3,000; a gift of Dr. Christopher V. Wright, Vanderbilt University). The levels of rPdx1 in sera were determined by immunoblotting as described above.

Total RNA was prepared from mouse tissues of the pancreas and liver using TRIZOL reagent and cDNA synthesized using Superscript II reverse transcriptase (Invitrogen) with random hexamer primers. Liver gene expression was determined by RT-PCR as previously described (31). The forward and reverse PCR primers were designed to be intron spanning, and their sequences and conditions are listed in Table 1. To eliminate false positives, no reverse transcriptase, positive, or blank controls were included. All data represent at least three measurements from at least four mice.

cDNA from pancreatic and liver tissues was subjected to three independent PCR reactions. Each reaction was performed in duplicate or triplicate in a Thermocycler Sequence detection system (DNA engine opticon 2; MJ Research, Springfield, MO) using SYBR green (Qiagen, Valencia, CA). Total pancreatic and liver RNA was used as templates for preparing cDNA. Tissues were snap-frozen in liquid nitrogen and RNA was extracted in RNase free tubes using Trizol reagent (Invitrogen). The integrity and stability of the RNAs from both pancreas and liver were confirmed by demonstrating the intact 28s and 18s bands on gel electrophoresis. cDNA was synthesized from 5 μg total RNA using Superscript reverse transcriptase enzyme (Invitrogen) with random hexamer primers according to the manufacturer's protocol. Amplication of the correct product was confirmed by gel electrophoresis. Conditions for real-time PCR were as follows: after initial denaturation at 95°C for 15 min to activate the enzyme, 38 cycles of PCR (denaturation 0.5 min at 94°C, annealing 0.5 min at 61°C [for liver] or 56°C [for pancreas], and elongation 0.5 min at 72°C with a final extension 5 min at 72°C) were carried out. Each gene was tested three times (3–4 mice/group). Relative gene expression in mouse pancreas treated by rPdx-1 and PTD-GFP on day 14 was calculated by 2^−ΔΔCt method. We used mouse β-actin as internal control and PTD-GFP treated day-14 pancreas cDNA as calibrator. First, both threshold cycle (C_T) values of target gene from rPdx1-treated day-14 pancreas cDNA (n = 3) and PTD-GFP–treated day-14 pancreas cDNA (n = 4) were normalized by C_T value of the internal control. Then, the former was subtracted by the latter, namely ΔΔC_{T rPdx1/GFP} = (C_{T, Target} − C_{T, Actin})_rPdx1 − (C_{T , Target} − C_{T , Actin})_GFP. Similarly, ΔΔC_T between rPdx1-treated day-14 liver cDNA (n = 3) and rPDX1-treated day-40 liver cDNA (n = 4, calibrator) was also calculated by ΔΔC_{T day 14/day 40} = (C_{T , Target} − C_{T, Actin})_{day 14} − (C_{T, Target} − C_{T, Actin})_{day 40}. The value of 2^−ΔΔCT, the fold change of gene expression, was plotted into the figure. The primers were designed in accordance to the real-time PCR conditions, and the sequences are listed in Table 2.

Paraffin blocks containing the pancreas and liver were obtained from at least 5 mice/group. Antigen retrieval from paraffin sections was done in trilogy solution (Cell Marque, Rocklin, CA) at 95°C for 30 min for unmasking nuclear antigens such as Pdx1 and Ki-67. Paraffin sections (5 μm) were incubated with anti-swine insulin (1:1,000), rabbit anti-Pdx1 (1:5,000), rabbit anti-human Ki67 (1:100; Novus Biologicals, Littleton, CO), and goat anti-glucagon (1:200) antibodies, followed by incubation with anti-mouse or rabbit IgG (1:5,000) conjugated with HRP and detection with DAB substrate kit (Vector Laboratories, Burlingame, CA) as previously described (31). For double immunofluorescence, tissue sections were incubated with anti-swine insulin (1:200) and goat anti-glucagon (1:150; Santa Cruz Biotechnology, Santa Cruz, CA) primary antibody overnight at 4°C, followed by donkey anti–guinea pig IgG conjugated with fluorescein isothiocyanate (1:1,000; RDI Research Diagnostics, Concord, MA) and donkey anti-goat IgG with Alexa flour 594 flourochrome (1:500; Invitrogen). For Ki67/insulin sequential immunostaining, the paraffin sections from pancreas were first immunostained for Ki67 nuclear antigen and developed in brown color using the HRP/DAB system. After being counterstained with hematoxylin to highlight cell nuclei (blue), the slides were immunostained for insulin using guinea pig anti-porcine insulin (cross-react with mouse insulin) antibody (DAKO) at a dilution of 1:200 for 16 min. Presence of insulin in pancreatic islet β-cells was visualized in red color with the Ventana Ultra View red detection kit (Ventana Medical System, Tucson, AR).

