Oncofusion

start / graph / ana-canela-2017-deposited-no-cohesin-arm-and-two-unread-human-end-seq-arms

Analysisestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q107-break-machinery-deposition

Canela 2017 deposited NO cohesin-depletion experiment - and it did deposit human END-seq in a B-cell-precursor leukaemia line, which nobody here had looked for

also called GSE99197 audit · GSE99194 END-seq manifest · the Nalm6 arm nobody read · Q107 deposition audit, 2017

This project believed the 2017 paper had removed a cohesin ring subunit and mapped the breaks. It did not - that experiment is in the 2019 paper. What the 2017 paper does have, and nobody here had noticed, is a break map in a human B-cell leukaemia line, which is much closer to the disease this lab studies.

The same thing, for a scientist

PMID 28735753 (Canela et al., Cell 2017) deposited SuperSeries **GSE99197** (61 samples) with SubSeries **GSE99194** [END-seq, 32 samples, *Homo sapiens*/*Mus musculus*/*Mus spretus*], **GSE99195** [ChIP-seq, 18 samples], **GSE99196** [nsRNA-seq, 11 samples]. **Two corrections to the state of belief in this graph.** (1) **There is no cohesin-depletion arm anywhere in the 2017 deposit.** GSE99195's 18 ChIP samples are CTCF, RAD21, TOP2A and TOP2B in mouse activated B, resting B, T cells and MEFs plus *Mus spretus* activated B - **zero SMC3 samples, zero SMC1 samples, zero human samples** - and the END-seq perturbations are `MEF_top2ko`, DRB, triptolide and the *Mus musculus*/*Mus spretus* CTCF-polymorphism comparison. The `Smc3f/f+CRE` MEF END-seq is **GSM3714900 in the 2019 deposit (GSE129524)**, alongside `Waplf/f+CRE` GSM3714902 [E1][E2]. (2) **The 2017 END-seq series contains two human arms - `END_seq_Nalm6_ETO`/`END_seq_Nalm6_NT` (GSM2635574/5) and `END_seq_MCF7_ETO`/`END_seq_MCF7_NT` (GSM2635568/9) - aligned to GRCh37/hg19.** Nalm6 is a human B-cell-precursor acute lymphoblastic leukaemia line. This is the closest thing in the whole loop-anchor fragility literature to a B-lineage human break map, and reading it puts the *KMT2A* exon-11 CTCF element at the **99.948th percentile of chr11** (see `ana-kmt2a-exon-11-element-is-a-top-0-05-percent-etoposide-dsb-site`) [E3].

Go deeper
The manifest

GSE99197 (SuperSeries) - 61 samples; Mus musculus; Homo sapiens; Mus spretus. Its own supplementary directory contains only GSE99197_RAW.tar (15,820,533,760 bytes) - no bigWig, no peak file, no table 1.

SubSeriesassaysamplesprocessed types depositedRAW.tar bytes
GSE99194END-seq3232 x .bw only9,486,120,960
GSE99195ChIP-seq18.bw only-
GSE99196nsRNA-seq11.bw only-

The 32 END-seq samples by organism 1:

organismnsamples
Mus musculus26aB12h, aB24h, DRBexp_aB12h (4), MEF, MEF_top2ko, NRN, PB, rB, T, Species_Mus_aB12h, TRPexp_aB12h (4), each with ETO and NT
Homo sapiens4END_seq_MCF7_ETO, END_seq_MCF7_NT, END_seq_Nalm6_ETO, END_seq_Nalm6_NT
Mus spretus2Species_Spr_aB12h_ETO, Species_Spr_aB12h_NT

The 18 ChIP-seq samples, in full, because the absence is the point 2: ChIP_seq_CTCF_aB, CTCF_MEF, CTCF_rB, CTCF_Spr_aB, CTCF_T, RAD21_aB, RAD21_MEF, RAD21_rB, RAD21_Spr_aB, RAD21_T, TOP2A_aB12h, TOP2A_aB24h, TOP2A_MEF, TOP2A_MEF_top2bKO, TOP2B_aB12h, TOP2B_aB24h, TOP2B_MEF, TOP2B_MEF_top2bKO. Every one is Mus musculus except the two Spr samples (Mus spretus). No SMC3, no SMC1, no NIPBL, no WAPL, no human.

Where the cohesin-loss experiments actually live

This project's dispatch brief for this beat asserted that Canela 2017 deposited "its mouse B-cell/MEF SMC3-CRE END-seq". It did not, and the assertion is checkable in one query. The three loss- or gain-of-cohesin END-seq arms in this literature are all in GSE129524, the 2019 deposit 2:

GSMtitlegenotype fieldspeciesdirection
GSM3714899END_seq_MEF_Smc3_PP_0_5hWTmousecontrol
GSM3714900END_seq_MEF_Smc3_FF_0_5hSmc3f/f+CREmousecohesin loss
GSM3714901END_seq_MEF_Wapl_FF_0_5h (control)Waplf/fmousecontrol
GSM3714902END_seq_MEF_Wapl_FF_CRE_0_5hWaplf/f+CREmousecohesin retention up
GSM3714903END_seq_HCT116_0_5hWThumancontrol
GSM3714904END_seq_HCT116__RAD21_AID_0_5hRAD21 AIDhumancohesin loss

All six are n = 1. The 2017 paper's contribution to the occupancy argument is a natural perturbation - CTCF-site polymorphisms between Mus musculus and Mus spretus that redistribute occupancy - not a depletion.

