THE DEVELOPMENTAL HANDOVER
SCHEMATICCHROMOSOME 11 · THE β-LIKE GLOBIN LOCUS
One locus.
A lifetime of expression.
Follow the handover from embryonic ε to fetal γ to adult β.
Then see where the Maltese KLF1 families depart from it.
“
Titus H. J. Huisman and Alex E. Felice.
From the haemoglobin variants to the mechanism.
With a historical thread through the work ofTitus H. J. Huisman and Alex E. Felice.
SELECT HEP DONOR
DEVELOPMENTAL SCHEMATIC
Fetal erythropoiesis
5′ → 3′ transcriptional order · reverse strand · spacing schematic Fetal γ Adult β Embryonic ε
REGULATOR RNA · CHANGES WITH THE CULTURE SLIDER
Fold vs the same-time low reference, combined nine-gene normalization. Arrows compare the relative fold at T48 with T0, not absolute RNA per cell. KLF1 is total RNA here; no TF protein activity is inferred.
SELECT A GENE ON THE LOCUS
lncRNA locusBGLT3
OBSERVED FAMILY DATA
The same mutation.
Different expression states.
M2 versus M4/M5 within FamF. M8 is a separate-family, cultured-HEP comparator.
LOCUS RNA
MEASURED ENDPOINTSWhere the high group stands out
Each cell is fold relative to the same-time low-reference geometric mean, for that gene and selected processing view. Colour is log₂ fold, clipped at ±5; text retains the value. Click a gene name to inspect it on the locus.
ACCESSIBILITY IS A SEPARATE MEASUREMENT
Promoter ATAC
Values: inherited log₂(fragment CPM + 0.5), usable-fragment normalization. Higher ATAC does not identify a bound factor or prove a loop. The RNA ordering is not reproduced by every promoter window.
TRANS-ACTING REGULATORS
RNA, not protein activity
Same-time low-reference fold within each gene, using the selected reference panel. The counting-view selector applies to locus RNA only. No unmeasured TF protein is inferred from RNA.
SPECIMENS MATTER
Blood HbF and cultured HbF
Protein corroboration, in a separate cohort
Family-adjusted carrier/control intensity ratios, 22 donors. These are not HbF percentages or measurements of the four HEP donors. KLF1, BCL11A, LRF, SOX6 and ZFP57 were not quantified in this panel.
What DNA narrows—and leaves open
PUBLIC HUMAN ERYTHROID DATA
Development,
with measurements attached.
Two independent studies. Compare sources within each study and matched culture day.
INDIVIDUAL LIBRARIES
A developmental programme is more than a single RNA gradient
KLF1 RNA does not show a universal fetal-to-adult rise in these panels. BGLT3 and HBBP1 also have different developmental patterns. The Maltese high-HbF state should not be labelled a complete return to a fetal state.
No fetal, cord and adult samples from different studies are pooled into one calibrated curve. Culture days are maturation times, not gestational or postnatal ages. A zero adult baseline is displayed as unavailable for fold calculations.
MOUSE MODELS · CROSS-SPECIES EVIDENCE
Shared regulators.
Different developmental clocks.
From the original EKLF knockouts to human-locus mice and public expression datasets. Compare what each model actually tests.
Which globin locus?
Which cell population?
Which perturbation?
DEVELOPMENTAL PROGRAMMES · ILLUSTRATIVE
Export comparison ↗Follow three biological contexts.
Steps compare developmental programmes; they are not matched ages, measured time courses or a conversion between human and mouse gestation. Gene shading is qualitative. Select a stage or press Play.
Native mouse ≠ fetal human
Native mouse β-like globins mainly move from embryonic to adult expression. There is no prolonged human-like fetal γ phase.
The human locus changes the question
Human β-globin transgenes let a mouse test human γ repression. The host’s developmental environment and the construct still matter.
Dosage and cell state matter
A complete Klf1 knockout causes severe erythroid failure. It is not the same perturbation as heterozygous K288X in the Maltese families.
What happens when a regulator is removed?
CONNECT TO THE MALTESE FAMILIES
Compare directions. Preserve the experiments.
The mouse interventions test regulatory capability. The family measurements describe variation among cultured HEPs. The data explorer places their within-study contrasts side by side without pooling their values.
Inspect individual samples and probe annotations
CROSS-SPECIES COMPARISON · SEPARATE DENOMINATORS
Mouse intervention and FamF observations
Mouse perturbation or lineage contrasts and human M2 / geometric mean of M4–M5 answer different biological questions. Family RNA uses combined-nine reference normalization. Matching a direction does not establish the family modifier. Mouse globin genes are not silently substituted for HBG1/HBG2.