Whole pancreas and liver organs were obtained from at least five mice/group. They were harvested, weighed, and immediately placed in acid-ethanol solution (180 mmol/l HCl in 70% ethanol) on ice with a corresponding tissue volume (1 ml buffer/0.1 g liver or 0.05 g pancreas) in accordance with a previously published procedure, with minor modifications (32). Tissue insulin levels were measured using an ultrasensitive mouse insulin enzyme-linked immunosorbent assay (ELISA) kit (ALPCO, Diagnostics, Salem NH). Absorbance was measured using a BIO-RAD 3550-UV microplate reader, with final results converted to nanograms insulin/milligrams pancreas tissue or nanograms insulin/grams liver. For measurement of serum insulin, both normal and treated mice were first fasted for 6 h, and blood samples were collected at 15-min intervals following intraperitoneal glucose (1 mg/g body wt) stimulation. Serum insulin levels were determined by ELISA.

Statistical significance was analyzed using an independent sample t test, requiring a P value <0.05 for the data to be considered statistically significant.

We constructed expression plasmids containing rat PDX1, rat PDX1-Δmut, or PTD-GFP cDNA, each also containing an additional nucleotide sequence coding for a His₆ tag for rapid purification using Ni²⁺-nitrilotriacetate columns. To obtain nearly homogeneous proteins in sufficient amounts for our in vitro and in vivo animal studies, we first optimized bacterial expression conditions to yield 10 mg highly pure rPdx1 per liter of growth medium. Figure 1A shows the structural organization of rPdx1 (a), PTD-GFP (b), and rPDX1-Δmut (c) and a Coomassie blue-stained SDS gel for rPdx1, PTD-GFP, and rPdx1-Δmut fusion proteins. These proteins consistently had purity >90–95%, based on densitometry. We also confirmed the identity of rPdx1, rPdx1-Δmut, and PTD-GFP proteins by Western blotting with anti-Pdx1 or anti–histidine-tag antibody.

To confirm that rPdx1 protein possessed the ability to penetrate cells due to antennapedia-like PTD within the homeodomain of Pdx1, WB cells were incubated with rPdx1 or rPdx1-Δmut protein (1 μmol/l) for specified times, after which the cells were washed three times with PBS. Cell lysates were separated by SDS-PAGE and blotted with anti-Pdx1 and anti-actin antibodies. The relative amount of rPdx1 in the cell blots was quantified by densitometry and normalized relative to actin. As shown in Fig. 1Ba, rPdx1 protein entry commenced within 5 min. As rPdx1 incorporation proceeded, cellular rPdx1 protein level reached peak values at 1–2 h and began to fall by 6 h. In contrast, rPdx1-Δmut with deletion of the 16aa antennapedia-like PTD failed to enter the cells, even though the cell culture medium still contained high levels of rPdx1-Δmut protein (Fig. 1Bb).

