Automatic Segmentation of the Eye in 3D Magnetic Resonance Imaging: A novel Statistical Shape Model for treatment planning of Retinoblastoma

Purpose: Proper delineation of ocular anatomy in 3D imaging is a big challenge, particularly when developing treatment plans for ocular diseases. Magnetic Resonance Imaging (MRI) is nowadays utilized in clinical practice for the diagnosis confirmation and treatment planning of retinoblastoma in infants, where it serves as a source of information, complementary to the Fundus or Ultrasound imaging. Here we present a framework to fully automatically segment the eye anatomy in the MRI based on 3D Active Shape Models (ASM), we validate the results and present a proof of concept to automatically segment pathological eyes. Material and Methods: Manual and automatic segmentation were performed on 24 images of healthy children eyes (3.29±2.15 years). Imaging was performed using a 3T MRI scanner. The ASM comprises the lens, the vitreous humor, the sclera and the cornea. The model was fitted by first automatically detecting the position of the eye center, the lens and the optic nerve, then aligning the model and fitting it to the patient. We validated our segmentation method using a leave-one-out cross validation. The segmentation results were evaluated by measuring the overlap using the Dice Similarity Coefficient (DSC) and the mean distance error. Results: We obtained a DSC of 94.90±2.12% for the sclera and the cornea, 94.72±1.89% for the vitreous humor and 85.16±4.91% for the lens. The mean distance error was 0.26±0.09mm. The entire process took 14s on average per eye. Conclusion: We provide a reliable and accurate tool that enables clinicians to automatically segment the sclera, the cornea, the vitreous humor and the lens using MRI. We additionally present a proof of concept for fully automatically segmenting pathological eyes. This tool reduces the time needed for eye shape delineation and thus can help clinicians when planning eye treatment and confirming the extent of the tumor.

The differential staurosporine mediated G1 arrest in normal versus tumor cells is dependent on the retinoblastoma and Chk1 proteins

The ability to regulate cell cycle progression is one of the differences that separates normal from tumor cells. A protein, which is frequently mutated or deleted in a majority of tumor cells, is the retinoblastoma protein (pRb). Previously, we reported that normal cells, which have a wild-type Rb pathway, can be reversibly arrested in the G1 phase of the cell cycle by staurosporine (ST), while tumor cells were unaffected by this treatment. As a result, ST may be used to protect normal cells against the toxic affects of chemotherapy. Here we set out to determine the mechanism(s) by which ST can mediate a reversible G1 arrest in pRb positive cells. To this end, we used an isogenic cell model system of normal human mammary epithelial cells (HMEC) with either intact pRb+ (p53-) or p53+ (pRb-) treated with ST. Our results show that pRb+ cells treated with low concentrations of ST, arrested in the G1 phase of the cell cycle; however, in pRb - cells there was no response. This was verified as a true G 1 arrest in pRb+ cells by two different methods for monitoring cell cycle kinetics and in two additional model systems for Rb (i.e. pRb -/- mouse embryo fibroblasts, and downregulation of RB with siRNA). Our results indicated that ST-mediated G1 arrest required pRb, which in turn initiated a cascade of events leading to inhibition of CDK4 and CDK2 activities and up-regulation of p21 protein. Further assessment of this pathway revealed the novel finding that Chk1 expression and activity were required for the Rb-dependent, ST-mediated G1 arrest. In fact, overexpression of Chk1 facilitated recovery from ST-mediated G1 arrest, an effect only observed in RB+ cells. Collectively, our data suggest pRb is able to cooperate with Chk1 to mediate a G1 arrest in pRb+ cells, but not in pRb- cells. The elucidation of this pathway can help identify novel agents that can be used to protect cancer patients against the debilitating affects of chemotherapy, by targeting only the normal proliferating cells in the body that are otherwise destroyed. ^

