Review

The potential role of transforming growth factor beta family ligand interactions in prostate cancer

  • The transforming growth factor beta (TGF-β) family plays an important role in embryonic development and control of the cell cycle. Members of the TGF-β family have pleiotropic functions and are involved in both the inhibition and progression of various cancers. In particular, deregulation of the TGF-β family has been associated with prostate cancer, as both a mechanism of disease progression and a possible therapeutic target. This review concentrates on the TGF-βs, activins and inhibins, bone morphogenetic proteins and NODAL and their connection to prostate cancer. Whilst most studies examine the family members in isolation, there are multiple interactions that may occur between members which can alter their function. Such interactions include ligand competition for receptor binding and shared intracellular pathways such as the Mothers against decapentaplegic (SMAD) proteins. Another mechanism for interaction within the TGF-β family is facilitated by their dimeric structure; heterodimers can form which exhibit different functional capabilities to their homodimeric counterparts. The potential formation of TGF-β family heterodimers has not been well examined in prostate cancer. The multiple methods of interrelations between members highlights the need for gross analysis of the TGF-β family and related factors in association with prostate cancer, in order to discover possible future avenues for TGF-β based diagnosis and treatments of the disease. This review describes the role of the TGF-β family of proteins in cancer and, in particular, prostate cancer. After a brief overview, the role of individual members of the family is considered and how these members may be involved in prostate cancer growth is discussed. The review highlights the complex interactions that occur between family members and that may contribute to the progression of prostate cancer.

    Citation: Kit P. Croxford, Karen L. Reader, Helen D. Nicholson. The potential role of transforming growth factor beta family ligand interactions in prostate cancer[J]. AIMS Molecular Science, 2017, 4(1): 41-61. doi: 10.3934/molsci.2017.1.41

    Related Papers:

    [1] Sakander Hayat, Bagus Imanda, Asad Khan, Mohammed J. F. Alenazi . Three infinite families of Hamilton-connected convex polytopes and their detour index. AIMS Mathematics, 2025, 10(5): 12343-12387. doi: 10.3934/math.2025559
    [2] Milica Anđelić, Tamara Koledin, Zoran Stanić . Notes on Hamiltonian threshold and chain graphs. AIMS Mathematics, 2021, 6(5): 5078-5087. doi: 10.3934/math.2021300
    [3] Wenke Zhou, Guo Chen, Hongzhi Deng, Jianhua Tu . Enumeration of dissociation sets in grid graphs. AIMS Mathematics, 2024, 9(6): 14899-14912. doi: 10.3934/math.2024721
    [4] Yubin Zhong, Sakander Hayat, Suliman Khan, Vito Napolitano, Mohammed J. F. Alenazi . Combinatorial analysis of line graphs: domination, chromaticity, and Hamiltoniancity. AIMS Mathematics, 2025, 10(6): 13343-13364. doi: 10.3934/math.2025599
    [5] Xia Liu . A related problem on s-Hamiltonian line graphs. AIMS Mathematics, 2022, 7(10): 19553-19561. doi: 10.3934/math.20221073
    [6] Shama Liaqat, Zeeshan Saleem Mufti, Yilun Shang . Newly defined fuzzy Misbalance Prodeg Index with application in multi-criteria decision-making. AIMS Mathematics, 2024, 9(8): 20193-20220. doi: 10.3934/math.2024984
    [7] Xia Hong, Wei Feng . Completely independent spanning trees in some Cartesian product graphs. AIMS Mathematics, 2023, 8(7): 16127-16136. doi: 10.3934/math.2023823
    [8] G. Nandini, M. Venkatachalam, Raúl M. Falcón . On the r-dynamic coloring of subdivision-edge coronas of a path. AIMS Mathematics, 2020, 5(5): 4546-4562. doi: 10.3934/math.2020292
    [9] Yunfeng Tang, Huixin Yin, Miaomiao Han . Star edge coloring of K2,t-free planar graphs. AIMS Mathematics, 2023, 8(6): 13154-13161. doi: 10.3934/math.2023664
    [10] Ali N. A. Koam, Adnan Khalil, Ali Ahmad, Muhammad Azeem . Cardinality bounds on subsets in the partition resolving set for complex convex polytope-like graph. AIMS Mathematics, 2024, 9(4): 10078-10094. doi: 10.3934/math.2024493
  • The transforming growth factor beta (TGF-β) family plays an important role in embryonic development and control of the cell cycle. Members of the TGF-β family have pleiotropic functions and are involved in both the inhibition and progression of various cancers. In particular, deregulation of the TGF-β family has been associated with prostate cancer, as both a mechanism of disease progression and a possible therapeutic target. This review concentrates on the TGF-βs, activins and inhibins, bone morphogenetic proteins and NODAL and their connection to prostate cancer. Whilst most studies examine the family members in isolation, there are multiple interactions that may occur between members which can alter their function. Such interactions include ligand competition for receptor binding and shared intracellular pathways such as the Mothers against decapentaplegic (SMAD) proteins. Another mechanism for interaction within the TGF-β family is facilitated by their dimeric structure; heterodimers can form which exhibit different functional capabilities to their homodimeric counterparts. The potential formation of TGF-β family heterodimers has not been well examined in prostate cancer. The multiple methods of interrelations between members highlights the need for gross analysis of the TGF-β family and related factors in association with prostate cancer, in order to discover possible future avenues for TGF-β based diagnosis and treatments of the disease. This review describes the role of the TGF-β family of proteins in cancer and, in particular, prostate cancer. After a brief overview, the role of individual members of the family is considered and how these members may be involved in prostate cancer growth is discussed. The review highlights the complex interactions that occur between family members and that may contribute to the progression of prostate cancer.


    Equity derivative market-making environments are known for their competitive pricing requirements in terms of both speed and accuracy. Large fluctuations of the underlying security price, together with other pricing parameters, can occur fast and without warning and are influenced by company news, as well as macroeconomic, financial, and political factors. If market makers cannot adjust the corresponding option prices instantaneously, arbitrage opportunities can occur.

    There are many reasons why a market maker could be slow in updating the price of a product. Firstly, high overall system latency can potentially cause a delay in the arrival of market data to the trading system. Secondly, the calibration process of an implied volatility surface (e.g., using a chosen parametric volatility model such as the Heston stochastic volatility model) based on new market price arrivals can represent another potential source of latency. The third step, however, is typically the most prevalent source of computational latency. Here, current market data and the corresponding fitted parametric volatility model are inputted into an algorithm to evaluate the price of a derivative. Depending on the option type, the computational performance of a pricer is affected by its underlying algorithm and the complexities in the payoff structure of the option.

    This paper proposes a method that accelerates the American-style option pricing mechanism. We do not investigate the closely connected data acquisition, calibration, and price publishing parts. We assume that a trading system gathers market data and other parameters during the data acquisition and calibration phases and provides them to the pricing engine. We focus on the pricing algorithm that calculates what is known as the fair price of the option. This is the price at which no theoretical arbitrage exists – the probabilistic expectation of a profit on both sides of a transaction is equal to zero. In practice, a bid-ask spread is then generated around the computed fair price, and both bid and ask prices are subsequently published on the market platform.*

    *Depending on the current inventory (i.e., risk position) of the market-making trading desk that is pricing the option, the trading system can allocate aggressiveness to either the bid or the ask side by adjusting the spread around the fair price. Aggressiveness settings allow a trading desk to reduce its risk without hedging by attracting a specific direction of transactions.

    Since the main objective of the proposed method is its adoption in market making, a key responsibility of the proposed algorithm lies in its compliance with regulatory guidelines. Even though we use neural networks to accelerate American option pricing, we ensure that our algorithm is not a black box. This is achieved by generating a training set using a standard well-known pricing method (e.g., the PDE-based pricer under the Heston stochastic volatility model) instead of alternative neural network-based methods employing market data in the training set. Subsequently, we utilize the generated training set for training a neural network over the assumed (e.g., Heston-based) parameter space. By training the neural network over a set of all possible parameter space configurations given a chosen parameter range offline, such an approach allows for: ⅰ) decreasing the latency in the pricing of American options to near-instantaneous and ⅱ) enhancing the transparency and explainability of the American option pricing algorithm. Furthermore, we show that the prices resulting from the proposed deep explainable pricer are sufficiently smooth with respect to the input parameters–one of the algorithmic criteria in eXplainable Artificial Intelligence.

    In pricing path-dependent options, conventional methods rely extensively on computationally expensive numerical techniques. The most common pricing approaches are binomial models, numerical partial differential equation (PDE) methods, and Monte Carlo (MC) methods. Since the difficulties in pricing American options lie in the question of how to evaluate optimal exercise, the key differences between the methods arise from how they approach the optimal exercise decision. Cox et al. (1979) proposed the initial binomial tree method, which is fast and reasonably accurate. However, it does not work in the case of stochastic volatility. Furthermore, a trinomial tree that converges faster than the binomial tree was developed by Boyle (1986). Tree models rely on constant volatility and determine the optimal exercise by comparing the immediate exercise value and the discounted probability-weighted cashflows at each node.

    To generalize to the non-constant volatility models, researchers developed methods based on numerical solutions to the Black-Scholes-Merton PDE, which holds in general for both European and American option styles. PDE methods for pricing American options admit the representation of the optimal exercise decision as a PDE boundary condition. The PDE with its boundary condition is solved numerically using various numerical solvers such as modified finite difference, finite element, and boundary projection methods. One of the pioneers in this area were Brennan and Schwartz (1977), who priced American put options using a boundary projection method. Later, Ikonen and Toivanen (2004) proposed an approach using the Crank-Nicolson method (i.e., a type of finite difference method, see Crank and Nicolson (1947)), which was found to be more efficient than the boundary projection method. Nevertheless, PDE methods are generally too slow for a market-making environment, especially in the presence of stochastic interest rates, stochastic volatility models, jump processes, multi-dimensional options, or derivatives with other complicated features.

    Furthermore, researchers analyzed the use of statistical simulation methods known as Monte Carlo (MC) methods for option pricing. These methods are slower than the other methods mentioned above. However, they proved to be more flexible in terms of their ability to price derivatives under different volatility dynamics, more complex path-dependent payoff structures, and further departures from the basic Black et al. (1973) framework. Initial MC methods for American option pricing suffered under the curse of dimensionality since the decision process of optimal exercise relied on performing an additional MC simulation for each in-the-money path as proposed by Boyle et al. (1997). Thus, the number of paths increased exponentially. Famously, Longstaff and Schwartz (2001) proposed an MC method where the exercise decision relies upon comparing the current exercise payoff and a least-squares approximation of the option continuation value. This method circumvents the dimensionality problem of the nested MC approach and is thus highly effective. However, the Longstaff-Schwartz (LS) method is too slow for a market-making environment. Recently, the area of rough volatility models has evolved, see Bayer et al. (2016). Due to the non-Markovian dynamics of rough volatility models, pricing with Monte Carlo methods is required.

