Our approach was based on the assumption that factors, whose targeting leads to a TIF phenotype, would be either directly involved in telomere protection or be regulators and modifiers of shelterin components. Furthermore we reasoned that suppression of a gene product that results in TIF formation would also cause a general DNA damage phenotype, when telomeric localization is not considered as defining criteria. We therefore focused on a group of genes that has been selected in a genome-wide approach designed to identify general players in the DNA damage response. When cells are subjected to ionizing radiation they form foci call IRIF (Ionizing Radiation Induced Foci) 
, consisting of proteins involved in damage recognition and repair at sites of lesions. A genome-wide screen designed to identify factors, whose suppression affect IRIF formation after ionizing irradiation 
allowed us to focus on a list of 520 target factors. These genes haven initially been identified by knock-down using the siGENOMERNAi pools from Dharmacon, but have been further verified using alternative siRNAs by the Durocher laboratory. The shelterin components TRF1 and TRF2 were identified as modifiers of the damage response by the aforementioned screen and therefore contained within this list, suggesting that proteins that play a role in telomere function can be isolated with this approach. We further adapted the initial gene list and excluded 134 genes that would be unlikely directly linked to genome stability and telomere function (e.g. ribosomal proteins), allowing us to focus on 386 potential factors (Table S1
We first verified that transfection with according siRNA pools resulted in efficient knock-down of TRF1 and TRF2 and that this suppression was sufficient to induce TIF. For these controls we used the OnTargetPlus (OTP) pools from Dharmacon, the next generation of siRNA pools which differ from the siGENOME pools by a novel target sequence selection algorithm and a unique single-strand RNA modification (http://www.dharmacon.com/product/productlandingtemplate.aspx?id=167&tab=0
). Transfection of HeLa 1.2.11 cells with the corresponding siRNA pools substantially reduced TRF1 and TRF2 protein levels already 24 hours post transfection and protein levels remained almost undetectable throughout the whole 72 hour time course (). Knock-down efficiency was unaltered, even if the transfection medium was changed after 24h (). As the siRNA pools for the actual screen were from the siGENOME library from Dharmacon and not the OTP pools, we also verified siRNA-mediated knockdown of TRF1 and TRF2 using siRNA pools from the siGENOME library. For these controls, we used a 24-well setup similar to the one eventually used for the actual screen (outlined in ). Transfection with the corresponding siRNA pools reduced TRF1 and TRF2 levels respectively, as evaluated by immunofluorescence (IF) 72h after transfection. Knock-down of TRF2 reduced TRF2 nuclear foci, whereas TRF1 was still clearly recognized in telomeric foci (Figure S1
, lower panel). However, knock-down of TRF1 also resulted in a reduced telomeric localization of TRF2 (Figure S1
, middle panel). In contrast, total protein levels of TRF2 were not affected by TRF1 knock-down, as has been observed previously (data not shown) 
Transfection with siRNA pools reduces TRF1 and TRF2 levels and induces a TIF phenotype.
We next tested whether cells transfected with siRNA pools against TRF2 and TRF1 developed TIF as a read out of telomere dysfunction 
. 72 hours after transfection a clear DNA damage response in the form of 53BP1 foci could be observed in cells transfected with siRNA pools against TRF1 or TRF2, whereas no increase in DNA damage foci was observed in cells transfected with a control siRNA against Renilla luciferase (RLUC) (). Furthermore, the DNA damage foci clearly represented TIF, as they were found almost exclusively at telomeres. The TIF response after TRF2 knock-down was even more pronounced than that after the suppression of TRF1 (). This differs from a general DNA damage response that could be seen when cells are exposed to ionizing radiation (IR), where DNA damage foci were randomly distributed throughout the nucleus and did not co-localize with telomeres ().
Based on these initial experiments we set up our screen as outlined in HeLa cells were transfected with siRNA pools on coverslips in a 24-well format and fixed in paraformaldehyde 72 hours post transfection. A combination of IF and Fluorescence In Situ Hybridization (IF-FISH) was then used to stain DNA damage foci and telomeres to detect TIF. On each 24-well plate we included a set of 4 controls: untransfected cells and RLUC siRNA as negative controls, TRF2 OTP pool as positive control, and siGLO transfection indicator from Dharmacon to check for successful transfection of the siRNA (see Material and Methods for details).