Creation of a retinoblastoma mutant with dominant -negative activity

The Retinoblastoma tumor suppressor gene (RB) plays a role in a variety of human cancers. Experimental analyses have indicated that the protein product of the RB gene (pRb) plays a role in cell cycle regulation, and that this protein is required in cellular differentiation, senescence, and cell survival. pRb function is dependent on its ability to bind to cellular factors. There are multiple protein binding domains within pRb. Mutations within these domains which eliminate the ability of pRb to bind its targets result in loss of function. Loss of pRb function leads to tumorigenesis, although uncontrolled cellular proliferation is not a universal response to pRb inactivation. The ultimate response to the loss of pRb is influenced by both the genetic and epigenetic environments. Targeted disruption of RB in mice results in embryonic lethality, demonstrating the requirement for functional pRb in development. Close examination of various tissues from the embryos which lack wildtype RB shows problems in differentiation as well as showing induction of apoptosis. Although disruption of RB has provided useful information, complete inactivation of a gene precludes the possibility of discovering the functions that separate domains may have within the system. Creation of a dominant negative mutant by domain deletion whose phenotype is expressed in the presence of the wildtype may provide information about the intermediate functions of the protein. In addition, tissue specific targeting of a dominant negative mutant of pRb allows for comprehensive analysis of pRb function in organogenesis. In this thesis, a series of RB deletion mutants were created and tested for dominant negative activity as well as cellular localization. A tissue culture assay for dominant negative activity was developed which screens for the phenotype of apoptosis due to loss of pRb function. Two mutants from this series scored positive for dominant negative activity in this assay. The effect of these mutants within the assay environment can be explained by a model in which pRb acts as a facilitator of cell fate pathway decisions. ^

A thyroid hormone receptor coactivator negatively regulated by the retinoblastoma protein

The retinoblastoma protein (Rb) plays a critical role in cell proliferation, differentiation, and development. To decipher the mechanism of Rb function at the molecular level, we have systematically characterized a number of Rb-interacting proteins, among which is the clone C5 described here, which encodes a protein of 1,978 amino acids with an estimated molecular mass of 230 kDa. The corresponding gene was assigned to chromosome 14q31, the same region where genetic alterations have been associated with several abnormalities of thyroid hormone response. The protein uses two distinct regions to bind Rb and thyroid hormone receptor (TR), respectively, and thus was named Trip230. Trip230 binds to Rb independently of thyroid hormone while it forms a complex with TR in a thyroid hormone-dependent manner. Ectopic expression of the protein Trip230 in cells, but not a mutant form that does not bind to TR, enhances specifically TR-dependent transcriptional activity. Coexpression of wild-type Rb, but not mutant Rb that fails to bind to Trip230, inhibits such activity. These results not only identify a coactivator molecule that modulates TR activity, but also uncover a role for Rb in a pathway that responds to thyroid hormone.

Hypo-phosphorylation of the retinoblastoma protein (pRb) by cyclin D:Cdk4/6 complexes results in active pRb

In cycling cells, the retinoblastoma protein (pRb) is un- and/or hypo-phosphorylated in early G1 and becomes hyper-phosphorylated in late G1. The role of hypo-phosphorylation and identity of the relevant kinase(s) remains unknown. We show here that hypo-phosphorylated pRb associates with E2F in vivo and is therefore active. Increasing the intracellular concentration of the Cdk4/6 specific inhibitor p15INK4b by transforming growth factor β treatment of keratinocytes results in G1 arrest and loss of hypo-phosphorylated pRb with an increase in unphosphorylated pRb. Conversely, p15INK4b-independent transforming growth factor β-mediated G1 arrest of hepatocellular carcinoma cells results in loss of Cdk2 kinase activity with continued Cdk6 kinase activity and pRb remains only hypo-phosphorylated. Introduction of the Cdk4/6 inhibitor p16INK4a protein into cells by fusion to a protein transduction domain also prevents pRb hypo-phosphorylation with an increase in unphosphorylated pRb. We conclude that cyclin D:Cdk4/6 complexes hypo-phosphorylate pRb in early G1 allowing continued E2F binding.

Incomplete penetrance of familial retinoblastoma linked to germ-line mutations that result in partial loss of RB function