    The use of artificial neural networks (ANNs) for option pricing started with Malliaris and Salchenberger (1993) and Hutchinson et al. (1994). The authors show that the Black-Scholes-based European-style option valuation can be successfully recovered with a neural network. These works serve as a valuable reference point that spurred a wide range of future research involving machine learning methods applied to option pricing and hedging. Lajbcygier and Connor (1997) study the difference between hybrid neural network-based option valuation and observed market option prices and show that the pricing bias can be reduced with bootstrap methods. Anders et al. (1998) show that applying statistical inference techniques to build neural network pricing models leads to superior out-of-sample pricing performance compared to the Black-Scholes model. Garcia and Gençay (2000) incorporate financial domain knowledge in the form of a homogeneous option pricing function directly into the construction of an ANN pricer and show that such an approach reduces the out-of-sample mean squared prediction error. On a similar note, Dugas et al. (2009) show that the generalization performance of a learning algorithm can be improved by incorporating prior knowledge into its architecture. The authors propose new types of function classes that incorporate the a priori constraints based on the knowledge that the function to be learned is non-decreasing in its two arguments and convex in one of them. Pricing of S & P-500 European call options with a non-parametric modular neural network model is investigated in Gradojevic et al. (2009).

    An extensive literature review of the applications of neural networks to option pricing is provided in Ruf and Wang (2020).

    Neural networks have also been studied in the context of option hedging as well as implied volatility smile and its calibration. The European-style option hedging is investigated in Chen and Sutcliffe (2012) and Shin and Ryu (2012). More recently, Buehler et al. (2019) present a deep hedging-based framework for hedging a portfolio of derivatives in the presence of market frictions. The implied volatility smile is, in a GARCH-neural network approach, studied in Meissner and Kawano (2001). Computation of implied volatilities is, among others, also studied in Andreou et al. (2010) and Liu et al. (2019b). More recently, Horvath et al. (2021) propose an application of neural networks in a two-step process where firstly, the European option pricing map is learned, and secondly, the European option implied volatility surfaces are calibrated to various parametric models. Deep learning calibration is studied also in Hernandez (2016), Ferguson and Green (2018), Liu et al. (2019a), and Itkin (2019).

    The literature presented up to now studies an application of ANNs to the pricing of European-style options. Additionally, applying machine learning methods to solving the American option pricing and hedging problem has already been explored. Kelly (1994) presents the first study of pricing and hedging American put options using neural networks. A similar analysis is presented also in Pires and Marwala (2004) and Morelli et al. (2004). More recently, Gaspar et al. (2020) use neural networks to price American put options on four large U.S. companies. Generally, the American option pricing approaches fall into the categories of ⅰ) regression-enhancement for basket options as in the case of Kohler et al. (2010), ⅱ) solving the optimal stopping problem as in the case of Becker et al. (2020), or ⅲ) a mix of both as in Chen and Wan (2021). Moreover, machine learning methods can be used to approximate value functions emerging in dynamic programming, for example, arising in the American option pricing problem, see Ye and Zhang (2019). Similarly, Fecamp et al. (2019) apply ANNs to solve the pricing and hedging problem under market frictions such as transaction costs.

    It should be noted that while our paper focuses on the pricing of American options, other types of options, such as exchange options, have been thoroughly explored in recent works (Pasricha and He, 2023; He and Lin, 2023a).

    This paper contributes to the literature on option pricing by introducing a direct application of deep neural networks to price American-style options, which is relevant as the majority of the single stock options traded on the market are of the American style. In contrast to existing literature that focuses on European-style options, such as Horvath et al. (2021), we extend our methodology to American-style options by learning not only asset price dynamics but also the dynamics of a chosen volatility model, as well as the American option optimal exercise strategy, which has not been explored in the existing literature. One of the key advantages of our approach is that it learns the direct map between model parameters and price, providing a more interpretable and transparent approach than previous methods that rely on an implied volatility surface.

    In addition to these contributions, our approach can be seen as an explainable method for pricing American-style options. By explicitly modeling the dynamics of the assumed volatility model and the corresponding optimal exercise strategy, our model provides a transparent and intuitive way to understand how American option prices are determined. Our results demonstrate that the proposed neural network pricer considerably improves upon the widely-used American-style option pricing models, such as PDE or Monte Carlo methods, in terms of both parameter interpretability and speed-accuracy trade-off.§ Overall, our model offers a more interpretable, transparent, and computationally efficient approach for pricing American-style options, which has the potential to be widely accepted by regulators and applicable in a market-making environment. In addition to its practical implications, our approach also opens up new avenues for future research in the field of Artificial Intelligence (AI) in finance.

    §Speed-accuracy trade-off in terms of derivative pricing, hedging, and fitting has also been confirmed by Spiegeleer et al. (2018). Moreover, Glau et al. (2020) extend the tensorized Chebyshev interpolation of conditional expectations in the parameter and state space and show that, at comparable accuracy, the interpolation in American option prices provides a promising gain in computational efficiency, leading to a speed-accuracy trade-off.

    In the last few years, methods from AI have been successfully applied in the field of finance. As more and more black-box approaches are being adopted, the demand for transparency from regulators, investors, developers, and managers has been increasing. This paper presents a contribution by proposing a method for pricing American-style options that falls within the field of eXplainable AI (XAI). The neural network, often considered a black-box algorithm, is, in our approach, applied solely as a function approximator that learns from a training set generated by a chosen (e.g., PDE-based) well-known pricing model. We show that by doing this, our pricer behaves in an interpretable, tractable, and trustworthy fashion. As such, our approach can be understood as a machine learning method that produces an accelerated and explainable option pricing algorithm while maintaining a high level of pricing accuracy. Model explainability is one of the crucial aspects that would ensure investors and regulators with understanding and trust to use the proposed machine learning pricing model in a market-making environment.

    An extensive review of the literature around XAI is provided in Arrieta et al. (2020).

    The remainder of the paper is organized as follows: Section 2 introduces the computational model and its underlying framework. In Section 3, the numerical results are presented and the speed-accuracy trade-off is analyzed. Finally, we conclude in Section 4.

    We propose a method where American-style option prices are generated using a standard pricing method, and the price data is subsequently utilized for training a neural network over the underlying parameter space. Since the option prices are constructed with a well-known pricing method, the neural network learns the pricing map by utilizing the generated training set that is by construction free of arbitrage. Essentially, this neural network provides the capability to learn the behavior of the American option price to changes within its parameter space. By generating our own training data and training the neural network offline (i.e., not during trading hours), the latency in the pricing of American options during trading hours can be reduced to near-instantaneous. Furthermore, the same neural network could be used for other purposes such as calculating implied volatilities or calculating the Greeks (i.e., a set of derivatives of an option with respect to various parameters), see Larguinho et al. (2022), at near-instantaneous speeds.

    We consider a typical parameter set prevalent in a derivative pricing framework Ψ:=(r,q,Tt,M,ψModelσ), where r denotes the interest rate, q the continuous dividend yield, Tt the time-to-maturity, M the moneyness, and ψModelσ the set of variables parameterizing a chosen volatility specification. A standard Black-Scholes-Merton setting represents a trivial case of constant volatility with ψBSσ:=σ0, whereas under the Heston stochastic volatility model, the parameterization is denoted by ψHestonσ:={v0,σ,κ,θ,ρ}, where the parameter definitions are provided in the model description below.

    For the purpose of this study, the Heston stochastic volatility model was chosen as the basis for the proposed approach to pricing American-style options. This choice of a stochastic volatility model is motivated by the need to capture the dynamic nature of implied volatility, which is crucial for pricing options with path-dependent payoffs. Although there are many stochastic volatility models in the literature, the Heston model is widely studied and well-understood, making it a popular choice in practice. Additionally, the Heston model is capable of reproducing some of the stylized facts present in observed financial returns, such as volatility clustering and mean reversion, see Cont (2001). However, the Heston model does have some limitations in fitting the short-end of the implied volatility surface, which can be addressed by including jumps (Xie et al., 2021) or a local volatility component (Farkas et al., 2022). Nonetheless, the Heston model remains a good choice for theoretical tractability and comparability with other studies in the literature. It is important to note that our proposed method can be extended to any parametric volatility model, including Lévy-type volatility models (Schoutens, 2003; Glasserman and Kim, 2011; Barndorff-Nielsen and Stelzer, 2013; Ascione et al., 2023; He and Lin, 2023b; Lin and He, 2023), as long as the volatility model is parametric. Therefore, the choice of the Heston model is primarily driven by its popularity and theoretical tractability, and our proposed method can be applied to a broad range of volatility models.

    By computing American option prices using, for example, a PDE method on a dense grid across the parameter set Ψ, we generate a training set over which a neural network pricing functional is trained. As standard methods are used offline to generate the training data for the neural network, a major component of this paper is the accurate generation of the training data. Due to various universal approximation theorems (discussed below), it can be assumed that the neural network is able to approximate the option pricing dynamics to, at best, the precision of the underlying standard model-generated data. Thus, the error of the standard model data serves as a lower bound to the neural network error, illustrating the requirement for accurate standard model input data.

    Definition 2.1 (Heston Stochastic Volatility Model). Consider a probability space {Ω,F,Q} adapted to a filtration Ft. Let St and vt denote the asset price and instantaneous variance at time t. Then, according to the Heston model, as presented in Heston (1993), the asset price and volatility processes are, under the pricing measure Q, defined by a system of stochastic differential equations (SDEs)

    dSt=(rq)Stdt+vtStdWt,dvt=κ[θvt]dt+σvtdZt,dZtdWt=ρdt, (1)

    where κ>0 denotes the speed of mean reversion, θ>0 the long term variance of the price process, σ>0 the volatility of volatility, 1ρ1 the correlation coefficient between the geometric Brownian motion processes Zt and Wt, and S0,v0>0 are the starting values of the stochastic processes St and vt, respectively. In order that the instantaneous variance vt>0 is strictly positive, the parameters need to obey the Feller condition 2κθ>σ2.