Keep the measurements and their provenance.
THE REGULATOR NETWORK · EVIDENCE BY CONTEXT
How far does each candidate travel?
“Not catalogued” means no qualifying result is included in this curated set; it does not mean the gene has no effect. A role in erythropoiesis alone is distinguished from a demonstrated globin-switch phenotype.
HOW THE PIECES CONNECT
Activation. Repression.
Contact. Translation.
Explore published perturbations and natural variants. Each link carries its experimental scope.
Explore the repression machinery in motion. MYB → KLF1 → BCL11A, NuRD modules and distinct routes to disruption.
THE REGULATORY NETWORK
CLICK A LINK CARDA connected model, with evidence attached
Solid links: perturbation-supported in published systems. Dashed link: proposed family interaction. Neither style denotes measured occupancy in these HEPs.
γ-promoter anchors: sequence is not occupancy
FROM REGULATOR TO CHROMATIN · A GUIDED MOLECULAR MODEL
Build the complex.
Find the point of release.
Follow MYB and KLF1 to BCL11A, then explore how NuRD assembles, engages the γ-globin promoter and can be uncoupled.
One mechanism.
Several distinct layers.Gene regulation → protein interactions → chromatin activity → globin output. Click a component to follow its evidence.
Several distinct layers.Gene regulation → protein interactions → chromatin activity → globin output. Click a component to follow its evidence.
THE UPSTREAM AXIS · REGULATORY ARROWS
→supports expression
→transcriptional activation
KLF1 → HBB
A second arm supports adult β-globin directly.
BCL11A + NuRD ⊣ HBG1 / HBG2
DNA recognition and corepressor engagement support repression.
Show the parallel routes
MYB → TR2 / TR4 (NR2C1 / NR2C2): a DRED-associated branch, alongside effects on erythroid cell state. The KLF1 branch is supported in human cells and is not universal across species.
LRF / ZBTB7A → NuRD-associated repression: a route independent of BCL11A. This network is not a single linear switch.
Suzuki et al., 2013 ↗ Masuda et al., 2016 ↗GUIDED WALKTHROUGH
Support repression
Illustrative positions, interactions and subunit families; no molecular scale or fixed assembly order. Scroll the figure on small screens. All components also have buttons below.
The parts, their roles, their evidence.
Module connectionRegulator engagementDisrupted interface
CONNECT TO YOUR MEASUREMENTS
OBSERVEDRead the RNA alongside the model.
Within-gene fold vs the same-time geometric mean of M3 / M7 / M9; combined nine-gene normalization. T0 and T48 are culture endpoints. Missing genes remain unavailable.
RNA does not measure complex formation. Protein abundance, interactions, promoter occupancy and enzyme activity require their own evidence. The animation is not fitted to these values.
HU links use historical array annotations. FamF PC/PH and Greek R/NR comparisons remain separate cohorts; these are not the RNA-seq T0/T48 samples.
NuRD is a family of complexes.
| Layer | Components | How to read this model |
|---|---|---|
| Deacetylase core | MTA1/2/3 · HDAC1/2 · RBBP4/7 | The main drawing retains family labels. The erythroid evidence particularly highlights MTA2, HDAC2 and RBBP4, with distinct repression and fitness findings. |
| Connector and bridge | MBD2 or MBD3 · GATAD2A/B | MBD2- and MBD3-containing assemblies are alternatives. The HbF-focused view uses MBD2a and GATAD2A. |
| Remodeling arm | CHD3/4/5 · CDK2AP1/2 | CHD4 is the erythroid focus. CDK2AP partners are included in the component explorer; no family-specific HbF effect is assigned to them. |
| Recruiters and partners | BCL11A · LRF/ZBTB7A · PRMT5/MEP50 | Recruiters and associated machinery are distinguished from core subunit families. This is not a claim that every partner occupies one complex simultaneously. |
NuRD = nucleosome remodeling and deacetylase. This is an HbF-focused synthesis, not an exhaustive catalog of every tissue-specific interactor. Abundance, assembly, recruitment and catalytic activity can change independently.
Follow each connection back to an experiment.
The walkway is an explanatory synthesis. “Uncoupling” describes disrupted protein interfaces; the sequence of frames is not a measured movie of physiological assembly or disassembly.
A CONNECTED HISTORY OF HAEMOGLOBIN RESEARCH
From the molecule
to the Maltese families.