To study the molecular basis for the clinical phenotype of incomplete penetrance of familial retinoblastoma, we have examined the functional properties of three RB mutations identified in the germ line of five different families with low penetrance. RB mutants isolated from common adult cancers and from classic familial retinoblastoma (designated as classic RB mutations) are unstable and generally do not localize to the nucleus, do not undergo cyclin-dependent kinase (cdk)-mediated hyperphosphorylation, show absent protein “pocket” binding activity, and do not suppress colony growth of RB(−) cells. In contrast, two low-penetrant alleles (661W and “deletion of codon 480”) retained the ability to localize to the nucleus, showed normal cdk-mediated hyperphosphorylation in vivo, exhibited a binding pattern to simian virus 40 large T antigen using a quantitative yeast two-hybrid assay that was intermediate between classic mutants (null) and wild-type RB, and had absent E2F1 binding in vitro. A third, low-penetrant allele, “deletion of RB exon 4,” showed minimal hyperphosphorylation in vivo but demonstrated detectable E2F1 binding in vitro. In addition, each low-penetrant RB mutant retained the ability to suppress colony growth of RB(−) tumor cells. These findings suggest two categories of mutant, low-penetrant RB alleles. Class 1 alleles correspond to promoter mutations, which are believed to result in reduced or deregulated levels of wild-type RB protein, whereas class 2 alleles result in mutant proteins that retain partial activity. Characterization of the different subtypes of class 2 low-penetrant genes may help to define more precisely functional domains within the RB product required for tumor suppression.

A maize cDNA encoding a member of the retinoblastoma protein family: involvement in endoreduplication.

Retinoblastoma (RB-1) is a tumor suppressor gene that encodes a 105-kDa nuclear phosphoprotein. To date, RB genes have been isolated only from metazoans. We have isolated a cDNA from maize endosperm whose predicted protein product (ZmRb) shows homology to the "pocket" A and B domains of the Rb protein family. We found ZmRb behaves as a pocket protein based on its ability to specifically interact with oncoproteins encoded by DNA tumor viruses (E7, T-Ag, E1A). ZmRb can interact in vitro and in vivo with the replication-associated protein, RepA, encoded by the wheat dwarf virus. The maize Rb-related protein undergoes changes in level and phosphorylation state concomitant with endoreduplication, and it is phosphorylated in vitro by an S-phase kinase from endoreduplicating endosperm cells. Together, our results suggest that ZmRb is a representative of the pocket protein family and may play a role in cell cycle progression. Moreover, certain plant monopartite geminiviruses may operate similarly to mammalian DNA viruses, by targeting and inactivating the retinoblastoma protein, which otherwise induces G1 arrest.

G1 arrest and down-regulation of cyclin E/cyclin-dependent kinase 2 by the protein kinase inhibitor staurosporine are dependent on the retinoblastoma protein in the bladder carcinoma cell line 5637.

The protein kinase inhibitor staurosporine has been shown to induce G1 phase arrest in normal cells but not in most transformed cells. Staurosporine did not induce G1 phase arrest in the bladder carcinoma cell line 5637 that lacks a functional retinoblastoma protein (pRB-). However, when infected with a pRB-expressing retrovirus [Goodrich, D. W., Chen, Y., Scully, P. & Lee, W.-H. (1992) Cancer Res. 52, 1968-1973], these cells, now pRB+, were arrested by staurosporine in G1 phase. This arrest was accompanied by the accumulation of hypophosphorylated pRB. In both the pRB+ and pRB- cells, cyclin D1-associated kinase activities were reduced on staurosporine treatment. In contrast, cyclin-dependent kinase (CDK) 2 and cyclin E/CDK2 activities were inhibited only in pRB+ cells. Staurosporine treatment did not cause reductions in the protein levels of CDK4, cyclin D1, CDK2, or cyclin E. The CDK inhibitor proteins p21(Waf1/Cip1) and p27 (Kip1) levels increased in staurosporine-treated cells. Immunoprecipitation of CDK2, cyclin E, and p2l from staurosporine-treated pRB+ cells revealed a 2.5- to 3-fold higher ratio of p2l bound to CDK2 compared with staurosporine-treated pRB- cells. In pRB+ cells, p2l was preferentially associated with Thrl6O phosphorylated active CDK2. In pRB- cells, however, p2l was bound preferentially to the unphosphorylated, inactive form of CDK2 even though the phosphorylated form was abundant. This is the first evidence suggesting that G1 arrest by 4 nM staurosporine is dependent on a functional pRB protein. Cell cycle arrest at the pRB- dependent checkpoint may prevent activation of cyclin E/CDK2 by stabilizing its interaction with inhibitor proteins p2l and p27.

Genomic structure of the human retinoblastoma-related Rb2/p130 gene.