    The value of an American call option with maturity T and strike K, is, at time t=0, given by

    V(S0,0)=supτT(T)EQ[erτ(SτK)+], (2)

    where T(T) denotes the set of stopping times τ with values in [0,T]. Assuming the system of SDEs from Equation (1), one can derive the solution to Equation (2) in terms of a PDE. The resulting Heston American call option PDE (expressed in terms of the option price) is given by

    Vt+12vS22VS2+ρσvS2VvS+12σ2v2Vv2rV+(rq)SVS+[κ(θv)λ(S,v,t)]Vv=0, (3)

    in the region D={0tT,0Stb(v,t),0v<}, subject to the boundary conditions

    V(S,v,T)=(SK)+,V(b(v,t),v,t)=b(v,t)K,limSb(v,t)VS=1,limSb(v,t)Vv=0,

    where b(v,t) represents the early exercise boundary at time t for the Heston volatility v, and λ(S,v,t) denotes the corresponding market price of volatility risk. See Heston (1993) and AitSahlia et al. (2010) for a detailed proof and more information.

    Algorithm 1 Neural network-based American option pricer under stochastic volatility
    Input: Chosen parameter ranges for r, q, v0, σ, κ, θ, ρ, M, and Tt.
    Output: Current price of an American-style option per underlying applicable in a market-making environment.
    (A) TRAINING DATA GENERATION: Evaluation of American option prices using a chosen standard pricing method across an entire parameter space Ψ.
    (B) LEARNING: Offline training of a neural network over a parameter-price data set.
    (C) PRICING: Application of a neural network pricer in a market-making environment.

     | Show Table
    DownLoad: CSV

    In Algorithm 1, the proposed American option pricing methodology is summarized in a three-step process that generates a direct explainable neural network pricer. In the case when the training data is generated with a standard PDE method, Step (A) reduces to solving the Heston PDE from Equation (3) using a finite difference method. The parameter ranges over which the learning takes place need to be chosen in this step. In our case, these are the Heston model parameters whereas the same algorithm works for any other parametric specification. We detail the chosen parameter ranges (assuring that the Feller condition is satisfied) in Section 3.1. Note that the chosen parameter ranges largely depend on the available computational capabilities. In Step (B), offline training takes place. One needs to choose the structure of the neural network and the corresponding hyperparameters in this step of the algorithm. We detail the architecture of our chosen neural network in Section 2.2 and provide the analysis of the training process in Section 3.1. The deep neural network here is utilized merely as a tool to learn the pricing map between the chosen standard (e.g., PDE-based) pricing method in Step (A) and the corresponding American option prices. Such an approach enhances the trust and transparency of our algorithm. Even though a neural network is applied, the algorithm itself is not a black box. As shown in the next section, our algorithm simply accelerates the computation of American option prices obtained by a well-understood pricing approach chosen in Step (A) of the algorithm. Finally, after the offline learning process from Step (B) is completed, the deep explainable (i.e., Heston-based) neural network pricer can be employed in a market-making environment in Step (C) of the algorithm. Given our aim for computational explainability, the proposed algorithm has a large potential of being understandable and trusted by humans, and as such accepted by regulators and investors.

    Since this neural network is targeted for market making, a key responsibility of any algorithm in a trading environment is its compliance with regulation such as the Markets in Financial Instruments Directive (MiFID) or Securities and Exchange Commission (SEC) guidelines. For example, MiFID specifically states that banks employing algorithms in trading should be able to "reconstruct efficiently and evaluate the strategies that algorithmic traders employ", see Directive-65/EU (2014). This implies a requirement that black-box pricing algorithms such as neural networks need to have an explainable element. As such, we find it important to stress that our XAI-based pricing method indeed fulfills this requirement since we can show, see Section 3.2, that the error is bounded on the underlying (i.e., Heston) parameter space.

    Note that one can use any underlying data-generating process as well as any American option pricing method for the training data generation since the calibration of the deep neural network works independently of the training price generation. As such, we provide a possible extension of the algorithm proposed above to pricing under the rough Bergomi volatility specification in Appendix 5.

    Rough Bergomi model improves upon the classical stochastic volatility counterparts by reflecting the implied volatility surface dynamics more accurately (e.g., especially for short maturities). Under the rough Bergomi model, derivative pricing must be performed via Monte Carlo simulations. Thus, the computation time could potentially be greatly reduced using our proposed accelerated pricing method.

    To improve upon the computationally slower standard American option pricing algorithms, we apply neural networks to serve as the functional evaluators–essentially substituting the underlying numerical pricing algorithm. This is primarily due to the ability of neural networks to model complex non-linear dynamics, differentiability, fast execution, and the ability to be evaluated using very basic low-level functionals. Furthermore, given the universal approximation properties of neural networks, there exists a neural network (trained upon a chosen basic pricing method) that is able to price American-style options to any desired accuracy up to the pricing error imposed by the generated training data.

    We aim to improve the process of American option pricing with respect to the speed-accuracy trade-off compared to the standard American option pricers. The motivation for improving the speed-accuracy trade-off lies in the ability of the neural network to learn the American option pricing map offline. Because of the simplicity of the neural network representation, their evaluation speed is unparalleled to the speed of other option pricing models. The boost in speed of the computation and the algorithm interpretability are the main arguments granting our approach a suitable candidate improving upon the currently used market-making pricing systems.

    We exploit a feed-forward neural network with a rectified linear unit (ReLU) activation function merely as a tool to map from a given parameter space to the American option prices. The near-instantaneous evaluation of neural networks and their ability to approximate continuous functions provide for the above-mentioned speed advantages over standard pricing methods. Neural networks can evaluate option prices in less than milliseconds, whereas PDE methods, binomial methods, or MC methods have significantly slower evaluation speeds. The deep neural network we employ is structured as a generic multi-layer perceptron as described in Caterini and Chang (2018).

    Definition 2.2 (Multi-Layer Perceptron). Given a data set with an input dimension of k and an output dimension of m, a multi-layer perceptron is defined as the neural network with an input layer of dimension n0:=k and an output layer of dimension nL+1:=m, where L specifies the number of hidden layers between the input and the output layer. Analogously, each layer between layers 0 and L+1 is of dimension ni. For each layer i, we denote the weight and bias matrices by WiRni+1×ni and biRni+1, respectively. Per layer, the following function is applied to provide the input to the subsequent layer: gi(xi;Wi,bi)=γi(Wixi+bi), where xiRni+1 and γi denotes the corresponding activation function of the layer.

    Some of the fundamental theorems of deep neural networks postulate that under a specific set of criteria, neural networks can approximate continuous functions to any degree of accuracy. These theorems are known as universal approximation theorems. Some of them focus on the depth (i.e., a number of hidden layers) of a neural network, and others on the width (i.e., the number of nodes per hidden layer). A widely known universal approximation theorem from Hornik et al. (1989) specifies that a neural network can approximate any continuous real-valued function on a compact set to an arbitrary accuracy ϵ>0 provided certain criteria for activation functions and depth hold. This is known as the depth-bounded universal approximation theorem. Extensions of this universal approximation theorem for ReLU activation functions are provided in Stinchcombe and White (1989). Moreover, Lu et al. (2017) propose a specific universal approximation function that provides an upper bound to the width of a neural network with ReLU activation functions. This result is relevant for this paper since we use the ReLU activation functions on the neural network's hidden layers. The corresponding theorem is stated below.

    Theorem 2.1 (Width-Bounded Universal Approximation Theorem for ReLU Networks). For any Lebesgue-integrable function f: RnR and any ϵ0, there exists a fully-connected ReLU network G with width dLn+4, such that the neural network function FG approximating f satisfies:

    Rn|f(x)FG|dx<ϵ. (4)

    Our choice of the ReLU activation function relies on some of its special properties. ReLU activation functions are special since for x<0, the activation function does not fire**—causing a sparser neural network. Furthermore, as x becomes large, the neurons do not become saturated†† as in popular sigmoid activation functions. Lastly, ReLU functions are known to train faster compared to other activation functions, as shown by Krizhevsky et al. (2017). One problem with ReLU activation functions noted by Caterini and Chang (2018) is that neurons can easily die, becoming what are known as "dead neurons" or neurons that always give a zero weight—a downside of allowing a sparse structure. This problem can be fixed by using a leaky ReLU activation function defined by

    γ(x):={x,x>00.01x,x0}, (5)

    **The so-called 'firing' of a node in a neural network is etymologically derived from the biological definition of neurons firing, i.e., sending a non-null impulse when an input is received.

    ††Saturated means that for a very large positive or negative number, the gradient of the sigmoid activation function is near zero.

    which returns a small negative value for x<0, allowing the neurons to recover.

    Following standard literature, we utilize the ADAM optimizer‡‡ for the training with the Mean Absolute Error (MAE) cost function:

    MAE:=1nΨ|VNN(ψk)VTS(ψk)|, (6)

    ‡‡ADAM optimizer offers many advantages compared to the typical stochastic gradient descent, see Kingma and Ba (2014) for more details.

    where n is the number of price samples, ψkΨ is a particular parameter specification, VNN(ψk) is the output of the neural network at each training iteration for a respective parameter combination ψk, and VTS(ψk) is the training data set for a respective parameter combination ψk. We decided to train with the MAE cost function because using an alternative cost function, for example, the widely-used mean squared error (MSE), would exaggerate the contribution induced by outliers by squaring their error in the cost function. This is relevant in our case since prices can vary widely within the parameter ranges (e.g., from 0.5 USD to 30 USD) and, consequently, the error varies as well.§§

    §§We also performed training with a Vega-Weighted Absolute Error (VWAE) cost function as presented in Hamida and Cont (2005). Thereby, the vega weight "converts" the errors in price into errors in implied volatility. The results, however, did not improve compared to the use of the conventional MAE cost function.

    The training of the neural network consists of the determination of the hyperparameters such as the learning rate, exponential learning rate adjustment (ELRA), batch size, epochs, and depth and width of the hidden layers. The final neural network structure is set to four hidden layers with 40 nodes per layer. We follow Horvath et al. (2021) and Liu et al. (2019b) in that we decide to use a neural network with four hidden layers ensuring that the depth requirement for the universal approximation theorems is reached.