Since this assay cannot distinguish genuine rPdx1 protein entry versus nonspecific binding to cell membrane surface, we determined whether internalized rPdx1 protein was biologically active. To test the transcriptional function of the rPdx1 protein, WB cells were first transduced with lentivirus-containing pNeuroD-GFP reporter gene, a direct downstream target gene of the Pdx1. The cells were then incubated in the presence or absence of rPdx1 (1 μmol/l) for 72 h. These cells were harvested for flow cytometry to evaluate the rPdx1-mediated NeuroD gene activity by determining the percentage of WB cells expressing pNeuroD-GFP. To compare the efficiency of externally administered rPdx1 vs. endogenous Pdx1, pNeuroD-GFP–containing WB cells were either transduced with LV-Pdx1 (serving as a positive control for endogenous Pdx1) or treated with rPdx1 protein for 72 h, and pNeuroD-GFP–expressing WB cells were determined by both fluorescence microscopy (Fig. 1Ca) and flow cytometry (Fig. 1Cb). A comparable transcription efficacy for LV-Pdx1–treated cells (21%) and rPdx1 protein–treated cells (19%) after 72 was observed. Both treatments showed statistically significant differences when compared with control cells containing pNeuroD-GFP vector only. These results clearly demonstrated that rPdx1 rapidly entered cells and efficiently activated its downstream NeuroD target gene. These findings also confirm that rPdx1 protein has the same or comparable transcriptional activity as endogenous Pdx1 protein produced by means of LV-Pdx1 transgene expression.

While the antennapedia-like PTD allows in vitro–administered rPdx1 to enter cells, the in vivo tissue distribution and kinetics of rPdx1 were unknown. To examine rPdx1 distribution and kinetics, BALB/c mice were intraperitoneally injected with 0.1 or 1 mg rPdx1 protein, blood samples were collected at various times, and rPdx1 was detected in sera by anti-Pdx1 immunoblotting. The rPdx1 became evident in sera as early as 1 h postinjection, reaching peak values at 2 h and then markedly falling by 6 h. No rPdx1 protein was detectable in the 24-h serum samples (Fig. 2A). Major organs were harvested at 1 or 24 h after intraperitoneal injection and probed with anti-Pdx1 antibody. As shown in the representative images of liver, pancreas, and kidney tissues (Fig. 2B), the rPdx1 was concentrated mainly in the nuclei of hepatocytes, and the greatest intensity of Pdx1-positive cells was found nearest the hepatic terminal veins. This distribution pattern is consistent with the predicted pathway for rapidly internalized rPdx1 via the portal vein system. The rPdx1 was also detected in peripheral acinar cells of the pancreas, possibly as a result of direct uptake or through local circulation along pancreatic terminal capillaries. In kidneys, rPdx1 was mostly concentrated in brush borders of proximal tubular cells. Little protein was seen in the cytoplasm and none in the nuclei. At 24 h postinjection, only faint rPdx1 protein was focally detected in liver, pancreas, and kidney tissue sections. A low level of rPdx1 was also detected in the tissues of the spleen, heart, lung, and brain at 1 h postinjection, but it became undetectable at 24 h postinjection (data not shown). Similar findings were also noted in mice treated with 1 mg rPdx1 protein. Control mice showed no detectable rPdx1 protein in any tissue section except for pancreatic islet β-cells. On the basis of these findings, we chose the 0.1-mg rPdx1 at 24 h intervals as our in vivo condition for assessing in vivo effects of rPdx1 on mice with streptozotocin-induced diabetes.

With this, we developed an experimental timeline for rPdx1 protein administration (Fig. 3A) in mice with streptozotocin-induced diabetes (27,31). Diabetic BALB/c mice (body weights ∼ 20 g) were intraperitoneally administered with 0.1 mg rPdx1 or nontherapeutic PTD-GFP protein (negative control) over 10 consecutive days. Fasting blood glucose levels were monitored as indicated. Mice receiving rPdx1 injections achieved near normoglycemia within 2 weeks of first injection (Fig. 3B); however, no amelioration of hyperglycemia was observed in control mice receiving the PTD-GFP.