The human Rb2/p130 gene shares many structural and functional features with the retinoblastoma gene and the retinoblastoma-related p107 gene. In the present study, we have cloned and partially sequenced the gene coding for the Rb2/p130 protein from human genomic libraries. The complete intron-exon organization of this gene has been elucidated. The gene contains 22 exons spanning over 50 kb of genomic DNA. The length of individual exons ranges from 65 to 1517 bp. The largest intron spans over 9 kb, and the smallest has only 82 bp. The 5' flanking region revealed a structural organization characteristic of promoters of "housekeeping" and growth control-related genes. A typical TATA or CAAT box is not present, but there are several GC boxes and potential binding sites for numerous transcription factors. This study provides the molecular basis for understanding the transcriptional control of the Rb2/p130 gene and for implementing a comprehensive Rb2/p130 mutation screen using genomic DNA as a template.

Regulation of the retinoblastoma protein-related protein p107 by G1 cyclin-associated kinases.

p107 is a retinoblastoma protein-related phosphoprotein that, when overproduced, displays a growth inhibitory function. It interacts with and modulates the activity of the transcription factor, E2F-4. In addition, p107 physically associates with cyclin E-CDK2 and cyclin A-CDK2 complexes in late G1 and at G1/S, respectively, an indication that cyclin-dependent kinase complexes may regulate, contribute to, and/or benefit from p107 function during the cell cycle. Our results show that p107 phosphorylation begins in mid G1 and proceeds through late G1 and S and that cyclin D-associated kinase(s) contributes to this process. In addition, E2F-4 binds selectively to hypophosphorylated p107, and G1 cyclin-dependent p107 phosphorylation leads to the dissociation of p107-E2F-4 complexes as well as inactivation of p107 G1 blocking function.

Retinoblastoma protein directly interacts with and activates the transcription factor NF-IL6.

The biological function of the retinoblastoma protein (RB) in the cell division cycle has been extensively documented, but its apparent role in differentiation remains largely unexplored. To investigate how RB is involved in differentiation, the U937 large-cell lymphoma line was induced to differentiate along a monocyte/macrophage lineage. During differentiation RB was found to interact directly through its simian virus 40 large tumor antigen (T antigen)-binding domain with NF-IL6, a member of the CAAT/enhancer-binding protein (C/EBP) family of transcription factors. NF-IL6 utilizes two distinct regions to bind to the hypophosphorylated form of RB in vitro and in cells. Wild-type but not mutant RB enhanced both binding activity of NF-IL6 to its cognate DNA sequences in vitro and promoter transactivation by NF-IL6 in cells. These findings indicate a novel biochemical function of RB: it activates, by an apparent chaperone-like activity, specific transcription factors important for differentiation. This contrasts with its sequestration and inactivation of other transcription factors, such as E2F-1, which promote progression of the cell cycle. Such disparate mechanisms may help to explain the dual role of RB in cell differentiation and the cell division cycle.

A potent transrepression domain in the retinoblastoma protein induces a cell cycle arrest when bound to E2F sites.

An intact T/E1A-binding domain (the pocket) is necessary, but not sufficient, for the retinoblastoma protein (RB) to bind to DNA-protein complexes containing E2F and for RB to induce a G1/S block. Indirect evidence suggests that the binding of RB to E2F may, in addition to inhibiting E2F transactivation function, generate a complex capable of functioning as a transrepressor. Here we show that a chimera in which the E2F1 transactivation domain was replaced with the RB pocket could, in a DNA-binding and pocket-dependent manner, mimic the ability of RB to repress transcription and induce a cell cycle arrest. In contrast, a transdominant negative E2F1 mutant that is capable of blocking E2F-dependent transactivation did not. Fusion of the RB pocket to a heterologous DNA-binding domain unrelated to E2F likewise generated a transrepressor protein when scored against a suitable reporter. These results suggest that growth suppression by RB is due, at least in part, to transrepression mediated by the pocket domain bound to certain promoters via E2F.

Opposing early and late effects of insulin-like growth factor I on differentiation and the cell cycle regulatory retinoblastoma protein in skeletal myoblasts.