    We generate the training sample (TS) parameter set ΨTS by choosing 12 equally spaced points for the parameter ranges M[90,110], Tt[1,255], κ(0,5], ρ[1,0], and σ(0,1], and 12 logarithmically spaced points for θ(0,0.41] and v0(0,0.41], since the long-term variance and the initial variance levels prevalently undertake lower (i.e., less than 50%) values.¶¶ Using this parameter set ΨTS as an input to the finite difference PDE method, we generate the training set on a very fine mesh of 400 points in time, 400 points in price space, and 100 points in volatility space (i.e., 400×400×100). This results in 127 price samples prior to the application of the Feller condition. Following the application of the Feller condition, we are left with a training set consisting of 24,634,368 parameter-price pairs. Moreover, our training set parameter ranges are in line with the ones studied in the existing literature, see Spiegeleer et al. (2018).

    ¶¶Note that r and q have been fixed to be 1% and 3%, respectively. We set r and q to be constant since they are parameters that should not change on an hour-by-hour basis. If necessary, the neural network can be retrained when necessary.

    To prepare the data for training, we standardize the input parameter data ΨTS as is a typically required practice in the training of neural networks since different input data have different absolute scales. We apply the same standardization (i.e., same mean and variance) to the parameter sets for all validation and test input data. Then, by applying the ADAM optimizer on the MAE cost function with the leaky ReLU activation function as described above, we train the neural network using an exponential learning rate adjustment. The training stops when the validation error does not improve for another 1000 iterations after reaching a running minimum. Naturally, during training, the difference between the validation and the training error is closely observed in order to prevent over-fitting to the input data.

    Due to instabilities in the ADAM optimizer, one typically applies an exponential learning rate adjustment that exponentially decreases the learning rate during the training session. This allows the optimizer to initially explore the cost function surface for the minima, and subsequently, by slowly reducing the learning rate, the likelihood that the optimizer jumps out of the minimum is reduced. In Figure 1, the effect of this adjustment for a sample training session is illustrated.

    Figure 1.  Training (in blue) and validation error (in red) per ADAM iteration with (upper subplot) and without (lower subplot) the exponential learning rate adjustment. The same pattern continues also after the 7000th iteration.

    After 23,000 iterations over the entire training set, the stopping rule was satisfied and the training was stopped. The applied neural network structure is presented in Table 1.

    Table 1.  This table presents the chosen neural network hyperparameters.
    Neural Network Structure:
    n 24,634,368
    Batch Size 1,231,719
    Initial Learning Rate 0.5
    ELRA 0.997
    Hidden Layers (Depth) 4
    Nodes per Layer (Width) 40
    Parameters to Train 5281
    Epoch Size 23,000
    Training Time 41 hours

     | Show Table
    DownLoad: CSV

    Having trained the neural network offline, we test its out-of-sample performance in this section. We generate the out-of-sample (OOS) parameter set ΨOOS by using the finite difference PDE pricer with the same mesh size as the one used for the training set, 400×400×100. Note that the OOS parameter set ΨOOS is independent of the training sample parameter set ΨTS so that the error bias is avoided. Using ΨOOS, we evaluate the neural network-based prices VNN and the corresponding near-exact (i.e., PDE-based) VPDE prices. By "exact" we imply that a highly fine-meshed finite difference method was applied to construct the option prices. In Table 2, the neural network prices are compared to the near-exact PDE prices in terms of mean absolute error (MAE), mean absolute percentage error (MAPE), and absolute percentage error (APE).

    Table 2.  Model Summary under the Heston Stochastic Volatility Model.
    Heston Stochastic Volatility Case:
    OOS n 14336 FD Grid Density 400×400×100
    NN OOS MAE 0.02 USD NN OOS MAPE 0.26%
    NN OOS SD of Error 0.014 USD Evaluation Time for OOS n 0.000026 sec.
    NN OOS Min APE 0.0000236% NN OOS Max APE 3.26%

     | Show Table
    DownLoad: CSV

    Now that we have an overview of the average performance of the neural network pricer, we take a closer look at how it performs for each area of the parameter space Ψ. Consider a regular pricing environment with parameter values of v0=5%, θ=15%, σ=0.3, κ=0.8, ρ=0.5, Tt=165 (i.e., in days), and M=100%. In Figure 2, we vary one parameter at a time over its respective parameter space and plot the resulting neural network price in comparison to the near-exact price as computed by the PDE method. Such an analysis demonstrates the explainability nature of our pricing algorithm.

    Figure 2.  The left part of the figure depicts a comparison between the prices computed by the neural network (NN) under the Heston stochastic volatility—in blue—and the training sample computed by the finite difference method (FDM)—in red—in dependence of different model parameters. The right part of the figure equivalently shows the relative error of the NN approach in comparison to the FDM.

    Notice that the neural network is able to learn the pricing function under the Heston stochastic volatility specification. Moreover, the well-studied Heston parametric structure is inherited by the deep explainable pricer with overall low relative pricing errors. Being able to provide such parametric interpretations positions our model in the field of XAI. Nevertheless, the neural network underperforms in some specific areas of the parameter space. Notice that, for example, ρ>0.4, M<95%, Tt<20, v0<0.05, and σ all exhibit above-average errors. One can tackle this issue by simulating more parameter-price pairs, especially in the under-performing parameter regions, and consequently improve the accuracy of the NN-based pricer. Moreover, one can also experiment with different cost functions or even train multiple neural networks for different parameter regions (e.g., one NN in the short time to maturity region and other for the rest). Given the stochastic nature of the learning algorithm, a certain level of pricing error is inevitable. However, as shown in Figure 2, the NN-based pricer learns the American option pricing map under the Heston stochastic volatility model with a low relative error and, above all, performs pricing in an explainable and interpretable fashion. Given the increasing demand for transparency from investors and regulators, this is an influential finding facilitating the potential compliance of the proposed pricing algorithm with regulatory guidelines.

    Figure 3 shows the implied volatility surface as computed by the deep explainable pricer and its PDE-based benchmark. One can observe that the deep explainable pricer reproduces the shape of the implied volatility surface by training from the data generated by the PDE-based pricer. The right subplot presents the error calculated as the value of the implied volatility as computed by the deep explainable pricer minus the value computed by the PDE-based pricer. Note that the errors are generally very low. The only area with larger deviations are the points close to maturity.

    Figure 3.  The left subplot depicts the implied volatility surface as computed by the deep explainable pricer. The middle subplot shows the implied volatility surface as computed by the benchmark PDE-based pricer. The right subplot presents the error in the implied volatility surface of the neural network pricer computed as the difference between the deep explainable and PDE-based implied volatility surfaces.

    The main incentive for applying neural networks in pricing American options are the benefits in terms of pricing speed. Next, we perform an analysis comparing the speed of evaluation of the proposed neural network pricer (over a set of 14,366 input parameter sets) with the speeds of standard methods such as the finite difference method, Monte Carlo simulation of Longstaff and Schwartz (2001), and binomial method of Cox et al. (1979) in Table 3. The computations were performed on an Intel Xeon Gold 6126 2.60GHz CPU.

    Table 3.  Evaluation speed of the neural network pricer in comparison to other pricing methods for Heston stochastic volatility.
    Pricing Method Average Evaluation Speed (secs) Compared to Neural Network Speed
    Neural Network Method 0.000026
    Finite Difference PDE Method 0.61 23,461 times slower
    Binomial Method (Constant Volatility) 7.3 280,769 times slower
    Longstaff-Schwartz Method 139.2 53,538,461 times slower

     | Show Table
    DownLoad: CSV

    The PDE method is implemented as proposed by Ikonen and Toivanen (2007) for an option of T=0.25. Naturally, the methods' speed differs given different maturities, step sizes, iterations, and a number of paths. Nevertheless, no other method is able to evaluate an American option price nearly as fast as the neural network pricer.

    To illustrate the speed-accuracy trade-off, we choose the FDM method (i.e., the fastest competitor of the NN) and reduce the fineness of its grid such that, on average, the mean absolute error of the neural network is matched with the FDM-based error. By comparing the two methods at the same error level, we analyze the practicality of both pricing approaches in a market-making environment where both speed and accuracy are of utmost importance. Table 4 shows that at the same level of accuracy, the neural network pricer exhibits a considerable speed improvement over the FDM-based pricing method. In particular, the FDM method is, at the same level of accuracy, 58 times slower compared to the deep explainable American option pricer proposed in this paper.

    Table 4.  Speed-accuracy trade-off of the neural network pricer in comparison to the FDM pricing method for Heston stochastic volatility.
    Model: Error Average Speed(secs)
    Neural Network 26bps 0.000026
    FDM (26×26) 26bps 0.0015

     | Show Table
    DownLoad: CSV

    Figure 4 depicts the overall speed-accuracy spectrum for constant volatility (CV) and Heston stochastic volatility specifications compared to the deep explainable pricer. Speed is measured in terms of the number of pricings per second and accuracy is given in terms of relative error expressed in basis points against a "true" value defined as the PDE method (for CV or Heston) with a very fine mesh grid. One can observe the outperformance of the neural network with respect to the alternative pricing methods. The deep explainable pricer not only significantly outperforms the PDE Heston benchmark, but also outperforms PDE, Longstaff-Schwartz, and binomial methods under constant volatility. Moreover, with an enhanced training set and longer training times, further improvements in error could be achieved–shifting the neural network point far to the left since the evaluation speed of a neural network is approximately constant. These results show that the proposed deep explainable pricer achieves a reasonable accuracy of computation while greatly reducing the speed of the computation. As such, it serves as a speedy alternative to more classical American option pricing methods currently used in a market-making environment.

    Figure 4.  Speed-accuracy trade-off for various American option pricing methods in both constant volatility (CV) and Heston stochastic volatility environments compared to our neural network (NN) pricer.

    Market-making in the equity derivatives markets requires high-precision and fast pricing algorithms. If the prices are not adjusted accurately and instantaneously, arbitrage opportunities occur. This paper presents an American-style option pricing method based on a deep neural network that accelerates the pricing while maintaining sufficient accuracy for market-making applicability.

    The proposed model is based on generating training data by using a standard American option pricing method. We utilize a PDE-based method and price under the Heston stochastic volatility model in order to construct the training data. The obtained training data is subsequently utilized to train a deep neural network over the underlying parameter space offline. Consequently, the underlying (i.e., Heston) parameter interpretability is inherited by the neural network pricer. Such interpretability improves the trust and transparency of the proposed deep explainable pricer, significantly increases the speed of American-style option pricing, and associates our algorithm with the area of XAI for finance.

