HEXIM2 - the paralogue that holds the bound P-TEFb pool constant when HEXIM1 is removed, which is why single-gene experiments on this axis read flat
also called HEXIM2 · HIS2
HEXIM2 is HEXIM1's near-twin and does the same job of switching P-TEFb off. When HEXIM1 is taken away, HEXIM2 steps in and keeps exactly the same amount of the enzyme switched off, which is why removing either one alone tells you very little.
The same thing, for a scientist
HEXIM2 (UniProt Q96MH2, 286 aa) binds P-TEFb through a 7SK-mediated interaction and, with 7SK, inhibits CDK9 kinase activity in vitro [E1]. In HEXIM1-knockdown cells HEXIM2 compensates functionally and quantitatively to hold the amount of 7SK/HEXIM-bound P-TEFb constant [E2]. The two form stable homo- and hetero-oligomers through their C-terminal domains, and a HEXIM multimer binds multiple P-TEFb units [E3]. The compensation is not symmetric in every respect: the P-TEFb-release feedback loop induces HEXIM1 transcription and **not** HEXIM2 [E4], and HEXIM1 and HEXIM2 have distinct expression patterns across tissues and cell lines [E1][E2]. In the KMT2A-rearranged panel this graph uses, HEXIM2 is expressed - median 7.2 nTPM, detected in all seven lines - so its near-zero dependency (mean Chronos -0.091 across 1,208 lines, rank 8,290 of 17,728) is **not** an expression artefact [E5][E6]. That is the same test that overturned a paralogue claim in this project before, applied here and passed.
The paralogue split, controlled for expression
This graph has repeatedly found paralogue splits informative and has also learned that a split is only real if the quiet paralogue is actually transcribed. Running the same control here:
| gene | median nTPM, 7 KMT2A-r lines | lines detected | Chronos, all lines | scan rank |
|---|---|---|---|---|
| HEXIM1 | 20.2 | 7/7 | -0.339 | 17,307 |
| HEXIM2 | 7.2 | 7/7 | -0.091 | 8,290 |
Both are expressed. Neither is a dependency. The split is therefore real and uninformative about fitness - which is exactly what the compensation experiments predict: two proteins that substitute for each other will both score near zero in a single-gene screen however important the pathway is. A double knockout has never been reported in any leukaemia cell, and it is the only screen that could speak to this axis.
Where the compensation is incomplete
Two asymmetries are on record and both cut against calling HEXIM2 redundant:
- The autoregulatory loop induces HEXIM1 only. Releasing P-TEFb raises HEXIM1 synthesis from an unannotated proximal promoter and does not raise HEXIM2 4.
- In AML cells, BET inhibition induces HEXIM1 while HEXIM2 levels remain unperturbed, and autophagic degradation in MLL-fusion AML likewise targets HEXIM1 7.
So HEXIM2 is a constitutive buffer and HEXIM1 is the inducible, regulated member. A pharmacology built on induction can only move HEXIM1.
| uniprot | Q96MH2 |
|---|---|
| length AA | 286 |
| gene symbol | HEXIM2 |
| uniprot verified | Q96MH2 = HEXI2_HUMAN, primary gene name HEXIM2, no listed synonyms, 286 aa, Homo sapiens; UniProt REST fetched and read back 2026-09-20 |
| paralogue | HEXIM1 (O94992, 359 aa) |
| breakpoint side | not applicable - no documented KMT2A contact |
paralogue split expression control
{
"question": "Is the flat HEXIM2 dependency an expression artefact?",
"answer": "No - detected in 7/7 KMT2A-rearranged panel lines, median 7.2 nTPM",
"source": "Human Protein Atlas rna_celline.tsv, filtered 2026-09-20",
"precedent": "analyses/expression-control-for-paralogue-splits"
}essentiality — the numbers behind the chart
{
"chronos_mean": -0.0911,
"n_lines": 1208,
"chronos_mean_kmt2ar_myeloid": -0.0903,
"chronos_mean_kmt2ar_lymphoid": -0.1411,
"chronos_mean_other_leukemia": -0.1108,
"scan_rank": 8290,
"scan_n_genes": 17728
}expression hpa celline nTPM — the numbers behind the chart
{
"SEM": 10.4,
"RS4;11": 58.9,
"MV4-11": 9.7,
"THP-1": 6.3,
"NOMO-1": 7.2,
"MOLM-13": 4.1,
"OCI-AML-2": 4,
"median": 7.2
}related entries
[ "gene-hexim1", "protein-7sk-snrnp" ]
depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": -0.116,
"chronos_mean_other_leukaemia": -0.1094,
"difference": -0.0066,
"pan_essentiality_all_lines": -0.0911,
"selectivity_rank": 8290,
"rank_denominator": 17728,
"provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
"how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}- Any HEXIM1-directed intervention is buffered by a paralogue that holds bound P-TEFb constant, and the buffering paralogue is not inducible by the same agents.