At days 14 and 40 postinjection, intraperitoneal glucose tolerance test (IPGTT) (Fig. 3C) showed that the mice receiving rPdx1 injection, in contrast to those diabetic mice receiving PTD-GFP, exhibited a much improved, nearly normal IPGTT curve. To further assess the ability of glucose-stimulated insulin release in the rPdx1-treated diabetic mice, healthy normal and treated mice were challenged at days 14 and 40 post-rPdx1 or–PTD-GFP injection with intraperitoneal bolus of glucose. Sera collected from normal, rPdx1-treated, or PTD-GFP–treated mice at 15 min postglucose injection were assayed for insulin and glucose (Fig. 3D). Serum insulin levels in rPdx1-treated mice were 6.9 times higher at day 14 and 11.3 times higher at day 40 than those in the PTD-GFP–treated group (Fig. 3E), indicating a significant improvement in the ability of the rPdx1-treated mice to handle glucose challenge. Overall, the serum insulin levels were inversely related to the blood glucose levels on days 14 and 40 postinjection in mice treated with rPdx1 or PTD-GFP protein. Although near euglycemia was achieved at day 40 in rPdx1-treated mice, the released insulin (2.6 μg/l) following 15 min of glucose stimulation was still much lower than that in normal nondiabetic mice (5.7 μg/l), suggesting either functional immaturity of newly formed islet β-cells or suboptimal β-cell mass in the pancreas or IPCs in nonpancreatic tissues for glucose homeostasis.

To explore whether rPdx1-mediated normoglycemia resulted from islet β-cell regeneration, rPdx1-treated mice were subjected to subtotal (>90%) pancreatectomy at day 30 (n = 4) postinjection. Blood glucose levels rose sharply following subtotal pancreatectomy (Fig. 3B), indicating that pancreatic cells (presumably the islet β-cells) played a dominant role in restoring normoglycemia at 30 days post–rPdx1 injection. To further determine the role of the pancreas at earlier stages of blood glucose normalization, subtotal pancreatectomy was performed at day 12 postinjection in rPdx1-treated mice (n = 4) and resulted in elevated blood glucose levels: from an average of 170 to 350 mg/dl within 48 h postoperation (data not shown). These results raise the distinct possibility that in vivo delivery of rPdx1 promoted endogenous β-cell regeneration.

Immunohistochemical examination confirmed vigorous islet β-cell regeneration, with larger and more abundant islets evident in rPdx1-treated mouse pancreata, in contrast to rare and scattered small islets in PTD-GFP–treated mice (Fig. 4A). Additionally, individual scattered insulin-positive cells were also noted within pancreatic acinar cells (data not shown). Such findings indicate that when administered in vivo in multiple doses, rPdx1 appears to promote endogenous β-cell regeneration via an as yet undefined molecular/cellular mechanism. To further characterize the regenerated islets, we conducted double immunofluorescence studies with anti-glucagon and anti-insulin antibodies. Significantly, we noted a change in the α-cell–to–β-cell ratio and distribution patterns (Fig. 4B) consistent with a dynamic process of islet cell regeneration and maturation. To determine the proliferation rate in the newly regenerated islets following rPdx1 treatment, we sequentially performed Ki67, followed by insulin immunostaining, on pancreas sections from various groups of animals. As shown in Fig. 4C, representative micrographs indicate that Ki67-positive pancreatic islet cells from normal mice were rarely seen (0–1 Ki67 positive cell/islet [Fig. 4C1]). Although few proliferating pancreatic islet cells were observed in the PTD-GFP–treated mice, it was difficult to identify normal-sized islets except for scant scattered small islets (Fig. 4C2). However, Ki67-positive islet cells were markedly increased in the newly regenerated islets in the pancreata of the rPdx1-treated mice compared with those of control mice. Furthermore, a much higher proliferation index of the islet cells in pancreatic sections was noted in day-14 (Fig. 4C3) than day-40 (Fig. 4C4) rPdx1-treated mice. Sequential double Ki67/insulin immunostaining (Fig. 4C5–7a) showed that both insulin-positive islet β-cells and non–β-cells were proliferating in islets from tissues obtained at days 14 and 40 post–rPdx1 treatment. Interestingly, pancreatic exocrine acinar cells from rPdx1-treated mice also had increased Ki67-positive cells (data not shown).

To determine the molecular events underlying rPdx1-mediated islet β-cell regeneration, we used real-time RT-PCR to examine the expression of key genes relevant to pancreas regeneration. rPdx1 treatment of diabetic mice (Fig. 4D) at day 14 resulted in markedly upregulated levels of INS-I (21.3 times), PDX1 (3.8 times), INGAPrP (14.5 times), Reg3γ (6.8 times), and PAP (34.3 times) relative to their corresponding control values (PTD-GFP–treated pancreas). A similar pattern of upregulation of the aforementioned genes was observed at day 40 posttreatment. Interestingly, expression of PAP, although upregulated at day 14 when compared with that in PTD-GFP–treated mice, was noticeably reduced at day 40 posttreatment. These results indicate that rPdx1 protein can promote islet β-cell regeneration, possibly by upregulating genes involved in pancreatic cell regeneration.