    The simulation analysis shows that our approach provides a substantial improvement with respect to the speed-accuracy trade-off over standard American option pricers. The improvement primarily arises due to the ability of neural networks to learn the American option pricing map offline. Moreover, given the simplicity of neural network representations, their evaluation speed is considerably higher compared to the speed of standard option pricing models. This speed increase is naturally more pronounced for Monte Carlo methods than for PDE methods due to the complexity of simulation methods. Also, we show that this increase in speed occurs without a significant loss in accuracy when methods are compared at the same precision level. Therefore, our model exhibits the speed-accuracy and explainability features required in a challenging market-making environment.

    Furthermore, potentially increased computing resources and the universal approximation properties of neural networks imply that the level of accuracy presented in this paper could be improved by ⅰ) increasing the number of simulated parameter-price data points over which the training is performed, ⅱ) experimenting with alternative neural network architectures, ⅲ) further optimizing the neural network hyperparameters, ⅳ) experimenting with alternative cost functions, or ⅴ) training multiple neural networks at different parameter regions.

    While these steps could potentially improve the accuracy of the neural network pricer, other interesting areas of future research could expand the presented deep explainable pricing framework. Particularly, since the speed improvements are greatest when compared to Monte Carlo methods, any pricing model that relies purely on MC methods would be a perfect candidate for neural network acceleration. One could, for example, expand the model to utilize large-deviation MC techniques for out-of-the-money and short-maturity options. Moreover, one could also consider alternative dynamics for the underlying asset and volatility processes, such as the rough volatility, possibly including cash dividends and more complex payoffs that are priceable only by the MC techniques.

    The authors are grateful to Jérôme Detemple, Walter Farkas, Damir Filipović, Isabella Kooij, Markus Leippold, Álvaro Leitao Rodríguez, Dan Lupu, Ana Mão de Ferro, Leander Gayda, Pit Götz, Ruggero Jappelli, Alexis Marchal, Ludovic Mathys, Michael Mi, Angelina Steffens, Nikola Vasiljević, Rudi Zagst, Žan Žurič, and the participants of the SFI Research Days 2021, the 7th International Young Finance Scholars Conference (2021), the Society of Financial Econometrics Summer School 2021, the 34th Australasian Finance and Banking Conference (2021), the XIX International Conference on Finance and Banking (2022), the 4th International Conference on Computational Finance 2022, and the International Risk Management Conference 2022 for helpful comments and suggestions.

    The authors declare no conflict of interest.

    Definition A.1 (Rough Bergomi Volatility Model). Let St and vt denote the asset price and instantaneous variance at time t. The rough Bergomi model, introduced by Gatheral et al. (2018), is, under a pricing measure, defined by a system of stochastic differential equations

    dStSt=(rq)dt+vt(ρdWt+1ρ2dWt),vt=ξ0(t)E(ηWα)t,

    where r is the interest rate, q the is continuous dividend yield, Wt and Wt are independent Brownian motions, 1ρ1, and the instantaneous variance vt is a product of a forward variance curve tξ0(t), known at time 0, and the Wick exponential E(ηWα)t=exp(ηWαt12Var(ηWαt)) for η>0 and a Gaussian random variable Wαt. The random variable Wαt is given by the Gaussian Riemann–Liouville process

    Wαt=2α+1t0(ts)αdWs,t0,

    where the parameter 12<α<0 controls the roughness of paths. The corresponding paths have a Hölder regularity of α+12 and locally behave like the paths of a fractional Brownian motion with H=α+12. The three time-homogeneous parameters, η, ρ, and α, can be interpreted as the smile, the skew, and the near-maturity explosion of the smile and the skew, expressed in terms of the implied volatility surface.

    In the case of rough Bergomi volatility, a Monte Carlo simulation is required to generate the initial training data needed for the neural network calibration. This means that the gains in speed due to training a neural network price evaluator are potentially greater than in the case of the Heston stochastic volatility model.

    In Algorithm 2, a neural network pricer for the case of Monte Carlo generated training data set is presented. In Step (A), the Euler-Maruyama method, see Maruyama (1955), is applied to simulate paths of the volatility and price processes for a fine mesh across the parameter set Ψ. On each path, optimal exercise opportunities are evaluated, and a corresponding payoff for each path is determined.

    For the evaluation of optimal exercise opportunities in Step (B), the methodology presented in Maruyama (1955) can be utilized. Their method of approximating the early exercise boundary allows for the computation of an early exercise boundary value for every point along the MC paths. The optimal exercise strategy is calculated by comparing the current simulated spot price level with the Andersen boundary value. By discounting and averaging over the payoffs of the simulated paths for each parameter combination, a Monte Carlo estimator of the American option price is obtained. Finally, by repeating this procedure for all combinations of parameters across Ψ, one arrives at a set of American option prices across the entire parameter space – Step (C). In Step (D), the offline training occurs, and the deep neural network rough Bergomi-based pricer can potentially be employed in a market-making environment in Step (E).

    Algorithm 2 Neural network-based American option pricer under rough volatility
    Input: ξ0, η, α, ρ, M, Tt.
    Output: Current price of an American-style option per underlying applicable in a market-making environment.
    (A) SIMULATION: Simulation of the volatility and price processes.
    (B) OPTIMAL EXERCISE: Evaluation of optimal exercise strategies for each simulated path.
    (C) TRAINING DATA GENERATION: Evaluation of American option prices across an entire parameter space Ψ.
    (D) LEARNING: Offline training of a neural network over a parameter-price data set.
    (E) PRICING: Application of a neural network pricer in a market-making environment.

     | Show Table
    DownLoad: CSV

    As mentioned in Andersen et al. (2016), under rough volatility models, the optimal exercise strategy should depend on past observations of the spot price and not only the current price. However, the authors concluded that there is no significant difference in the computed price due to including past spot prices in the exercise decision. Based on this reasoning, one could assume that the current spot price contains sufficient information to decide upon the American option exercise. However, further investigation is necessary.

