In the last few years, cells have been increasingly used as vehicles for the delivery of therapeutics. The cell-based approach is particularly attractive for the delivery of biotherapeutic agents that are difficult to synthesize, have limited tissue penetrance, or are rapidly inactivated upon direct in vivo introduction. Some of the key factors for the success of this type of therapeutic delivery have been established, such as the means and efficiency of cell loading with a therapeutic payload, and the nature of therapeutics that the cells can carry. However, the issue of biodistribution of injected cell carriers in vivo still remains an important aspect of cell-based delivery that has yet to be fully investigated. Importantly, different types of cell vehicles may have specific biodistribution or cell homing patterns and, therefore, may provide a special advantage to achieve site-specific or targeted delivery of therapeutics.

The ability of injected cells to either passively concentrate in specific organs or actively home to disease sites supports the rationale for targeted delivery of therapeutics by cell vehicles. There is growing evidence that sites of injury or growing tumors favor active homing of endogenous and exogenous stem or progenitor cells 1 2 . The first observation of this phenomenon was published by Studeny et al, using MSCs as vehicles delivering IFNβ 3 . This and a subsequent study by the same group 4 reported MSC localization in lung tumors after systemic injection of these cells. The recognition that the tumor microenvironment or tumor cytokine profile is similar to that of inflammatory sites evoked a search for the tumor attracting signals. Despite still incomplete knowledge of these cues, the practical aspects of cell-based delivery of therapeutics to specific sites have been actively exploited. A growing number of studies have used MSCs as cell vehicles to exploit their native ability to target tumors, as a means to track malignant tissues or for the delivery of anticancer agents to tumors 2 5 6 7 8 9 . Several studies investigated MSCs as cell vehicles for the delivery of various clinically relevant anticancer factors, including cytokines, interferons, pro-drugs and replication competent adenoviruses, with noted benefits 10 11 12 13 . The native tumor homing phenomenon of MSCs was confirmed in different experimental systems 2 12 . Other cell types, such as umbilical cord matrix stem cells (UCMS) 14 , neural stem cells 15 16 and endothelial progenitor cells 17 18 have also demonstrated the inherent ability to migrate toward tumors or other pathologies.

Along with using native cell homing properties, modification of the cell membrane by expressing appropriate receptors was also proposed as a means to obtain targeted cell vehicles. Much of the groundwork for such targeting approaches has previously been established for immune cells (T-cells, NK cells, CIK cells), where lymphocyte populations were modified to express artificial receptors (T-bodies) with distinct binding specificities to target cells. Artificial or chimeric receptors (AR) have been derived from the binding domains of antibodies (usually the single chain antibody, scFv) or T-cell receptors. An array of chimeric receptors, mostly with specificity for different tumor markers, has been tested for biological function in vitro 19 20 and in vivo 21 22 . This approach is often termed "targeted" adoptive immunotherapy, since the active targeting mechanism was added to redirect the native killing function of an immune cell to a defined target cell. Remarkably, the added affinity to retarget cell killing function was found to enhance localization of the modified cells to the target sites. Several studies demonstrated that AR-modified lymphocytes are detected in higher numbers in tumors that express the cognate receptor, compared to untargeted cells 23 24 .

Despite showing its potential, the AR-based strategy has not been translated to other cell types that may serve as promising cell vehicles. Only a few applications have demonstrated the feasibility of using AR as a binding moiety in non-immune cell contexts 25 26 . In other examples, surface-expressed scFvs served as artificial receptors for viral infection 27 or enhanced the tumor cell binding 28 . Therefore, applying the AR strategy to other cell types and investigating the potential targeting benefits holds promise as a means to increase cell concentration in desired sites. Of note, most of the studies using native MSC homing did not quantitatively determine the level of cells that home to tumors or other sites. The tumor homing behavior of MSCs was demonstrated by the mere presence of these cells in the sites of interest and/or lack of such cells in other organs 8 11 13 29 . The few studies that did attempt to quantitatively estimate MSC numbers localized in tumors have reported low to moderate numbers 3 5 10 . Since increasing the number of cell vehicles in tumors would parallel therapeutic efficacy, investigation of native or artificial means of cell homing to tumors are of high therapeutic importance.