We also assessed whether in vivo intraperiteonal rPdx1 delivery affects the liver by searching for the presence of IPCs in treated mice at days 14 and 40 postinjection. Representative liver images from mice treated with PTD-GFP or rPdx1 at day 14 posttreatment (Fig. 5A) indicate that most of the scattered insulin-staining–positive liver cells were distributed along the edges of hepatic terminal veins (H.T.V.), a pattern consistent with rPdx1 tissue distribution in the liver (see Fig. 2B). There were scattered individual insulin-positive hepatocytes (arrows) having small bilobed nuclei and dark condensed chromatin, hinting at a more mature cell pattern. No IPCs were observed in the control GFP-treated mouse liver.

We next investigated expression profile of pancreatic genes in rPdx1- and PTD-GFP–treated livers (Fig. 5B). As expected, the rPdx1-treated livers at day 14 posttreatment (lane 6) expressed many Pdx1-target pancreatic genes, including PDX1, INS-I, GLUC, ELAS, and IAPP. Upregulation of other pancreatic endocrine genes (INS-II, SOM, NeuroD, and ISL-I) and pancreatic exocrine genes (p48 and AMY), was noted in the livers of the rPdx1-treated mice compared with that in the PTD-GFP–treated mice. However, Ngn3 gene expression was not detectable at either time point. Interestingly, by day 40, INS-I, GLUC, ELAS, and IAPP gene expression became undetectable, whereas expression of the aforementioned pancreatic genes continued, albeit at reduced levels. To quantitatively compare the changes of the pancreatic genes in the livers between days 14 and 40, we performed real-time RT-PCR for selected mRNAs including INS-I, GLUC, PDX1, p48, AMY, and ELAS. As indicated in Fig. 5C, a several-fold increase was noted at day 14 vs. day 40 mice post–rPdx1 treatment.

Given the intrinsic ability of rPdx1 protein to penetrate cells indiscriminately, we also examined the tissue specificity of rPdx1's effect on expression of pancreatic PDX1, INS-I, GLUC, and AMY genes by RT-PCR. We found little or no pancreatic gene activation at day 14 post–rPdx1 treatment in kidney, brain, heart, lung, small bowel, or spleen (Fig. 5D). These results suggest that rPdx1 selectively promotes liver expression of pancreatic genes and pancreatic β-cell regeneration without detectable evidence of undesired expression in other tissues.

To determine the relative contribution of pancreas and liver tissue–derived insulin to ameliorating blood glucose levels following rPdx1 treatment, pancreas and liver tissue insulin content of both normal and treated mice was measured at day 14 or 40 by ELISA. Pancreatic insulin content in rPdx1-treated diabetic mice at days 14 and 40 postinjection was ∼44 and 68%, respectively, of the normal pancreatic levels (Fig. 6A). These levels were also 6.7 times higher at day 14 (P < 0.01, Student's t test) and 15.8 times higher at day 40 (P < 0.001) posttreatment compared with those in the PTD-GFP–treated diabetic mice.

There is a marked increase (∼16 times) in the liver tissue insulin content at day 14 posttreatment in the rPdx1-treated mice over that of the PTD-GFP–treated mice (P < 0.001) and nearly a ninefold increase over that of normal liver (Fig. 6B). Interestingly, there was sharply reduced liver insulin content at day 40 post–rPdx1 treatment, although it was still 7.3 times higher than that in PTD-GFP–treated mice (P < 0.01) and ∼2 times higher than that in normal liver. These findings confirmed that in vivo rPdx1 treatment promoted pancreatic islet β-cell regeneration and transient liver cell transdifferentiation into IPCs, suggesting that liver and pancreas both contributed to achieving glucose homeostasis in a compensatory fashion.

