Hypoxia-induced human endonuclease G expression suppresses tumor growth in a xenograft model

Abstract

We have developed a hypoxia-inducible gene therapy approach for the expression of the mature form of human endonuclease G to facilitate cell death in hypoxic regions of the tumor. The chimeric therapeutic gene is placed under the control of a hypoxia response element based promoter and contains a translocation motif linked in frame to an oxygen-dependent degradation domain and the endonuclease G gene. Transient expression of the chimeric therapeutic gene in breast and prostate cancer cell lines resulted in efficient cell death under hypoxia-mimetic conditions. Stable MDA-MB-435 cells expressing the chimeric therapeutic gene under 1% O2 showed an increase in stable HIF-1alpha protein levels and synthesis of the endonuclease G protein in a time-dependent manner. In normoxic conditions, these stable transgenic cells exhibited no change in growth rate, invasion and motility when compared to parental cells. Moreover, xenografts generated using the transgenic cells exhibited highly significant suppression of tumor growth in a preclinical cancer model compared to the parental cell line. Thus, the hypoxia-modulated endonuclease G expression has the potential to be used as a gene-based-therapy system to kill malignant cells within hypoxic regions of tumors.

Figure 1

Ilustration of the chimeric hEndoG toxic protein and verification of its killing function during transient expression in tumor cells, (a) A linear representation of the chimeric hEndoG construct is depicted. The N-terminus of the fusion protein contains a translocation motif (TLM) that is fused in-frame with a linker (L) that is predicted to be a random coil. An exploded view of this portion of the protein is shown beneath the illustration and provides the amino acid and nucleotide sequence for the novel N-terminus. The underlined nucleotide sequences indicate Xma I and Age I restriction sites. The Xma I site demarcates the junction between the TLM and L. The Age I site provided an in-frame fusion site with the oxygen-dependent degradation domain (ODD). The two proline residues designated 402 and 564 are the positions that are hydroxylated in the native HIF-1α protein. The ODD is followed by another L sequence, which is connected with the mature form of hEndoG. The 435 (b–e), PC-3 (f–i) and MCF-7 (j–m) cells were transiently transfected with the p(ΔEF-1α)-5HRE-TLM-ODD-hEndoG plus pEF-1α tdTomato and grown without CoCl₂ (b, c, f, g, j and k) or in the presence of CoCl₂ (d, e, h, i, I and m). Most transfected cells were undergoing death, have detached from the substratum, and are thus are seen as floating.

Figure 2

Induction of hEndoG in stable 435-hEndoG cells promotes cell death. Representative phase contrast photomicrographs are shown of parental 435 cells (a–d) and a stable clone of 435-hEndoG cells (e–h) obtained post CoCl₂ treatment (0, 24, 48 and 72 h). Adherent live cell counts were obtained for each time point during different experiments and average cell numbers determined. The average live cell counts at each time point were normalized to zero time points (100%) as shown in plots of these resulting average percentages (±s.d.; n = 8). The difference between the means at the 24, 48 and 72 h time points was significant as determined by a one-tailed t-test (P-value <6 × 10⁻⁷ for each case). Three independent clones were analyzed.

Figure 3

Viability of 435 and 435-hEndoG cells following CoCl₂ treatment. The 435 (a and b) and 435-hEndoG (c and d) cells were treated with CoCl₂ for 72 h. Following CoCl₂ incubation, the cells were cultured for 5 days in normal media conditions (b and d) and stained with Calcein-AM (insets) to determine cell viability. As shown, 435-hEndoG cells were unable to completely recover as is evident from the sparse cell density as compared to the 435 cells.

Figure 4

In vitro characterization of 435-hEndoG. (a) Bar graphs showing percentage of cells exhibiting motility across a transwell membrane or invasion through Matrigel for 435 (white bars) and 435-hEndoG (black bars). It is seen that the both cell lines had very similar motilities and invasiveness in these assays (n = 3±s.d.). (b) Growth curves (n = 2 for 12 & 24 h and n = 3 for 48 and 72 h ±s.d.) for 435 (filled squares) and 435-hEndoG (filled triangles) shows that these cell lines have identical growth rates, (c) Longitudinal tracking (0, 24, 48 and 72 h) of migration into and filling of a ‘wound’ in a wound healing assay by 435 or 435-hEndoG cells. The top photographs shows that initially both cell lines were nearly confluent. Within each time point photograph, red lines demarcate the boundaries of the introduced wound.

Figure 5

Kinetics of expression of hEndoG and its affect on genomic DNA. (a) The 435 and 435-hEndoG cells were maintained under normoxic condition (20% O₂) or cultured in moderate hypoxia (1% O₂). Cells were harvested at the indicated time points and whole cell protein extracts prepared. The membrane was scored for HIF-1α and hEndoG with monoclonal anti-HIF-1α-ODD. The ribosomal protein 36B4 was used as a loading control and scored for with an in house generated polyclonal anti-36B4. (b) Following addition of CoCl₂, genomic DNA was extracted at 12, 24, 48, 72 and 96 h as well as from untreated control cells. Following electrophoresis in a 0.6% agarose gel, the DNA smear pattern was predominant in 435-hEndoG lanes as compared to the 435 lanes.

Figure 6

Stable 435-hEndoG cells exhibit greatly attenuated growth as xenografts. The same animals (1–5) are shown in both photographs, that is, 435 cells were injected into the second mammary fat pad on the left side (ventral side up) whereas 435-hEndoG cells were injected into the contra-lateral mammary fat pad. All mice received 4 × 10⁶ cells per mammary fat pad and photographs were taken 6 weeks later. In all cases, tumors are encircled in yellow, which shows that tumors generated from 435 cells are visibly much larger than tumors generated from 435-hEndoG cells. Growth curves for 435 (filled squares) and 435-hEndoG cells (open squares) generated tumors for each mouse (1–5) are presented beneath the photographs. The far right-hand graph presents mean tumor sizes (error bars are ± s.d.) for mice 1–4 except at week 7 where averages of tumor growth on mice 2 and 4 are presented. One-tailed t-test P-values are: P = 0.0009, P = 0.001, P = 0.01 and P = 0.0005, for weeks 3, 4, 5 and 6, respectively. Neither cell line produced a tumor of substantial size on mouse 5.

Figure 7

Histological comparison of 435 and 435-hEndoG tumor sections reveals differences in composition. Formalin-fixed 5µm sections of 435 (a–d) and 435-hEndoG (e–h) tumors were hematoxylin and eosin stained. All four 435 tumors are shown to have a necrotic (N) core that is surrounded by tumor (T) cells. In contrast, necrotic regions are completely (e, f and h) or largely (g) absent from the much smaller 435-hEndoG tumors. Arrows in e and h point to residual ductal structures whereas asterisks in these images indicate adipocytic tissue.