The present study tested the hypothesis that artificial receptors with affinities to target sites can be added to cell vehicles and the new cell binding properties can be utilized to increase cell vehicle levels in the target sites. Specifically, we investigated the possibility of increasing the number of MSCs in ovarian tumors by expressing a tumor antigen-binding receptor on the MSC surface. This would provide an additional means to increase the number of tumor-associated MSCs beyond their native tumor homing potential. To this end, we have created MSCs that express an artificial receptor (AR) that binds to erbB2, a frequent marker of tumor cells (MSC-AR). We have shown that these AR-expressing MSCs (MSC-AR) have enhanced binding to erbB2-expressing cells in vitro. Furthermore, we tested erbB-2 targeting of MSC-AR in model systems in vivo and demonstrated that addition of the AR increased retention of circulating MSC-AR in erbB2-expressing sites. We also confirmed an increased concentration of MSC-AR, compared to MSC, in erbB2 positive ovarian tumors.

These data show that the number of tumor-associated MSCs can be increased via affinity-based targeting, which can potentially serve to improve therapeutic delivery. Broadly, we demonstrated that an artificial cell targeting strategy can be beneficial to MSC-based cell vehicles and suggests that this strategy could also have potential for other cell types that lack native homing abilities.

A growing number of studies utilize engineered MSCs as a tool to track malignant tissues and deliver anticancer agents within the tumor microenvironment. MSC homing to tumors has been confirmed in a variety of experimental models, however the homing efficiency is clearly model-dependent and generally modest 3 10 . Additional cell targeting efforts may enhance the efficiency of tumor homing and consequently deliver more therapeutics. These cell targeting efforts may include physical cell routing, utilization of physiological forces for cell concentration and strategies that involve intrinsic or engineered cell homing/targeting mechanisms 40 . Targeting strategies can be used singly or in combinations to maximize cell vehicle concentration in the target site. For instance, combined native MSC tumor homing with preconditioning of the tumor site by irradiation has been shown to enhance MSC homing to irradiated tumors 41 . Native cell homing can also be combined with other types of cell targeting means 42 . The current study investigated whether native tumor targeting of MSCs can be enhanced by engineered targeting via expressing an artificial tumor-binding receptor.

Our study applied affinity-based targeting to cell vehicles that lack immune recognition. To date, only a few applications have demonstrated the feasibility of using scFvs as binding moieties in non-immune cell contexts. One example is where an artificial chimeric receptor was applied to primary human monocytes to target monocytes to CEA-expressing tumor cells 25 . Another study used gpi-anchored anti-CD20 scFv fragments exposed on red blood cells (RBC) and evaluated binding of targeted erythrocytes to CD20 positive tumors 26 . In our in vitro experiments, MSCs grafted with anti-erbB2 artificial receptors demonstrated increased binding to cells overexpressing erbB2 (40% in experimental group versus 8% in control). The only available study that investigated similar erbB2-based cell binding interactions 28 reported increased cell binding numbers that are in a good agreement with our results (20% in experimental group compared to 6-8% in control).

The next important question was whether the enhanced MSC-AR binding ability would translate to an in vivo tumor localization advantage, compared to unmodified cells. Given complexity of the processes of biodistribution and homing of injected cells, we reasoned that an effect of engineered cell targeting would be more pronounced and better detectable in model systems. For instance, an isolated heart model was used to detect the difference of MSC homing to normal versus infarcted myocardium 43 . Thus, for initial testing we choose a transient transgenic mouse model previously used to validate targeting of the affinity-modified adenoviral vectors 44 . Expression of erbB2 tumor marker in the mouse lungs ensures its easy accessibility to systemically injected cells and direct cell-marker contact. In addition, this model allows the dissection of only the affinity-related component of cell targeting, since native homing of MSC to lungs has not been reported. Of note, this model is easily manipulated whereby other markers can be tested in similar fashion.

It is not accidental that most studies detecting tumor homing of intravenously introduced MSC were performed on lung tumor models 4 10 13 45 46 . This mode of cell introduction utilizes two cell-targeting mechanisms, temporal physiological accumulation of cells in the lungs and native MSC tumor homing, whereby lung-concentrated MSCs actively migrate to local lung tumors. It was expected that accumulation of MSC in the lungs after systemic injection would be the same for modified and unmodified MSCs due to the first-pass effect 39 . However, AR-expressing cells by virtue of enhanced cell-cell interactions may show different levels of cell retention and kinetics of subsequent lung evacuation. In two subsequent experiments we have shown that MSC and MSC-AR have a different pattern of interaction with erbB2-lungs. An increased number of MSC-AR was detected in the lungs at several time points compared to MSC numbers. The time window, where the differences in experimental groups were detectable, was relatively short (14-32 hrs). At more distant time points (52, 72 hrs) MSCs were not detected in the lungs using this method. We believe that the major reason for this is MSC destruction. The hCAR transgenic mice are immunocompetent and xenogeneic (human) MSCs introduced into immunocompetent mice are likely to be killed by immune-based mechanisms over time. Despite the short window of opportunity for detecting differences, this model, nevertheless, gave us an indication that modified cells have different behavior in the model system and in vivo cell-cell interactions result in a detectable cell retention effect.

A more relevant and stringent model to test potential benefits of additional MSC targeting is the ovarian tumor model. We have previously demonstrated the native ability of MSC to home to SKOV3ip1 xenografts 12 . The high level of erbB2 expression makes the SKOV3ip1 model appropriate to test our double-targeting strategy, which engages both mechanisms of MSC-AR tumor targeting: native and engineered. Multiple primary and metastatic tumor nodules with generally poor developed stromal structures may again offer better accessibility of tumor markers to cell vehicles expressing AR and allows the detection of the benefit of affinity-based targeting. In the pilot experiment, a substantial increase (8-folds) in the number of tumor-associated MSC-AR versus MSC was detected. To validate this initial observation and to more accurately establish the timing of MSC homing, we investigated the kinetics of MSC tumor targeting. The speed and pattern of cell vehicles homing to target sites are important parameters to consider in designing therapeutic delivery strategies, as these values may differ considerably. For instance, homing of systemically injected CD34+ cells to bone marrow is very fast; these cells reach the bone marrow in 1 hr 47 . However, there is not much data on the efficiency and speed of MSC homing to tumors. Upon systemic injection MSC tumor homing is apparently delayed and diminished due to trapping in the lung vasculature. It is reported that upon systemic injection, MSC can stay in the lungs for two or more days 7 48 , thus the intravenous route of MSC introduction is slow and inefficient. Recent quantitative studies found less than 1% of systemically injected MSC is able to reach distant sites 49 50 . In the ip tumor settings, MSCs did not have to pass the lungs, thus the anticipated time for MSC tumor homing is expected to be much shorter and efficiency better. Despite this prediction, the actual kinetics of MSC homing was unknown. In our experiment, we detected preferential homing of MSCs to tumors as early as 2 hrs, while the maximum homing of MSCs to ovarian tumors was observed at 24 hrs after cell injection. The kinetics of homing is an important parameter for our future strategy of using MSC-based vehicles to deliver oncolytic adenoviruses. Quick homing ensures that cells have enough time to reach tumors before they are killed by virus replication.

We detected preferential tumor localization of MSCs in both MSC and MSC-AR groups, which confirmed the native tumor homing abilities of these cells and assured that these functions are not perturbed by AR expression. In two separate experiments we detected an increase in the number of tumor-associated MSC-AR versus MSC-GFP starting from 24 hrs using two methods of luciferase quantitation.

The analysis of MSC distribution in the peritoneum demonstrated that intraperitoneally developed tumors are the major cell homing sites in both groups and across all time points tested. The ability to localize to even minor tumor metastasis at the surface of organs is a remarkable property of MSCs that can be exploited for diagnostic or treatment endpoints. Although addition of the AR might not influence the actual process of MSC moving towards a chemotactic homing gradient, it was able to "strengthen" binding to tumors and resulted in increased tumor-associated cell numbers and the tumor/liver ratio. The mechanism of these effects is potentially mediated by increased adherence to tumor cells, which affected both the efficacy (number of cells attached to tumors) as well as the specificity (ratio of tumor-bound versus other organ bound cells).

In both models we capitalized on the accessibility of targeting tumor markers to cells expressing AR. The necessity of the accessibility may dictate a careful selection of the marker and corresponding targeting moiety. For instance, under systemic injection, circulated cells are most likely to have physical contact with endothelial cells, supporting the targeting of specific endothelium markers. Such selective binding of circulating cells to key neovasculature markers has been described 51 52 . An avenue for utilizing tumor markers may be facilitated by irregular and atypical tumor vasculature allowing direct contact of circulating cell vehicles with tumor cells. Further studies to identify such markers and to test if cell targeting to these markers increases their retention in tumors are needed.

Another important issue concerning the therapeutic use of cells as vehicles is the quantity of cells reaching the desired destination, as well as an understanding of how these numbers translate to therapeutic benefits. Some applications might need a maximum possible cell number to achieve a therapeutic benefit 6 7 53 , while others may benefit from delivery just a few cells to trigger the desired effect 45 . Thus, knowledge of the quantitative characteristics of cell homing in different models is useful and needed for further translation of these strategies to the clinic.

The majority of studies exploiting MSC tumor homing have only demonstrated the presence of labeled MSCs in the tumor parenchyma 8 11 13 29 . The quantitative aspect of cell homing or targeting to some extent is present in the available literature, however, it is not usually the major subject of these studies and, therefore, is not systematically approached. Meaningful information on homing efficiency can only be extracted when comparisons are performed within the same study. For instance, such comparisons are reported on different routes of MSC injection 54 or homing of MSC to non-irradiated versus irradiated tumors in single animal 41 55 or comparing MSC and 3T3 tumor homing 9 .

Among the studies that attempted quantitation of MSC numbers in particular sites, only moderate cell numbers in the targeted tissues were reported 3 10 . MSC numbers in these studies are mostly expressed as relative units, thus, preventing the calculation of the actual homing efficiencies as a proportion of the injected cell dose. To date, most studies attempting MSC quantitaion were performed on lung tumors or lung metastases. Despite moderate cell numbers in lung tumors reported, these studies demonstrated the therapeutic effect of local cell-based transgene delivery. This is an important observation, as it demonstrates that even moderate cell numbers in lung tumors are sufficient to show a therapeutic benefit to this approach. Studies investigating MSC homing to distant (subcutaneous) tumors after systemic injection reported more controversial numbers 5 . While the presence of MSCs in subcutaneous tumors after iv introduction in general was demonstrated, only one study has reported on the therapeutic efficacy of MSC-based delivery to sc tumors 8 . The benefits of therapeutic treatment in such settings remain to be proved. Thus, despite the fact of the recruitment of MSC to tumors has been established in a variety of experimental models, the efficacy of this process in each case varies and is still presents a subject for investigation.

The major purpose of our study was to investigate the differences that AR-modified cells achieve in tumor targeting. Therefore, thorough quantitative evaluation of absolute cell numbers per tumor or other organs was not performed. Of note, some features of the ovarian tumor model influence accurate quantitative estimation of tumor cell homing and have to be accounted for. Multiple tumor nodules and metastasis hamper accurate collection of the entire tumor sample, which results in underestimation of total MSC tumor homing levels. On the contrary, metastases to organs (spleen, liver, intestine surfaces), if not identified and dissected out, may incorrectly attribute MSC homing to these organs, especially using bulk assays such as luciferase activity of organ lysate. This leads to an overestimation of MSC distribution to off-tumor sites, while, in fact, this is also tumor-related homing. Therefore, accurate quantitative analysis would require more attention to both, the procedure of organ collection for analysis, and the ability to better visualize tumor nodules.

Nevertheless, the importance of quantitative assessment, as well as developing an accurate methodology for determination of absolute cells numbers in organs has been recognized. Thus, tumor homing data were presented as cells per mg of protein, which gives a ballpark estimation of cell numbers present in tumors at these conditions. The level of native MSC tumor targeting was roughly estimated as 10% of the injected cell dose, while addition of the artificial receptor increased this efficiency 1.5-2 times. Based on this estimation, we believe that MSC-based therapeutic delivery has more practical utility after tumor debulking. In the residual disease it may provide much higher vehicle/tumor cell ratio than cases in which large primary tumors exist. Also, a marker targeted by MSC-AR will be better exposed on small tumor nodules without prominent stromal component or on patches of disseminated tumor cells.

Our study confirmed that modification of cell carriers via expressing artificial receptor mediates the numbers of injected cell carriers in the organ or site of interest. We have demonstrated the practical relevance of our strategy in an ovarian tumor model, and showed that the number of modified MSC carriers increased in intraperitoneal tumors. Artificial receptor strategy can be applicable to other cell types, especially to circulating cells lacking native homing abilities.

The authors declare that they have no competing interests.

SK, LP carried out all study including in vitro and in vivo experiments needed to test porposed strategy. JR helped with in vitro studies, drafting and editing the manuscript. RA carried out general supervision regarding ovarian model used, involved in drafting the manuscript. DTC oversaw the project, have made contribution to study design and discussion of ideas and results. LP have made the major contribution to developing the concept of cell targeting, carried out all study design, acquisition and interpretation of data, wrote and edited the manuscript. All authors read and approved the final version of the manuscript.