Why this matters beyond bookkeeping. The graph's mechanism-loop-anchor-dsb-hotspots treats 2017 and 2019 as two studies with a shared method and a disagreement about transcription. The deposition audit sharpens that: 2019 is the paper with the directed cohesin experiments and 2017 is the paper with the cell-type and species breadth. An argument about whether cohesin causes the breaks has to be made from 2019, and 2019's human arm is a single colorectal-carcinoma library.

The Nalm6 arm is the nearest thing to a B-lineage human break map that exists, and it was never named in this graph

Nalm6 is a human B-cell-precursor acute lymphoblastic leukaemia line (peripheral blood, 19-year-old male, ETV6::PDGFRB). Its END-seq etoposide/untreated pair is GSM2635574/GSM2635575, aligned to GRCh37/hg19, deposited as 335,785,861-byte and 260,214,650-byte bigWigs 3.

It was invisible to every query this graph has run, and the reason is exactly the one the dispatch brief flagged in general terms: the 2017 abstract and title name neither KMT2A nor any cell line, so no gene-scoped and no cell-line-scoped PubMed query can reach it. The route that did reach it is elink dbfrom=pubmed db=gds on the PMID - four GEO series, one of which is END-seq - followed by reading the per-sample organism field. That route costs two API calls and should be run on every mechanism paper this graph cites, before any bounded absence about deposited data is written.

Caveat that is not small: Nalm6 is a leukaemia line that already carries a fusion, is not infant, is not KMT2A-rearranged, and is B-precursor rather than fetal-liver HSC. It is a proxy, and it is the best one in this literature.

Facts and figures
scriptmanifest audit via eutils elink/esummary and GEO acc.cgi text views; no committed script
statusrun
result summaryNo cohesin-depletion arm exists in the 2017 deposit; GSE99195 is CTCF/RAD21/TOP2A/TOP2B in mouse and Mus spretus only, with zero human and zero SMC3/SMC1 samples. The Smc3f/f+CRE and Waplf/f+CRE END-seq arms are in the 2019 deposit (GSM3714900, GSM3714902). The 2017 END-seq series does contain four human samples - Nalm6 and MCF7, each ETO and NT, hg19, bigWig only - and the Nalm6 pair is the only B-lineage human break map in this literature.
genome build depositedGRCh37/hg19 for human, GRCm38p2/mm10 for mouse
brief premise correctedThe dispatch brief stated that Canela 2017 deposited 'its mouse B-cell/MEF SMC3-CRE END-seq'. It did not. There is no SMC3 sample of any assay type in the 2017 deposit; the Smc3f/f+CRE MEF END-seq is GSM3714900 in the 2019 deposit.
inputs
[
  "GEO GSE99197 (SuperSeries, PMID 28735753), 61 samples",
  "GEO GSE99194 [END-seq], 32 samples, filelist.txt",
  "GEO GSE99195 [ChIP-seq], 18 samples, full sample listing",
  "GEO GSE99196 [nsRNA-seq], 11 samples",
  "GEO GSE129524 GSM3714899-GSM3714904 for the comparison"
]
accessions
[
  "GSE99197",
  "GSE99194",
  "GSE99195",
  "GSE99196",
  "GSM2635568",
  "GSM2635569",
  "GSM2635574",
  "GSM2635575",
  "SRX2843836"
]
How this could bite us back
  • The corrected picture makes the evidence base for 'cohesin occupancy causes breaks at KMT2A' thinner than the graph assumed, not thicker: one unreplicated human library in a colorectal line, plus two unreplicated mouse fibroblast libraries. Any intervention argument resting on that chain should say so in its first sentence.
Still open
  1. What etoposide dose and duration were used for Nalm6 and MCF7? The GSM records quote a shared treatment protocol that does not name the human arms' conditions, so the human END-seq arms of 2017 cannot currently be matched in dose to the 2019 HCT116 arm.
  2. Does Canela 2017's paper contain a cohesin manipulation that left no GEO footprint? This graph's mechanism-loop-anchor-dsb-hotspots attributes a 'requires pre-existing cohesin binding' claim to the pair of papers; establishing which paper it is in needs the 2017 full text, which is open access at PMC6133249 and was not read here.
  3. Nalm6 is available and widely used. Would a RAD21 or SMC3 degron in Nalm6, with END-seq or sBLISS, be a tractable single-experiment answer to premise (i)? Nothing in the three deposits attempts it.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    PMID 28735753 deposited GEO SuperSeries GSE99197 (61 samples) whose own supplementary directory contains only GSE99197_RAW.tar (15,820,533,760 bytes), with SubSeries GSE99194 [END-seq, 32 samples, deposited as 32 bigWig files plus GSE99194_RAW.tar of 9,486,120,960 bytes], GSE99195 [ChIP-seq, 18 samples, bigWig only] and GSE99196 [nsRNA-seq, 11 samples, bigWig only]; of the 32 END-seq samples exactly four are Homo sapiens - END_seq_MCF7_ETO, END_seq_MCF7_NT, END_seq_Nalm6_ETO, END_seq_Nalm6_NT - two are Mus spretus and twenty-six are Mus musculus.