The mechanisms by which insulin-like growth factors (IGFs) can be both mitogenic and differentiation-promoting in skeletal myoblasts are unclear because these two processes are believed to be mutually exclusive in this tissue. The phosphorylation state of the ubiquitous nuclear retinoblastoma protein (Rb) plays an important role in determining whether myoblasts proliferate or differentiate: Phosphorylated Rb promotes mitogenesis, whereas un- (or hypo-) phosphorylated Rb promotes cell cycle exit and differentiation. We hypothesized that IGFs might affect the fate of myoblasts by regulating the phosphorylation of Rb. Although long-term IGF treatment is known to stimulate differentiation, we find that IGFs act initially to inhibit differentiation and are exclusively mitogenic. These early effects of IGFs are associated with maintenance of Rb phosphorylation typical of proliferating cells; upregulation of the gene expression of cyclin-dependent kinase 4 and cyclin D1, components of a holoenzyme that plays a principal role in mediating Rb phosphorylation; and marked inhibition of the gene expression of myogenin, a member of the MyoD family of skeletal muscle-specific transcription factors that is essential in muscle differentiation. We also find that IGF-induced inhibition of differentiation occurs through a process that is independent of its mitogenic effects. We demonstrate, thus, that IGFs regulate Rb phosphorylation and cyclin D1 and cyclin-dependent kinase 4 gene expression; together with their biphasic effects on myogenin expression, these results suggest a mechanism by which IGFs are initially mitogenic and subsequently differentiation-promoting in skeletal muscle.

Lack of functional retinoblastoma protein mediates increased resistance to antimetabolites in human sarcoma cell lines.

Growth inhibition assays indicated that the IC50 values for methotrexate (MTX) and 5-fluorodeoxyuridine (FdUrd) in HS-18, a liposarcoma cell line lacking retinoblastoma protein (pRB), and SaOS-2, an osteosarcoma cell line with a truncated and nonfunctional pRB, were 10- to 12-fold and 4- to 11-fold higher, respectively, than for the HT-1080 (fibrosarcoma) cell line, which has wild-type pRB. These Rb-/- cell lines exhibited a 2- to 4-fold increase in both dihydrofolate reductase (DHFR) and thymidylate synthase (TS) enzyme activities as well as a 3- to 4-fold increase in mRNA levels for these enzymes compared to the HT-1080 (Rb+/+) cells. This increase in expression was not due to amplification of the DHFR and TS genes. Growth inhibition by MTX and FdUrd was increased and DHFR and TS activities and expression were correspondingly decreased in Rb transfectants of SaOS-2 cells. In contrast, there was no significant difference in growth inhibition among these cell lines for the nonantimetabolites VP-16, cisplatin, and doxorubicin. A gel mobility-shift assay showed that parental SaOS-2 cells had increased levels of free E2F compared to the Rb-reconstituted SaOS-2 cells. These results indicate that pRB defective cells may have decreased sensitivity to growth inhibition by target enzymes encoded by genes whose transcription is enhanced by E2F proteins and suggest mechanisms of interaction between cytotoxic agents and genes involved in cell cycle progression.

Induction of a retinoblastoma phosphatase activity by anticancer drugs accompanies p53-independent G1 arrest and apoptosis.

DNA-damaging agents induce accumulation of the tumor suppressor and G1 checkpoint protein p53, leading cells to either growth arrest in G1 or apoptosis (programmed cell death). The p53-dependent G1 arrest involves induction of p21 (also called WAF1/CIP1/SDI1), which prevents cyclin kinase-mediated phosphorylation of retinoblastoma protein (RB). Recent studies suggest a p53-independent G1 checkpoint as well; however, little is known about its molecular mechanisms. We report that induction of a protein-serine/threonine phosphatase activity by DNA damage signals is at least one of the mechanisms responsible for p53-independent, RB-mediated G1 arrest and consequent apoptosis. When two p53-null human leukemic cell lines (HL-60 and U-937) were treated with a variety of anticancer agents, RB became hypophosphorylated, accompanied with G1 arrest. This was followed immediately (in less than 30 min) by apoptosis, as determined by the accumulation of pre-G1 apoptotic cells and the internucleosomal fragmentation of DNA. Addition of calyculin A or okadaic acid (specific serine/threonine phosphatase inhibitors) or zinc chloride (apoptosis inhibitor) prevented the G1 arrest- and apoptosis-specific RB dephosphorylation. The levels of cyclin E- and cyclin A-associated kinase activities remained high during RB dephosphorylation, supporting the involvement of a chemotherapy-induced serine/threonine phosphatase(s) rather than p21. Furthermore, the induced phosphatase activity coimmunoprecipitated with the hyperphosphorylated RB and was active in a cell-free system that reproduced the growth arrest- and apoptosis-specific RB dephosphorylation, which was inhibitable by calyculin A but not zinc. We propose that the RB phosphatase(s) might be one of the p53-independent G1 checkpoint regulators.