    [1] da Silva HB, Amaral EP, Nolasco EL, et al. (2013) Dissecting major signaling pathways throughout the development of prostate cancer. Prostate Cancer 2013: 920612.
    [2] Chang H, Brown CW, Matzuk MM (2002) Genetic analysis of the mammalian transforming growth factor-β superfamily. Endocr Rev 23: 787-823. doi: 10.1210/er.2002-0003
    [3] Massagué J (2008) TGFβ in cancer. Cell 134: 215-230. doi: 10.1016/j.cell.2008.07.001
    [4] Zhang YE (2009) Non-Smad pathways in TGF-β signaling. Cell Res 19: 128-139. doi: 10.1038/cr.2008.328
    [5] Heldin C-H, Moustakas A (2016) Signaling Receptors for TGF-β Family Members. Cold Spring Harb Perspect Biol 8: a022053. doi: 10.1101/cshperspect.a022053
    [6] Morikawa M, Derynck R, Miyazono K (2016) TGF-β and the TGF-β Family: Context-Dependent Roles in Cell and Tissue Physiology. Cold Spring Harb Perspect Biol 8: a021873. doi: 10.1101/cshperspect.a021873
    [7] Jenkins G (2008) The role of proteases in transforming growth factor-beta activation. Int J Biochem Cell Biol 40: 1068-1078. doi: 10.1016/j.biocel.2007.11.026
    [8] Johnson KE, Makanji Y, Temple-Smith P, et al. (2016) Biological activity and in vivo half-life of pro-activin A in male rats. Mol Cell Endocrinol 422: 84-92. doi: 10.1016/j.mce.2015.12.007
    [9] Walton KL, Kelly EK, Chan KL, et al. (2015) Inhibin biosynthesis and activity are limited by a prodomain-derived peptide. Endocrinology 156: 3047-3057. doi: 10.1210/en.2014-2005
    [10] Walton KL, Makanji Y, Harrison CA (2012) New insights into the mechanisms of activin action and inhibition. Mol Cell Endocrinol 359: 2-12. doi: 10.1016/j.mce.2011.06.030
    [11] De Caestecker M (2004) The transforming growth factor-β superfamily of receptors. Cytokine Growth Factor Rev 15: 1-11. doi: 10.1016/j.cytogfr.2003.10.004
    [12] Brodin G, ten Dijke P, Funa K, et al. (1999) Increased smad expression and activation are associated with apoptosis in normal and malignant prostate after castration. Cancer Res 59: 2731-2738.
    [13] Shi Y, Massagué J (2003) Mechanisms of TGF-β signaling from cell membrane to the nucleus. Cell 113: 685-700. doi: 10.1016/S0092-8674(03)00432-X
    [14] Sun PD, Davies DR (1995) The cystine-knot growth-factor superfamily. Annu Rev Biophys Biomol Struct 24: 269-291. doi: 10.1146/annurev.bb.24.060195.001413
    [15] Lebrun J-J (2012) The dual role of TGF in human cancer: from tumor suppression to cancer metastasis. ISRN Mol Biol 2012.
    [16] Bierie B, Moses HL (2006) Tumour microenvironment: TGFβ: the molecular Jekyll and Hyde of cancer. Nat Rev Cancer 6: 506-520. doi: 10.1038/nrc1926
    [17] Xie L, Law BK, Chytil AM, et al. (2004) Activation of the Erk pathway is required for TGF-β1-induced EMT in vitro. Neoplasia 6: 603-610. doi: 10.1593/neo.04241
    [18] Beck C, Schreiber H, Rowley DA (2001) Role of TGF-β in immune-evasion of cancer. Microsc Res Tech 52: 387-395.
    [19] Margheri F, Schiavone N, Papucci L, et al. (2012) GDF5 regulates TGFß-dependent angiogenesis in breast carcinoma MCF-7 cells: in vitro and in vivo control by anti-TGFß peptides. PLoS One 7: e50342. doi: 10.1371/journal.pone.0050342
    [20] Futakuchi M, Nannuru KC, Varney ML, et al. (2009) Transforming growth factor-β signaling at the tumor-bone interface promotes mammary tumor growth and osteoclast activation. Cancer Sci 100: 71-81. doi: 10.1111/j.1349-7006.2008.01012.x
    [21] Sugatani T, Alvarez U, Hruska K (2003) Activin A stimulates IκB‐α/NFκB and rank expression for osteoclast differentiation, but not AKT survival pathway in osteoclast precursors. J Cell Biochem 90: 59-67. doi: 10.1002/jcb.10613
    [22] Gupta DK, Singh N, Sahu DK (2014) TGF-β Mediated Crosstalk Between Malignant Hepatocyte and Tumor Microenvironment in Hepatocellular Carcinoma. Cancer Growth Metastasis 7: 1.
    [23] Marino FE, Risbridger G, Gold E (2013) The therapeutic potential of blocking the activin signalling pathway. Cytokine Growth Factor Rev 24: 477-484. doi: 10.1016/j.cytogfr.2013.04.006
    [24] Siegel PM, Massagué J (2003) Cytostatic and apoptotic actions of TGF-β in homeostasis and cancer. Nat Rev Cancer 3: 807-820. doi: 10.1038/nrc1208
    [25] Irigoyen J, Munoz-Canoves P, Montero L, et al. (1999) The plasminogen activator system: biology and regulation. Cell Mol Life Sci 56: 104-132. doi: 10.1007/PL00000615
    [26] Li WY, Chong SS, Huang EY, et al. (2003) Plasminogen activator/plasmin system: a major player in wound healing? Wound Repair Regen 11: 239-247. doi: 10.1046/j.1524-475X.2003.11402.x
    [27] Pulukuri SM, Gondi CS, Lakka SS, et al. (2005) RNA interference-directed knockdown of urokinase plasminogen activator and urokinase plasminogen activator receptor inhibits prostate cancer cell invasion, survival, and tumorigenicity in vivo. J Biol Chem 280: 36529-36540. doi: 10.1074/jbc.M503111200
    [28] Hartwell KA, Muir B, Reinhardt F, et al. (2006) The Spemann organizer gene, Goosecoid, promotes tumor metastasis. Proc Natl Acad Sci U S A 103: 18969-18974. doi: 10.1073/pnas.0608636103
    [29] Padua D, Massagué J (2009) Roles of TGFβ in metastasis. Cell Res 19: 89-102. doi: 10.1038/cr.2008.316
    [30] Mundy GR (2002) Metastasis: Metastasis to bone: causes, consequences and therapeutic opportunities. Nat Rev Cancer 2: 584-593. doi: 10.1038/nrc867
    [31] Logothetis CJ, Lin S-H (2005) Osteoblasts in prostate cancer metastasis to bone. Nat Rev Cancer 5: 21-28. doi: 10.1038/nrc1528
    [32] Chen G, Deng C, Li Y-P (2012) TGF-beta and BMP signaling in osteoblast differentiation and bone formation. Int J Biol Sci 8: 272-288. doi: 10.7150/ijbs.2929
    [33] Tsuchida K, Nakatani M, Hitachi K, et al. (2009) Activin signaling as an emerging target for therapeutic interventions. Cell Commun Signal 7: 10.1186. doi: 10.1186/1478-811X-7-10
    [34] Buijs J, Van Der Horst G, Van Den Hoogen C, et al. (2012) The BMP2/7 heterodimer inhibits the human breast cancer stem cell subpopulation and bone metastases formation. Oncogene 31: 2164-2174. doi: 10.1038/onc.2011.400
    [35] Schnepp B, Hua X (2003) A Tumor Suppressing Duo: TGFβ and Activin Modulate a Similar Transcriptome. Cancer Biol Ther 2: 171-172. doi: 10.4161/cbt.2.2.334
    [36] Mittl PR, Priestle JP, Cox DA, et al. (1996) The crystal structure of TGF‐β3 and comparison to TGF‐β2: implications for receptor binding. Protein Science 5: 1261-1271. doi: 10.1002/pro.5560050705
    [37] Gatherer D, Ten Dijke P, Baird DT, et al. (1990) Expression of TGF-isoforms during first trimester human embryogenesis. Development 110: 445-460.
    [38] Millan FA, Denhez F, Kondaiah P, et al. (1991) Embryonic gene expression patterns of TGF beta 1, beta 2 and beta 3 suggest different developmental functions in vivo. Development 111: 131-143.
    [39] Pelton RW, Saxena B, Jones M, et al. (1991) Immunohistochemical localization of TGF beta 1, TGF beta 2, and TGF beta 3 in the mouse embryo: expression patterns suggest multiple roles during embryonic development. J Cell Biol 115: 1091-1105. doi: 10.1083/jcb.115.4.1091
    [40] López-Casillas F, Wrana JL, Massagué J (1993) Betaglycan presents ligand to the TGFβ signaling receptor. Cell 73: 1435-1444. doi: 10.1016/0092-8674(93)90368-Z
    [41] Piek E, Heldin C-H, Ten Dijke P (1999) Specificity, diversity, and regulation in TGF-β superfamily signaling. FASEB J 13: 2105-2124.
    [42] David D, Nair SA, Pillai MR (2013) Smurf E3 ubiquitin ligases at the cross roads of oncogenesis and tumor suppression. Biochim Biophys Acta 1835: 119-128.
    [43] Robertson IB, Horiguchi M, Zilberberg L, et al. (2015) Latent TGF-β-binding proteins. Matrix Biology 47: 44-53. doi: 10.1016/j.matbio.2015.05.005
    [44] Butz H, Rácz K, Hunyady L, et al. (2012) Crosstalk between TGF-β signaling and the microRNA machinery. Trends Pharmacol Sci 33: 382-393. doi: 10.1016/j.tips.2012.04.003
    [45] Zhong H, Wang H-R, Yang S, et al. (2010) Targeting Smad4 links microRNA-146a to the TGF-β pathway during retinoid acid induction in acute promyelocytic leukemia cell line. Int J Hematol 92: 129-135. doi: 10.1007/s12185-010-0626-5
    [46] Ottley E, Gold E (2014) microRNA and non-canonical TGF-β signalling: Implications for prostate cancer therapy. Crit Rev Oncol Hematol 92: 49-60. doi: 10.1016/j.critrevonc.2014.05.011
    [47] van den Bosch MH, Blom AB, van Lent PL, et al. (2014) Canonical Wnt signaling skews TGF-β signaling in chondrocytes towards signaling via ALK1 and Smad 1/5/8. Cell Signal 26: 951-958. doi: 10.1016/j.cellsig.2014.01.021
    [48] Grönroos E, Kingston IJ, Ramachandran A, et al. (2012) Transforming growth factor β inhibits bone morphogenetic protein-induced transcription through novel phosphorylated Smad1/5-Smad3 complexes. Mol Cell Biol 32: 2904-2916. doi: 10.1128/MCB.00231-12
    [49] Liu IM, Schilling SH, Knouse KA, et al. (2009) TGFβ‐stimulated Smad1/5 phosphorylation requires the ALK5 L45 loop and mediates the pro‐migratory TGFβ switch. EMBO J 28: 88-98. doi: 10.1038/emboj.2008.266
    [50] Perry KT, Anthony CT, Steiner MS (1997) Immunohistochemical localization of TGFβ1, TGFβ2, and TGFβ3 in normal and malignant human prostate. Prostate 33: 133-140.
    [51] Danielpour D (2005) Functions and regulation of transforming growth factor-beta (TGF-beta) in the prostate. Eur J Cancer 41: 846-857. doi: 10.1016/j.ejca.2004.12.027
    [52] Dallas SL, Zhao S, Cramer SD, et al. (2005) Preferential production of latent transforming growth factor beta-2 by primary prostatic epithelial cells and its activation by prostate-specific antigen. J Cell Physiol 202: 361-370. doi: 10.1002/jcp.20147
    [53] Laverty H, Wakefield L, Occleston N, et al. (2009) TGF-β3 and cancer: a review. Cytokine Growth Factor Rev 20: 305-317. doi: 10.1016/j.cytogfr.2009.07.002
    [54] Parada D, Arciniegas E, Moreira O, et al. (2003) Transforming growth factor-beta2 and beta3 expression in carcinoma of the prostate. Arch Esp Urol 57: 93-99.