    What this evidence cannot show

    This is a manifest audit. No RAW.tar was opened, so a peak-call file could sit inside one of the archives. The GSE99195 and GSE99196 per-file listings were read via the series record rather than a separate filelist fetch, so their exact file sizes are not quoted here; their deposited processed type (BW) comes from the esummary suppfile field and the series record. Sample counts are GEO's, and the END-seq series' 32 samples map exactly to 32 deposited bigWigs with no merged pileups.

    Evidence that argues against this (1)
    • The 2017 deposit not containing peak calls is a weak criticism for the same reason as in the 2019 audit - bigWig is the complete quantitative record. And the absence of a cohesin-depletion arm is not a deficiency of the 2017 paper, whose designed perturbation was a natural one: CTCF-motif polymorphisms between Mus musculus and Mus spretus, deposited as GSM2635582-GSM2635585 END-seq plus ChIP_seq_CTCF_Spr_aB and ChIP_seq_RAD21_Spr_aB. That is arguably a cleaner test of occupancy-versus-breakage than an acute degron, because it changes occupancy at individual sites without changing the global cohesin pool. NCBI Gene Expression Omnibus 2017
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GSE99195, the only ChIP-seq series of PMID 28735753, contains exactly 18 samples - CTCF in mouse activated B, MEF, resting B and T plus Mus spretus activated B; RAD21 in the same five; TOP2A in activated B at 12 h and 24 h, MEF and Top2b-knockout MEF; and TOP2B in the same four - with no SMC3, SMC1, NIPBL or WAPL sample and no Homo sapiens sample; whereas the cohesin loss- and gain-of-function END-seq arms of this literature are GSM3714900 (END_seq_MEF_Smc3_FF_0_5h, genotype Smc3f/f+CRE), GSM3714902 (END_seq_MEF_Wapl_FF_CRE_0_5h, genotype Waplf/f+CRE) and GSM3714904 (END_seq_HCT116__RAD21_AID_0_5h, genotype RAD21 AID), all three deposited in GSE129524 with PMID 31202577.

    What this evidence cannot show

    This bounds the DEPOSIT, not the paper: an experiment performed and reported only as a blot or an image would leave no GEO trace, so 'no SMC3 sample in GSE99195' is not the same claim as 'Canela 2017 performed no SMC3 experiment'. The full text of PMID 28735753 was not read for this entry, so the paper-level claim is deliberately not made. The genotype strings are the submitters' free text.

    Evidence that argues against this (1)
    • This graph's own entry mechanism-loop-anchor-dsb-hotspots reports that Canela 2017 showed loop-anchor breakage requires pre-existing cohesin binding. If that is in the 2017 paper rather than the 2019 one, then a cohesin manipulation of some kind exists in 2017 without a GEO footprint - which is precisely why this item is scoped to the deposit. Resolving it needs the 2017 full text, which this beat did not read. Canela A et al. 2017
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GSM2635574 (END_seq_Nalm6_ETO) and GSM2635575 (END_seq_Nalm6_NT) are Homo sapiens END-seq libraries from Nalm6, genotype WT, aligned to GRCh37/hg19 with bowtie 0.12.7 and deposited as bigWig coverage tracks of 335,785,861 and 260,214,650 bytes; their SRA experiment for the etoposide arm is SRX5657096-independent accession SRX2843836, and the deposited data processing states peaks were called with --nolambda --nomodel --keep-dup=all against the matched non-treated sample keeping >10-fold enriched peaks, but no peak file is deposited.

    What this evidence cannot show

    n = 1 for each of the etoposide and untreated arms; no replicate exists. The etoposide dose and duration for the Nalm6 arm are not stated in the fields quoted here. Nalm6 is a cultured B-cell-precursor ALL line carrying its own rearrangement and is neither infant-derived nor KMT2A-rearranged, so it is a lineage proxy and not a model of the index disease. The sample's own record does not name a cell-cycle synchronisation step, unlike the HCT116 arm of the 2019 series.

    Evidence that argues against this (1)
    • Calling Nalm6 'the nearest thing to a B-lineage human break map' may understate what Gothe 2019 offers: TK6 is a human lymphoblastoid line with TWO END-seq-equivalent replicates plus isogenic repair mutants, and Gothe additionally has primary human CD34+ bone-marrow progenitors with two replicates. On replication and on primary-cell grounds the Gothe deposit is stronger; Nalm6's advantage is only that it is B-precursor-committed and that it comes from the END-seq side of the dispute. NCBI Gene Expression Omnibus 2019
Connected to