Does all cancer cells come from a single clone of cells or are there evidences of polyclonal cancer cells too?
Most tumors are probably not strictly clonal; all cancer cells in a tumor seldom have the exact same set of mutations. Tumors tend to be subclonal, with all cancer cells having a single common ancestor but with later evolution resulting in some mutations shared among all cancer cells but others restricted to sub-populations of cancer cells. This fact and its implications for therapy were noted 40 years ago by Nowell (Science 194:23-28, 1976), and verified by much recent work as reviewed for example by McGranahan and Swanton (Cancer Cell 27:15-26, 2015).
A tumor might need to have a billion ($10^9$) or so cancer cells to be clinically detectable. Thus many years and many cell divisions have passed since the time that a single cell developed the set of mutations originally needed to produce the invasive tumor. With a probability of some uncorrected mutation at each cell division, particularly in cancer cells that often have defects in DNA replication and repair and in chromosome separation, it is essentially impossible for all of the billion cancer cells in the tumor to have identical sets of mutations. See, for example, the work of Bozic et al (eLife 2: e000747,2013). So although there is a clonal origin of the tumor, the tumor is typically a collection of related but genetically distinct populations best considered to be subclones.
Differences among subclones present a major barrier to traditional cytotoxic therapies, to more modern targeted therapies, and to the immunotherapies that presently receive so much attention. Even if most of the tumor is sensitive and responds, a subclone whose mutations somehow provide resistance to the therapy will be able to repopulate the tumor thereafter. Bozic and Nowak (PNAS 111:15964-8, 2014) estimated that a clinically detectable tumor may have 10 or more subclones resistant to any targeted therapy, due to mutations accumulated during the subclinical growth of the tumor.
As examples from immunotherapy for cancer, McGranahan et al recently showed (Science 351:1463-1469, 2016) that responses to immune checkpoint inhibition in melanoma and non-small cell lung cancer were best in patients with clonal rather than subclonal tumor neoantigens for the immune system to attack. Zaretzky et al (New Engl J Med 2016, epub ahead of print) identified subclonal mutations in interferon-receptor signaling and antigen presentation in some patients, mutations enriched in the recurrent tumors following immune checkpoint inhibition and thus likely accounting for relapse following that therapy.
It is possible in principle for a single physical tumor to have multiple cells of origin. Tumors often develop from a region of tissue that has been subject to field cancerization, where carcinogenic challenges have produced many cells with mutations that lead to dysregulated but not invasive behavior. This can lead to multiple independent cancers arising from the same general location, as individual cells later develop the full complement of mutations needed for an invasive tumor. In principle, such multiple cancers might coalesce into a single tumor mass so that the tumor is polyclonal, but I'm not aware of any well documented cases. So it seems that tumors are seldom truly polyclonal with multiple independent cells of origin.