Follow the experiments that connected haemoglobin chemistry, inherited variation and gene regulation—with a dedicated thread through Alex E. Felice and Joseph Borg’s work.
IN MEMORY OF
ALEX E. FELICE
ALEX E. FELICE
From Malta, Augusta and Caltech to the molecular genetics of the Maltese families: the work of Felice, Huisman and Schroeder helped make haemoglobin variation an instrument of discovery.
Read Felice’s historical account, 2023 ↗A curated chronology, reviewed 23 October 2026. Original studies, reviews, resources and conference abstracts are labelled separately. Sequence shows the development of ideas; it does not imply that every later study directly derives from an earlier one. Publication dates are retained, including genuine intervals between studies.
TRACEABLE BY DESIGN
Follow every claim
back to its evidence.
Measured data, published mechanisms and the explanatory animation remain identifiable throughout.
Core atlas dataJSON · measurements, evidence, provenanceFamily locus RNACSV · all 18 processing combinationsLocus accessibilityCSV · retained windows and normalizationsRegulator RNACSV · all retained reference panels
You are using the offline atlasSelf-contained HTML · keep this file private
Mouse ModelsJSON · comparative evidence and measurementsNuRD mechanism & evidenceJSON · components, states, primary sourcesWhat this atlas establishes
The family data support a perturbed adult programme and a graded BGLT3/HBE1 RNA state. Independent perturbations establish regulatory capabilities. Together they support a mechanistic interpretation; they do not uniquely identify the cause of between-carrier variation or close every step of the universal switch.
Primary evidence index
RECOVERED EXPERIMENTS · JAN / MAR 2009 · RMA REPROCESSING 2012
The same cells.
A drug perturbation.
Paired untreated → hydroxyurea HEP expression. Compare inherited KLF1 insufficiency with responder and nonresponder patient cultures.
FAMF12
paired donors · 24 arrays
GREEK / CYPRUS15
paired patients · 30 arrays
PRIMARY COMPARISON4 × 4
WT / KLF1 / R / NR donors
EACH MATRIX54,675
probe sets · original RMA log₂
Confirmed treatment key: PC = untreated; PH = HU. Here every treatment effect is HU − untreated. Positive means higher with HU. These experiments are separate from the atlas’s RNA-seq T0/T48 series.
INDIVIDUAL DONORS · PAIRED DESIGN
Loading recovered expression…
Lines join the same donor’s untreated and HU cultures. RMA log₂ signal; compare within a probe and cohort. Group effects use only the four prespecified donors. Separate batches are not pooled. p values are exploratory; BH q values account for all finite probe tests in each contrast.
What these arrays can measure
Probe assignments are historical source annotations, not current sequence-remapped validation. HBG/HBB saturation and homologous-probe cross-hybridization limit quantitative interpretation. Borg et al., 2010 explicitly discuss globin saturation ↗ No probe assignment means unestablished coverage, not zero expression. Enhancers, promoter sites and chromatin contacts have no new measurement in this array block.
Blood response and cultured-cell response
Values retained as recorded in the clinical workbook’s HbF fields. Blood baseline / semester and culture pre / post are distinct specimens. R/NR comes from the explicit historical groups, not a new threshold. Numeric patient IDs agree between the two source workbooks; original specimen identity remains a provenance assumption.
A perturbation layer, with boundaries
Ask whether HU changes a regulator in the same direction in WT, KLF1, responders and nonresponders. The interaction compares paired changes, rather than untreated group differences.
Different genotypes, maturation states, culture conditions and small groups can influence the contrasts. A shared RNA response does not establish that the regulator causes HbF induction.
Do not equate George 8 with M8. The original FamF numbering has not been linked to the modern M-series donors. The Malta metadata also labels George 15 wt/wt despite “Middle HbF”. These labels are retained as supplied.
Original filenames and assignments
Seven extra Greek pairs retain R? / NR? labels, and four extra FamF pairs have no genotype assignment in the sample workbook. They are excluded from primary group statistics. HU dose and exposure duration remain unresolved.
Recovered source data and analysis
The TXT attachments contain processed expression matrices referencing CEL filenames. The raw CEL intensity files themselves have not been supplied. Legacy filtered lists are retained for annotation and audit; their “up/down” labels are not substituted for recalculated paired HU effects.
Complete HU matrices, group summaries and provenance downloads are available from the hosted atlas. A focused globin, regulator and stress-response panel is embedded in this offline copy; its plots and CSV exports work offline. The hosted atlas searches all 54,675 probes.