    [55] Djonov V, Ball RK, Graf S, et al. (1997) Transforming growth factor‐β3 is expressed in nondividing basal epithelial cells in normal human prostate and benign prostatic hyperplasia, and is no longer detectable in prostate carcinoma. Prostate 31: 103-109.
    [56] Hisataki T, Itoh N, Suzuki K, et al. (2004) Modulation of phenotype of human prostatic stromal cells by transforming growth factor‐betas. Prostate 58: 174-182. doi: 10.1002/pros.10320
    [57] Ao M, Williams K, Bhowmick NA, et al. (2006) Transforming growth factor-β promotes invasion in tumorigenic but not in nontumorigenic human prostatic epithelial cells. Cancer Res 66: 8007-8016. doi: 10.1158/0008-5472.CAN-05-4451
    [58] Wikström P, Stattin P, Franck‐Lissbrant I, et al. (1998) Transforming growth factor β1 is associated with angiogenesis, metastasis, and poor clinical outcome in prostate cancer. Prostate 37: 19-29.
    [59] Steiner MS, Barrack ER (1992) Transforming growth factor-beta 1 overproduction in prostate cancer: effects on growth in vivo and in vitro. Mol Endocrinol 6: 15-25.
    [60] Steiner MS, Zhou Z-Z, Tonb DC, et al. (1994) Expression of transforming growth factor-beta 1 in prostate cancer. Endocrinology 135: 2240-2247.
    [61] Mu Y, Sundar R, Thakur N, et al. (2011) TRAF6 ubiquitinates TGF [beta] type I receptor to promote its cleavage and nuclear translocation in cancer. Nat Commun 2: 330. doi: 10.1038/ncomms1332
    [62] Wan Y, Yang M, Kolattukudy S, et al. (2010) Activation of cAMP-responsive-element-binding protein by PI3 kinase and p38 MAPK is essential for elevated expression of transforming growth factor β2 in cancer cells. J Interferon Cytokine Res 30: 677-681. doi: 10.1089/jir.2009.0117
    [63] Bang Y-J, Kim S-J, Danielpour D, et al. (1992) Cyclic AMP induces transforming growth factor beta 2 gene expression and growth arrest in the human androgen-independent prostate carcinoma cell line PC-3. Proc Natl Acad Sci U S A 89: 3556-3560. doi: 10.1073/pnas.89.8.3556
    [64] Cho HJ, Baek KE, Saika S, et al. (2007) Snail is required for transforming growth factor-β-induced epithelial–mesenchymal transition by activating PI3 kinase/Akt signal pathway. Biochem Biophys Res Commun 353: 337-343. doi: 10.1016/j.bbrc.2006.12.035
    [65] Walker L, Millena AC, Strong N, et al. (2013) Expression of TGFβ3 and its effects on migratory and invasive behavior of prostate cancer cells: involvement of PI3-kinase/AKT signaling pathway. Clin Exp Metastasis 30: 13-23. doi: 10.1007/s10585-012-9494-0
    [66] Vo BT, Morton Jr D, Komaragiri S, et al. (2013) TGF-β effects on prostate cancer cell migration and invasion are mediated by PGE2 through activation of PI3K/AKT/mTOR pathway. Endocrinology 154: 1768-1779. doi: 10.1210/en.2012-2074
    [67] Williams RH, Stapleton A, Yang G, et al. (1996) Reduced levels of transforming growth factor beta receptor type II in human prostate cancer: an immunohistochemical study. Clin Cancer Res 2: 635-640.
    [68] Demagny H, De Robertis EM (2015) Smad4/DPC4: A barrier against tumor progression driven by RTK/Ras/Erk and Wnt/GSK3 signaling. Mol Cell Oncol: e989133.
    [69] Hahn SA, Schutte M, Hoque AS, et al. (1996) DPC4, a candidate tumor suppressor gene at human chromosome 18q21. 1. Science 271: 350-353. doi: 10.1126/science.271.5247.350
    [70] Papageorgis P (2015) TGFbeta Signaling in Tumor Initiation, Epithelial-to-Mesenchymal Transition, and Metastasis. J Oncol 2015: 587193.
    [71] Hofmann TG, Stollberg N, Schmitz ML, et al. (2003) HIPK2 regulates transforming growth factor-β-induced c-Jun NH2-terminal kinase activation and apoptosis in human hepatoma cells. Cancer Res 63: 8271-8277.
    [72] Edlund S, Bu S, Schuster N, et al. (2003) Transforming Growth Factor-β1 (TGF-β)–induced Apoptosis of Prostate Cancer Cells Involves Smad7-dependent Activation of p38 by TGF-β-activated Kinase 1 and Mitogen-activated Protein Kinase Kinase 3. Mol Biol Cell 14: 529-544. doi: 10.1091/mbc.02-03-0037
    [73] Kang H-Y, Huang K-E, Chang SY, et al. (2002) Differential modulation of androgen receptor-mediated transactivation by Smad3 and tumor suppressor Smad4. J Biol Chem 277: 43749-43756. doi: 10.1074/jbc.M205603200
    [74] Micke P (2004) Tumour-stroma interaction: cancer-associated fibroblasts as novel targets in anti-cancer therapy? Lung Cancer 45: S163-S175. doi: 10.1016/j.lungcan.2004.07.977
    [75] Sheng J, Chen W, Zhu HJ (2015) The immune suppressive function of transforming growth factor-beta (TGF-beta) in human diseases. Growth Factors 33: 92-101. doi: 10.3109/08977194.2015.1010645
    [76] Mu Y, Gudey SK, Landström M (2012) Non-Smad signaling pathways. Cell Tissue Res 347: 11-20. doi: 10.1007/s00441-011-1201-y
    [77] Thompson TB, Cook RW, Chapman SC, et al. (2004) Beta A versus beta B: is it merely a matter of expression? Mol Cell Endocrinol 225: 9-17. doi: 10.1016/j.mce.2004.02.007
    [78] Ryu B, Kern SE (2003) The essential similarity of TGFβ and activin receptor transcriptional responses in cancer cells. Cancer Biol Ther 2: 164-170. doi: 10.4161/cbt.2.2.276
    [79] Loomans HA, Andl CD (2014) Intertwining of Activin A and TGFβ Signaling: Dual Roles in Cancer Progression and Cancer Cell Invasion. Cancers 7: 70-91. doi: 10.3390/cancers7010070
    [80] Gold E, Marino FE, Harrison C, et al. (2013) Activin‐βc reduces reproductive tumour progression and abolishes cancer‐associated cachexia in inhibin‐deficient mice. J Pathol 229: 599-607. doi: 10.1002/path.4142
    [81] Marino FE, Risbridger G, Gold E (2014) The inhibin/activin signalling pathway in human gonadal and adrenal cancers. Mol Hum Reprod 20: 1223-1237. doi: 10.1093/molehr/gau074
    [82] Mellor SL, Cranfield M, Ries R, et al. (2000) Localization of Activin βA-, β B-, andβ C-Subunits in Human Prostate and Evidence for Formation of New Activin Heterodimers ofβ C-Subunit 1. J Clin Endocrinol Metab 85: 4851-4858.
    [83] Mellor SL, Ball EM, O'Connor AE, et al. (2003) Activin βC-subunit heterodimers provide a new mechanism of regulating activin levels in the prostate. Endocrinology 144: 4410-4419. doi: 10.1210/en.2003-0225
    [84] Stenvers KL, Findlay JK (2010) Inhibins: from reproductive hormones to tumor suppressors. Trends Endocrinol Metab 21: 174-180. doi: 10.1016/j.tem.2009.11.009
    [85] Walton KL, Makanji Y, Wilce MC, et al. (2009) A common biosynthetic pathway governs the dimerization and secretion of inhibin and related transforming growth factor β (TGFβ) ligands. J Biol Chem 284: 9311-9320. doi: 10.1074/jbc.M808763200
    [86] Johnson KE, Makanji Y, Temple-Smith P, et al. (2016) Biological activity and in vivo half-life of pro-activin A in male rats. Mol Cell Endocrinol 422: 84-92. doi: 10.1016/j.mce.2015.12.007
    [87] Sheth A, Panse G, Vaze A, et al. (1981) Inhibin in the human prostate. Arch Androl 6: 317-321. doi: 10.3109/01485018108987543
    [88] Sathe VS, Sheth NA, Phadke MA, et al. (1987) Biosynthesis and localization of inhibin in human prostate. Prostate 10: 33-43. doi: 10.1002/pros.2990100107
    [89] Thomas TZ, Wang H, Niclasen P, et al. (1997) Expression and localization of activin subunits and follistatins in tissues from men with high grade prostate cancer. J Clin Endocrinol Metab 82: 3851-3858. doi: 10.1210/jcem.82.11.4374
    [90] Balanathan P, Ball EM, Wang H, et al. (2004) Epigenetic regulation of inhibin alpha-subunit gene in prostate cancer cell lines. J Mol Endocrinol 32: 55-67. doi: 10.1677/jme.0.0320055
    [91] Mellor SL, Richards MG, Pedersen JS, et al. (1998) Loss of the Expression and Localization of Inhibinα-Subunit in High Grade Prostate Cancer 1. J Clin Endocrinol Metab 83: 969-975.
    [92] Risbridger G, Shibata A, Ferguson K, et al. (2004) Elevated expression of inhibin α in prostate cancer. J Urol 171: 192-196. doi: 10.1097/01.ju.0000100048.98807.b7
    [93] Matzuk M, Finegold M, Mather J, et al. (1994) Development of cancer cachexia-like syndrome and adrenal tumors in inhibin-deficient mice. Proc Natl Acad Sci U S A 91: 8817-8821. doi: 10.1073/pnas.91.19.8817
    [94] Coerver KA, Woodruff TK, Finegold MJ, et al. (1996) Activin signaling through activin receptor type II causes the cachexia-like symptoms in inhibin-deficient mice. Mol Endocrinol 10: 534-543.
    [95] Dalkin A, Gilrain J, Bradshaw D, et al. (1996) Activin inhibition of prostate cancer cell growth: selective actions on androgen-responsive LNCaP cells. Endocrinology 137: 5230-5235.
    [96] Liu QY, Niranjan B, Gomes P, et al. (1996) Inhibitory effects of activin on the growth and morphogenesis of primary and transformed mammary epithelial cells. Cancer Res 56: 1155-1163.
    [97] Valderrama-Carvajal H, Cocolakis E, Lacerte A, et al. (2002) Activin/TGF-β induce apoptosis through Smad-dependent expression of the lipid phosphatase SHIP. Nat Cell Biol 4: 963-969. doi: 10.1038/ncb885
    [98] Leto G, Incorvaia L, Badalamenti G, et al. (2006) Activin A circulating levels in patients with bone metastasis from breast or prostate cancer. Clin Exp Metastasis 23: 117-122. doi: 10.1007/s10585-006-9010-5
    [99] Sottnik JL, Keller ET (2013) Understanding and targeting osteoclastic activity in prostate cancer bone metastases. Curr Mol Med 13: 626-639. doi: 10.2174/1566524011313040012
    [100] Gold E, Jetly N, O'Bryan MK, et al. (2009) Activin C antagonizes activin A in vitro and overexpression leads to pathologies in vivo. Am J Pathol 174: 184-195. doi: 10.2353/ajpath.2009.080296
    [101] Wakefield LM, Hill CS (2013) Beyond TGFβ: roles of other TGFβ superfamily members in cancer. Nat Rev Cancer 13: 328-341. doi: 10.1038/nrc3500
    [102] Katagiri T, Watabe T (2016) Bone Morphogenetic Proteins. Cold Spring Harb Perspect Biol 8: a021899. doi: 10.1101/cshperspect.a021899
