L calc program

The CFC assay was described previously [25] (link). Briefly, freshly isolated total bone marrow cells were suspended in Methocult GF M3434 methylcellulose medium (StemCell Technologies, Vancouver, B.C., Canada). Triplicate cultures were set up for each specimen according to the manufacturer's instructions. After 14 days of culture, the total colony- forming unit (CFU) were counted.
The CAFC assay was described previously [25] (link). Briefly, a monolayer of FBMD-1 stromal cells was plated in 96-well plates and grown to confluent; whole bone marrow cells were seeded at 81,000, 27,000, 9000, 3000, 1000 or 333 cells per well in CAFC medium (Iscove's MDM supplemented with 20% horse serum, 10⁻⁵ M hydrocortisone, 10⁻⁵ M 2-mercaptoethanol, 100 U/ml penicillin, and 100 μg/ml streptomycin). Twenty replicate wells per cell number were counted. The frequency of CAFC was determined on days 7, 14, 21, 28 and 35. Wells were scored positive if at least one phase dark hematopoietic clone (containing five or more cells) was seen. The frequencies of CAFCs were calculated using the L-Calc program (StemCell Technologies).

Inducible shRNA-ALDH1A3 UMSCC47 cells were suspended in 50:50 DMEM:Matrigel and implanted subcutaneously into the flank of 6–8 week old, female NOD/SCID mice. Subsequently, for each cell dilution, mice were randomly assigned to two treatment arms; control diet (n = 5) or doxycycline (200 mg/kg)-containing diet (n = 5) ad libitum. Tumor incidence was monitored for 49 days following tumor cell implantation. Cancer initiating cell frequency was calculated using the L-Calc program (STEMCELL Technologies Inc., Vancouver, Canada). Sample size estimate was not performed for this experiment. All animals were included in our analysis. Investigative team was not blinded to the group allocation during the experiment and when assessing the endpoint of tumor incidence. Animal experiments were conducted in compliance with ethical regulations and under an approved protocol from The Ohio State University.

Both the L-calc program (StemCell Technologies Inc, #28425 Version 1.1.1, October 2005) and/or the ELDA program (http://bioinf.wehi.edu.au/software/elda/) [76 (link)] were used to determine the frequencies of CICs and gave extremely similar results. The ELDA program was preferentially used. The significance between the median of the test of CIF between NOD/SCID and NSG mice and the number of CICs in the human population from the CA9-sorted populations were determined with a paired two-tailed Student t-test using the GraphPad software (GraphPad software Inc., San Diego, California). For the CA9^+/− populations, the differences between the CIF of sorted cells was determined using a one-way analysis of variance test included in the ELDA web tool. The linear regression and correlation analyses were generated using the GraphPad software. A two-tailed Spearman analysis was performed for all correlations. The significance between the fold change of expression of different markers between normoxic (CA9⁻) and hypoxic (CA9⁺) populations have been previously described [35 (link)].