- A near-zero CRISPR score for a buffered gene is uninformative in both directions and should not be quoted as evidence that HEXIM2 is dispensable.
- What does a HEXIM1/HEXIM2 double knockout do to a KMT2A-rearranged cell? It has never been reported, and it is the only experiment that can tell this axis from noise.
- Is the HEXIM1:HEXIM2 ratio different in infant KMT2A::AFF1 B-ALL from adult AML, given that the two paralogues have distinct tissue expression patterns?
- Can HEXIM2 be induced by anything? Every known inducer of this axis moves HEXIM1 only.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
HEXIM2 has a regulated association with P-TEFb, forms complexes containing 7SK and P-TEFb, and in conjunction with 7SK inhibits P-TEFb kinase activity; as HEXIM1 is knocked down HEXIM2 functionally compensates for its association with P-TEFb.
HEXIM2, a HEXIM1-related protein, regulates positive transcription elongation factor b through association with 7SK. · Byers SA, Price JP, Cooper JJ, Li Q, Price DH · J Biol Chem 2005What this evidence cannot show
HeLa and Jurkat cells; gradient analysis, EMSA and in vitro kinase assays. The authors' own conclusion is that use of HEXIM1 versus HEXIM2 can be differentially regulated in vivo, which is the opposite of redundancy.
Evidence that argues against this (1)
- The compensation is measured as the amount of P-TEFb in the bound fraction, not as a growth or transcription phenotype, so 'compensates' is a biochemical statement and not a functional equivalence. Yik JH et al. 2005
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In HEXIM1-knocked-down cells HEXIM2 compensates functionally and quantitatively to maintain a constant level of 7SK/HEXIM-bound P-TEFb, and the two proteins show distinct expression patterns across human tissues and established cell lines.
Compensatory contributions of HEXIM1 and HEXIM2 in maintaining the balance of active and inactive positive transcription elongation factor b complexes for control of transcription. · Yik JH, Chen R, Pezda AC, Zhou Q · J Biol Chem 2005What this evidence cannot show
Knockdown rather than knockout, in HeLa; the distinct expression patterns are shown by blot across a tissue panel, not quantified.
Evidence that argues against this (1)
- If the compensation were complete, HEXIM1 manipulation would have no phenotype - and in MOLM-13 it clearly does, with knockdown raising c-Myc and growth and ectopic expression suppressing growth. The buffering is therefore partial in at least one KMT2A-rearranged cell. Devaraj SG et al. 2016
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
HEXIM1 and HEXIM2 interact directly through their C-terminal domains and form stable homo- and hetero-oligomers, and multiple P-TEFb modules bind a single HEXIM multimer.
Transcription-dependent association of multiple positive transcription elongation factor units to a HEXIM multimer. · Dulac C, Michels AA, Fraldi A, Bonnet F, Nguyen VT, Napolitano G, Lania L, Bensaude O · J Biol Chem 2005What this evidence cannot show
Yeast two-hybrid, transfection, hydrodynamics and cross-linking; no structure. Whether HEXIM1 and HEXIM2 occupy the same particle in vivo is not established.