    [103] Wozney JM, Rosen V, Celeste AJ, et al. (1988) Novel regulators of bone formation: molecular clones and activities. Science 242: 1528-1534. doi: 10.1126/science.3201241
    [104] Balboni AL, Cherukuri P, Ung M, et al. (2015) p53 and ΔNp63α Coregulate the Transcriptional and Cellular Response to TGFβ and BMP Signals. Mol Cancer Res 13: 732-742. doi: 10.1158/1541-7786.MCR-14-0152-T
    [105] Herpin A, Lelong C, Favrel P (2004) Transforming growth factor-beta-related proteins: an ancestral and widespread superfamily of cytokines in metazoans. Dev Comp Immunol 28: 461-485. doi: 10.1016/j.dci.2003.09.007
    [106] Pils D, Wittinger M, Petz M, et al. (2010) BAMBI is overexpressed in ovarian cancer and co-translocates with Smads into the nucleus upon TGF-ß treatment. Gynecol Oncol 117: 189-197. doi: 10.1016/j.ygyno.2009.12.034
    [107] Cao X, Chen D (2005) The BMP signaling and in vivo bone formation. Gene 357: 1-8. doi: 10.1016/j.gene.2005.06.017
    [108] Onichtchouk D, Chen Y-G, Dosch R, et al. (1999) Silencing of TGF-β signalling by the pseudoreceptor BAMBI. Nature 401: 480-485. doi: 10.1038/46794
    [109] Gamer LW, Cox K, Carlo JM, et al. (2009) Overexpression of BMP3 in the developing skeleton alters endochondral bone formation resulting in spontaneous rib fractures. Dev Dynam 238: 2374-2381. doi: 10.1002/dvdy.22048
    [110] Gamer LW, Nove J, Levin M, et al. (2005) BMP-3 is a novel inhibitor of both activin and BMP-4 signaling in Xenopus embryos. Dev Biol 285: 156-168. doi: 10.1016/j.ydbio.2005.06.012
    [111] Stewart A, Guan H, Yang K (2010) BMP‐3 promotes mesenchymal stem cell proliferation through the TGF‐β/activin signaling pathway. J Cell Physiol 223: 658-666.
    [112] Olsen OE, Wader KF, Hella H, et al. (2015) Activin A inhibits BMP-signaling by binding ACVR2A and ACVR2B. Cell Commun Signal 13: 27. doi: 10.1186/s12964-015-0104-z
    [113] Harris SE, Harris MA, Mahy P, et al. (1994) Expression of bone morphogenetic protein messenger RNAs by normal rat and human prostate and prostate cancer cells. Prostate 24: 204-211. doi: 10.1002/pros.2990240406
    [114] Masuda H, Fukabori Y, Nakano K, et al. (2004) Expression of bone morphogenetic protein‐7 (BMP‐7) in human prostate. Prostate 59: 101-106. doi: 10.1002/pros.20030
    [115] Lamm ML, Podlasek CA, Barnett DH, et al. (2001) Mesenchymal factor bone morphogenetic protein 4 restricts ductal budding and branching morphogenesis in the developing prostate. Dev Biol 232: 301-314. doi: 10.1006/dbio.2001.0187
    [116] Ye L, Lewis-Russell JM, Kyanaston HG, et al. (2007) Bone morphogenetic proteins and their receptor signaling in prostate cancer. Histol Histopathol 22: 1129-1147.
    [117] Dai J, Keller J, Zhang J, et al. (2005) Bone morphogenetic protein-6 promotes osteoblastic prostate cancer bone metastases through a dual mechanism. Cancer Res 65: 8274-8285. doi: 10.1158/0008-5472.CAN-05-1891
    [118] Liao A, Wang W, Sun D, et al. (2014) Bone morphogenetic protein 2 mediates epithelial-mesenchymal transition via AKT and ERK signaling pathways in gastric cancer. Tumour Biol: 1-6.
    [119] Lai TH, Fong YC, Fu WM, et al. (2008) Osteoblasts‐derived BMP‐2 enhances the motility of prostate cancer cells via activation of integrins. Prostate 68: 1341-1353. doi: 10.1002/pros.20799
    [120] Feeley BT, Krenek L, Liu N, et al. (2006) Overexpression of noggin inhibits BMP-mediated growth of osteolytic prostate cancer lesions. Bone 38: 154-166. doi: 10.1016/j.bone.2005.07.015
    [121] Kwon SJ, Lee GT, Lee JH, et al. (2014) Mechanism of pro‐tumorigenic effect of BMP‐6: Neovascularization involving tumor‐associated macrophages and IL‐1α. Prostate 74: 121-133. doi: 10.1002/pros.22734
    [122] Piccirillo S, Reynolds B, Zanetti N, et al. (2006) Bone morphogenetic proteins inhibit the tumorigenic potential of human brain tumour-initiating cells. Nature 444: 761-765. doi: 10.1038/nature05349
    [123] Buijs JT, Rentsch CA, van der Horst G, et al. (2007) BMP7, a putative regulator of epithelial homeostasis in the human prostate, is a potent inhibitor of prostate cancer bone metastasis in vivo. Am J Pathol 171: 1047-1057. doi: 10.2353/ajpath.2007.070168
    [124] Bokobza SM, Ye L, Kynaston HG, et al. (2010) GDF‐9 promotes the growth of prostate cancer cells by protecting them from apoptosis. J Cell Physiol 225: 529-536. doi: 10.1002/jcp.22235
    [125] Bokobza SM, Ye L, Kynaston H, et al. (2011) Growth and differentiation factor 9 (GDF-9) induces epithelial–mesenchymal transition in prostate cancer cells. Mol Cell Biochem 349: 33-40. doi: 10.1007/s11010-010-0657-5
    [126] Yan C, Wang P, DeMayo J, et al. (2001) Synergistic roles of bone morphogenetic protein 15 and growth differentiation factor 9 in ovarian function. Mol Endocrinol 15: 854-866. doi: 10.1210/mend.15.6.0662
    [127] Wu X, Matzuk MM (2002) GDF-9 and BMP-15: oocyte organizers. Rev Endocr Metab Disord 3: 27-32. doi: 10.1023/A:1012796601311
    [128] Peng J, Li Q, Wigglesworth K, et al. (2013) Growth differentiation factor 9: bone morphogenetic protein 15 heterodimers are potent regulators of ovarian functions. Proc Natl Acad Sci U S A 110: E776-E785. doi: 10.1073/pnas.1218020110
    [129] Brown DA, Stephan C, Ward RL, et al. (2006) Measurement of serum levels of macrophage inhibitory cytokine 1 combined with prostate-specific antigen improves prostate cancer diagnosis. Clin Cancer Res 12: 89-96. doi: 10.1158/1078-0432.CCR-05-1331
    [130] Aw Yong KM, Zeng Y, Vindivich D, et al. (2014) Morphological effects on expression of growth differentiation factor 15 (GDF15), a marker of metastasis. J Cell Physiol 229: 362-373. doi: 10.1002/jcp.24458
    [131] Kojima M, Troncoso P, Babaian RJ (1998) Influence of noncancerous prostatic tissue volume on prostate-specific antigen. Urology 51: 293-299. doi: 10.1016/S0090-4295(97)00497-4
    [132] Gritsman K, Talbot WS, Schier AF (2000) Nodal signaling patterns the organizer. Development 127: 921-932.
    [133] Gray PC, Harrison CA, Vale W (2003) Cripto forms a complex with activin and type II activin receptors and can block activin signaling. Proc Natl Acad Sci U S A 100: 5193-5198. doi: 10.1073/pnas.0531290100
    [134] Harrison CA, Gray PC, Vale WW, et al. (2005) Antagonists of activin signaling: mechanisms and potential biological applications. Trends Endocrinol Metab 16: 73-78. doi: 10.1016/j.tem.2005.01.003
    [135] Yeo C-Y, Whitman M (2001) Nodal signals to Smads through Cripto-dependent and Cripto-independent mechanisms. Mol Cell 7: 949-957. doi: 10.1016/S1097-2765(01)00249-0
    [136] Chen C, Shen MM (2004) Two modes by which Lefty proteins inhibit nodal signaling. Curr Biol 14: 618-624. doi: 10.1016/j.cub.2004.02.042
    [137] Belo JA, Bachiller D, Agius E, et al. (2000) Cerberus-like is a secreted BMP and nodal antagonist not essential for mouse development. Genesis 26: 265-270.
    [138] Lawrence MG, Margaryan NV, Loessner D, et al. (2011) Reactivation of embryonic nodal signaling is associated with tumor progression and promotes the growth of prostate cancer cells. Prostate 71: 1198-1209. doi: 10.1002/pros.21335
    [139] De Silva T, Ye G, Liang Y-Y, et al. (2012) Nodal promotes glioblastoma cell growth. Frontiers Endocrinol 3: 1-6.
    [140] Aykul S, Ni W, Mutatu W, et al. (2014) Human cerberus prevents nodal-receptor binding, inhibits nodal signaling, and suppresses nodal-mediated phenotypes. PLoS One 10: e0114954.
    [141] Vo BT, Khan SA (2011) Expression of nodal and nodal receptors in prostate stem cells and prostate cancer cells: autocrine effects on cell proliferation and migration. Prostate 71: 1084-1096. doi: 10.1002/pros.21326
    [142] Yingling JM, Blanchard KL, Sawyer JS (2004) Development of TGF-β signalling inhibitors for cancer therapy. Nat Rev Drug Discov 3: 1011-1022. doi: 10.1038/nrd1580
    [143] Isaacs MJ, Kawakami Y, Allendorph GP, et al. (2010) Bone morphogenetic protein-2 and-6 heterodimer illustrates the nature of ligand-receptor assembly. Mol Endocrinol 24: 1469-1477. doi: 10.1210/me.2009-0496
    [144] Neugebauer JM, Kwon S, Kim H-S, et al. (2015) The prodomain of BMP4 is necessary and sufficient to generate stable BMP4/7 heterodimers with enhanced bioactivity in vivo. Proc Natl Acad Sci U S A 112: E2307-E2316. doi: 10.1073/pnas.1501449112
    [145] Yeo C-Y, Whitman M (2001) Nodal signals to Smads through Cripto-dependent and Cripto-independent mechanisms. Mol Cell 7: 949-957. doi: 10.1016/S1097-2765(01)00249-0
    [146] Fuerer C, Nostro MC, Constam DB (2014) Nodal. Gdf1 heterodimers with bound prodomains enable serum-independent nodal signaling and endoderm differentiation. J Biol Chem 289: 17854-17871.
    [147] Harrington AE, Morris‐Triggs SA, Ruotolo BT, et al. (2006) Structural basis for the inhibition of activin signalling by follistatin. EMBO J 25: 1035-1045. doi: 10.1038/sj.emboj.7601000
    [148] Guo X, Wang X-F (2009) Signaling cross-talk between TGF-β/BMP and other pathways. Cell Res 19: 71-88. doi: 10.1038/cr.2008.302
  • This article has been cited by:

    1. Kooi-Kuan Yoong, Roslan Hasni, Gee-Choon Lau, Muhammad Ahsan Asim, Ali Ahmad, Reflexive edge strength of convex polytopes and corona product of cycle with path, 2022, 7, 2473-6988, 11784, 10.3934/math.2022657
    2. Yuanyuan Shen, Xinhui An, Baoyindureng Wu, Zuonong Zhu, Hamilton-Connected Mycielski Graphs∗, 2021, 2021, 1607-887X, 1, 10.1155/2021/3376981
    3. Yubin Zhong, Sakander Hayat, Suliman Khan, Vito Napolitano, Mohammed J. F. Alenazi, Combinatorial analysis of line graphs: domination, chromaticity, and Hamiltoniancity, 2025, 10, 2473-6988, 13343, 10.3934/math.2025599
  • Reader Comments
  • © 2017 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(8083) PDF downloads(1183) Cited by(2)

Figures and Tables

Figures(3)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog