The wide spread of mobile devices has enabled a new paradigm of innovation called Mobile Crowdsourcing (MCS) where the concept is to allow entities, e.g., individuals or local authorities, to hire workers to help from the crowd of connected people, to execute a task or service. Some complex tasks require the collaboration of multiple workers to ensure its successful completion. In this context, the task requester needs to hire a group of socially connected and collaborative workers that, at the same time, have sufficient skills to accomplish the task. In this paper, we develop two recruitment strategies for collaborative MCS frameworks in which, virtual teams are formed according to four different criteria: level of expertise, social relationship strength, recruitment cost, and recruiter's confidence level. The first proposed strategy is a platform-based approach which exploits the platform knowledge to form the team. The second one is a leader-based approach that uses team members' knowledge about their social network (SN) neighbors to designate a group leader that recruits its suitable team. Both approaches are modeled as integer linear programs resulting in optimal team formation. Experimental results show a performance trade-off between the two virtual team grouping strategies when varying the members SN edge degree. Compared to the leader-based strategy, the platform-based strategy recruits a more skilled team but with lower SN relationships and higher cost.