Given that absolute and relative insulin deficiencies, respectively, form the basis of type 1 and 2 diabetes, the identification of a means for restoring functional β-cell mass would hold immense promise as a means for curing these disorders. In this study, we made several novel findings: 1) in vivo administration of recombinant Pdx1 ameliorates hyperglycemia in diabetic mice, 2) amelioration of hyperglycemia is attended by both pancreatic β-cell regeneration and liver cell transdifferentiation, and 3) the observed therapeutic effect is likely to require Pdx1 to have an intact protein transduction domain. Our experiments therefore constitute a proof-of-principle demonstration that protein therapy in the form of in vivo Pdx1 delivery to whole animals can be a highly effective therapeutic strategy, one that exploits intrinsic properties of a naturally occurring pancreatic transcription factor and that avoids undesirable effects typically associated with viral vector-mediated gene therapies.

Specifically, our results indicate that in vivo delivery of rPdx1 can promote both β-cell regeneration as well as liver cell transdifferentiation into IPCs. Pdx1-mediated pancreatic islet β-cell regeneration appears to be the dominant effect on glucose homeostasis, since marked hyperglycemia was observed in mice receiving nearly total pancreatectomy. Although the exact cellular and molecular events responsible for rPdx1-mediated β-cell regeneration and liver cell transdifferentiation remain to be defined, we believe that, based on our findings, rPdx1 protein enters the circulation via terminal veins and capillaries and penetrates cellular membranes to gain nucleus entry into target cells in the liver and pancreas (Fig. 2B), resulting in activation of rPdx1-dependent transcription factor cascade. The notion that rPdx1 promotes pancreatic β-cell regeneration is supported by the presence of numerous pancreatic large islets (Fig. 4A and B), an increasing number of proliferating islet cells (Fig. 4C), an ensuing increase in the level of pancreatic tissue insulin (Fig. 6A), and the significant upregulation of several key genes related to pancreatic cell regeneration in rPdx1-treated mice (Fig. 4D). The fact that rPdx1 vigorously promoted pancreatic islet cell proliferation and regeneration raises intriguing questions about the type of pancreatic cells that are the targets of rPdx1 (e.g., what is the cell origin for the newly regenerated islets?). Possible mechanisms include residual islet cell proliferation, exocrine acinar cell transdifferentiation, and pancreatic ductal/stem cell neogenesis. Presently, our approach limits us from tracking the cells of origin into/within the target tissue, resulting in the newly formed islets.

Several genes (INGAPrP, Reg-3γ, and PAP) upregulated by rPdx1 treatment are members of the pancreatic regenerating (Reg) gene family originally identified in animal models of β-cell regeneration (33). Their gene products play important roles in the maintenance of progenitors in the process of pancreas regeneration (10). Despite variation between individual samples, expression levels of the Reg genes determined by RT-PCR correlated well with an earlier study involving conditional expression of PDX1 in a transgenic mouse model (14).

The observed functional effects of rPdx1 on the liver are consistent with the published results of ectopic expression of PDX1 gene via adenovirus vectors resulting in liver cell transdifferentiation (17,32). In our experimental models, liver and pancreas appear to work in a sequential, compensatory manner to ameliorate hyperglycemia and ultimately to restore euglycemia in rPdx1-treated mice. The kinship between the liver and pancreas in controlling glucose homeostasis is also supported by a recent study using liver and pancreas double-injury animal models (34). The early phase of rPdx1-induced hepatic insulin production is supported by an intense expression of the insulin I gene (Fig. 5B and C) and a nearly 18-fold increase in liver tissue insulin in comparison with that of control liver (Fig. 6B). While the precise mechanism underlying the rPdx1-mediated surge in hepatic insulin during early-stage glucose homeostasis remains to be elucidated, possible explanations include the following: 1) action of hepatic insulin on pancreatic progenitor cells via the insulin signaling pathway to promote β-cell regeneration via IRS2-Akt-Pdx1–mediated signal transduction (35); 2) insulin-mediated facilitation of β-cell neogenesis, involving amelioration of hyperglycemic toxic effects on residual β-cell regeneration (36); and 3) rPdx1-mediated hepatic insulin production, resulting in an increased rate of glucose clearance by the liver, perhaps by promoting glucokinase expression and/or by insulin's stimulatory action on glycogen synthase, thereby lowering blood glucose levels (37).