Here we want to point out that we detected a low number of cells exhibiting clear TIF in both the RLUC siRNA transfected and untransfected controls. Despite the fact that the overall occurrence of TIF was rare in control cells (), they always appeared in a similar pattern. Within our controls we could detect clusters of cells that exhibited a very low intensity of telomeric FISH signal, which is indicative of short telomeres. Such clusters of cells frequently exhibited a TIF phenotype (Figure S2A
). We speculate that these are occurrences of spontaneous TIF that might arise in sub-populations of cells with short telomeres 
. For our screen we used a clonal HeLa cell line with long telomeres (HeLa1.2.11), however we suggest that some cells in the culture acquire critically short telomeres over time, which attract DNA damage factors.
List of identified candidate factors.
During the actual screen of the 386 candidate factors we determined the number of TIF in up to 196 cells for each individual knock-down. We obtained data for 382 of the 386 genes tested (Table S1
), since four candidates failed to deliver interpretable results due to broken cover slips. Stringent criteria were applied to identify genes as potential candidates for TIF formation. We only considered genes that displayed an average of at least 3 TIF per cell, an arbitrary number that corresponds to the average number of TIF per cell for all genes tested plus two standard deviations and led to the identification of 11positive hits only (, ). Furthermore, between 2–3 TIFs/cell the steepness of the curve in increases, which suggested that this is an appropriate point of separation between noise and real hits. As expected from the bias in selection of the initial set of candidate factors, many of the genes tested exhibited a strong DNA damage phenotype after siRNA-mediated knock-down. Also, the suppression of a number of candidates induced cell death, and as such we could not evaluate the phenotype accurately (data not shown). The analysis of the 11 factors with a clear TIF phenotype was independently repeated twice. In the repetitions all of the hits except YAF2 led to reproducible phenotypes (Table S2
Only one target displayed an overwhelmingly TIF-specifc phenotype. Knock-down of LOC283523 induced a strong DNA damage response, and DNA damage focico-localized exclusively with telomeres (, middle panels). This gene was also the highest scoring gene in the screen (). However, LOC283523 has been annotated as a TRF1-pseudogene (telomeric repeat binding factor [NIMA-interacting] 1 pseudogene; Gene ID 283523). Due to the high sequence identity of this gene to TRF1, the siRNA pool against LOC283523 also contains 2 siRNAs with a perfect match to the bona fide TRF1 sequence. As a consequence, the observed TIF-phenotype is most likely due to suppression of endogenous TRF1 gene products, supported by the finding that TRF1 expression is found down regulated in cells transfected with LOC283523 siGENOME pool (Figure S2B
TIF induction by isolated candidate factors.
All other candidates listed in lead to a weak TIF phenotype, where telomere specific damage foci were observed among a general, non-telomeric DNA damage response. Some of these gene products have well-established roles in DNA metabolism (e.g. RRM1) or replication (POLA, PCNA) and as such it is not surprising that their suppression results in a strong damage response. As a consequence, it is difficult to establish whether the weak TIF phenotype is a direct consequence of the respective knock-down or only occurs as an indirect secondary effect.
Within the set of factors leading to a general DNA damage with a weak TIF phenotype we identified several gene products involved in splicing (NHP2L1, MGC13125, SKIIP, SF3A1) (). Genes associated with splicing and RNA processing have previously been reported to be the most enriched functional group within factors, whose suppression mediates DNA damage 
. Here we focused on SKIIP (also known as SKIP or SNW1; Gene ID: 22938), since this protein has been shown to associate with telomeres in a large-scale proteomics approach, in which whole telomeres were purified and associated proteins were identified by mass spectrometry 
. The knock-down of SKIIP by transfection with the corresponding siGENOME siRNA pool induced general DNA damage in the affected nuclei. However, some of the DNA damage foci were localized at telomeres and thus were considered as TIFs (, bottom panel). To verify this phenotype, we repeated the suppression using the OTP siRNA pool against SKIIP. SKIIP protein levels were clearly diminished 72 hours after transfection with this siRNA pool () and we again observed a few, clear TIFs in the background of a general nucleus-wide DNA damage response ().
Knock-down of SKIIP using OTP siRNA pools.
We next tested whether the SKIIP protein can localize to telomeres by immunofluorescence. Cells were stained with an antibody against SKIIP and an antibody against the telomeric marker TRF2. This approach pointed out that SKIIP was present throughout the nucleus, with less intense staining in the nucleolus () suggesting that the strong expression of SKIIP in the nucleus masked a potential telomere-specific localization. However, when the nuclei were pre-extracted before fixation to remove the soluble protein fraction 
, we could detect distinguished SKIIP foci in the nucleus () and some of these foci co-localized with telomeric signals (), suggesting that a fraction of SKIIP binds specifically to chromosome ends.
Nuclear localization of SKIIP and TRF2.