Evidence that argues against this (1)
- No structure of the assembled 7SK snRNP exists, so the stoichiometry is model-dependent and should not be quoted as a fixed number. Pham VV et al. 2022
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Release of P-TEFb from the 7SK snRNP increases synthesis of HEXIM1 but not HEXIM2.
Release of positive transcription elongation factor b (P-TEFb) from 7SK small nuclear ribonucleoprotein (snRNP) activates hexamethylene bisacetamide-inducible protein (HEXIM1) transcription. · Liu P, Xiang Y, Fujinaga K, Bartholomeeusen K, Nilson KA, Price DH, Peterlin BM · J Biol Chem 2014What this evidence cannot show
HeLa cells; measured after treatment with P-TEFb-releasing compounds rather than a physiological stimulus.
Evidence that argues against this (1)
- HEXIM2 transcription may be regulated by stimuli this study did not use; an absence of induction by one class of compound is not evidence that HEXIM2 is unregulated. Byers SA et al. 2005
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
HEXIM2 is detected in all seven KMT2A-rearranged panel lines the Human Protein Atlas carries, median 7.2 nTPM, so its near-zero dependency score is not explained by the gene being silent.
Human Protein Atlas cell-line RNA nTPM, rna_celline.tsv (24,315,372 rows, 1,206 cell lines), downloaded and filtered for this beat on 2026-09-20 · Human Protein Atlas · Database 2026What this evidence cannot show
Bulk poly-A cell-line RNA. RS4;11 returns 58.9 nTPM against 4.0-10.4 in the other six and is a high outlier for several unrelated genes in the same pull, so the median rather than the mean is quoted.
Evidence that argues against this (1)
- mRNA detection does not establish that HEXIM2 protein is present at a level that could buffer HEXIM1 loss in these particular cells; no blot exists for any KMT2A::AFF1 line. Human Protein Atlas 2026
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
HEXIM2 has mean Chronos -0.091 across 1,208 DepMap lines and ranks 8,290th of 17,728 on index-minus-comparator difference, with the KMT2A-rearranged myeloid half at -0.090 and the lymphoid half at -0.141 against -0.111 in other leukaemia lines.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API, re-queried for this beat on 2026-09-20 and split by lineage inside the KMT2A-rearranged arm; genome-scale ranks read from this repository's analyses/depmap-kmt2a-selective-scan-full/results.json (17,728 genes, 1,208 lines) · Broad Institute DepMap; Open Targets Platform; Oncofusion KG internal scan · Databases 2026What this evidence cannot show
Six myeloid and four lymphoid lines. All values are near zero, so this is a statement that nothing happens rather than a measured effect. A single-gene CRISPR screen cannot detect a function buffered by a paralogue, which is precisely the situation here.
Evidence that argues against this (1)
- The compensation literature predicts exactly this null, so a null here is consistent with HEXIM2 mattering a great deal and with it mattering not at all. Only a HEXIM1/HEXIM2 double knockout would separate them, and none has been reported in leukaemia. Yik JH et al. 2005
- E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In AML cells, BET inhibition induces HEXIM1 while HEXIM2 levels remain unperturbed.
HEXIM1 induction is mechanistically involved in mediating anti-AML activity of BET protein bromodomain antagonist. · Devaraj SG, Fiskus W, Shah B, Qi J, Sun B, Iyer SP, Sharma S, Bradner JE, Bhalla KN · Leukemia 2016What this evidence cannot show
MOLM-13 and one patient-derived AML sample; read from the PMC full text because the PubMed record carries no abstract. HEXIM2 was assayed by immunoblot as a control rather than as the subject of the study.
Evidence that argues against this (1)
- An unchanged total protein level does not exclude a change in how much HEXIM2 is loaded onto 7SK, which is the quantity that matters and which was not measured. Yik JH et al. 2005
reversibly associated inhibitory subunit; buffers HEXIM1 loss
paralogue; compensates quantitatively for HEXIM1 loss, and both are expressed in all seven KMT2A-rearranged panel lines
inhibits P-TEFb kinase activity in conjunction with 7SK
supplies the epigenetics record gene-hexim2 lacks, and the experimental-design consequence of the buffering