Pdx1 expression has also been reported to be associated with β-cell neogenesis in rodent pancreas injury models (10,14,35,38). Although our studies showed roughly a 2 to 4× increase of PDX1 gene expression in the pancreas of rPdx1- over PTD-GFP–treated mice, pancreatic β-cell function appears to be exquisitely sensitive to small changes in PDX1 gene expression levels in both humans and mice (38,39). The rPdx1 protein can positively regulate PDX1 gene expression, as evidenced by upregulation of endogenous PDX1 gene expression in the livers of rPdx1-treated mice. These observations are consistent with the findings of others indicating that Pdx1 binds to its own promoter and positively regulates its own gene expression (40). These results suggest that rPdx1-based protein therapy may not require a large-dose or long-term treatment and, thus, may reduce or eliminate potential dosage-related systemic toxicity.

Although we have shown that rPdx1 effectively reverses hyperglycemia in diabetic mice, there are potential obstacles to clinical translation. One concern is that rPdx1 could be partially degraded by serum proteases. To assess this possibility, further studies on rPdx1 stability in whole blood and plasma would be helpful. Moreover, detailed pharmacokinetic studies are needed to optimize dosages, routes of delivery, and the interval between treatments. The polyclonal anti-Pdx1 antiserum used in this study precludes distinguishing between intact and any partially degraded rPdx1 protein. Significantly, the ability of rPdx1 to ameliorate hyperglycemia in diabetic mice, along with its nuclear localization in liver and pancreatic acinar cells (Fig. 2B), indicates in vivo availability of a sufficient amount of intact rPdx1 protein or biologically active degradation products capable of translocation into cells. Another concern is the potential toxicity of rPdx1 to off-target organs via its PTD, since the rPdx1 protein has the potential to enter almost any tissue or cell type. We can, however, exclude the activation of rPdx1 target genes in tissues other than liver and pancreas at day 14 posttreatment. Moreover, the animals appeared normal, without evidence of weight loss or abnormal organ morphology. In fact, the diabetic mice treated with rPdx1 gained body weight, showed an improved IPGTT, and exhibited markedly reduced blood glucose levels. Nonetheless, a full toxicity profile of rPdx1, especially at earlier time points, is required to address this question. These studies are beyond the scope of the present manuscript and will be pursued in the future.

Interestingly, the distribution of rPdx1 in the kidney is quite different from that in the liver and pancreas at the early 1-h time point (Fig. 2B). Instead of being present in the cell nuclei, rPdx1 was localized near the brush border of the proximal tubular cells (Fig. 2B3 and 2B6). Such a distribution was not observed at 24 h (Fig. 2B9). As the rPdx1 protein rapidly enters the bloodstream (Fig. 2A) after intraperitoneal injection, it may be filtered through glomerular capillaries into the urinary spaces via the fenestrated capillary endothelial cells and glomerular basement membrane. Alternatively, rPdx1 may gain entry via cells by virtue of its on-board PTD. Once in the urinary spaces, it would not be surprising if the cationic rPdx1 protein interacts electrostatically with polyanions (e.g., sialic acid and phopholipids) present on the apical surface of proximal tubular epithelial cells.

In conclusion, this demonstration that in vivo rPdx1 delivery into diabetic mice rapidly restores euglycemia exploits the intrinsic properties of this key pancreatic transcription factor (e.g., its built-in antennapedia-like PTD, its positive autoregulation, [28,41], its vital role in pancreatic cell development and regeneration [10,14], and its role in maintaining pancreatic β-cell function) (14,15,42). Indeed, we believe that Pdx1-based protein therapy should allow for a redirection (or reactivation) of pancreatic stem/precursor cell differentiation and transdifferentiation (or reprogramming) from nonpancreatic cells along pancreatic β-cell developmental pathways, a feature that could prove beneficial in treating patients with diabetes.

This work is supported in part by National Institutes of Health Grants DK064054 and DK071831 (to L.J. Yang).

We thank Drs. C.V. Wright and W.B. Coleman for providing Pdx1 antibody and WB cell line; L.J. Chang for lentivirus production; D.Q. Tang, S.A. Litherland, and E. Dooley for technical assistance; and J.C. Crawford for critical editorial and scientific inputs.