Chemical Replacement of Noggin with Dorsomorphin Homolog 1 for Cost-Effective Direct Neuronal Conversion

The direct conversion of adult human skin fibroblasts (FBs) into induced neurons (iNs) represents a useful technology to generate donor-specific adult-like human neurons. Disease modeling studies rely on the consistently efficient conversion of relatively large cohorts of FBs. Despite the identification of several small molecular enhancers, high-yield protocols still demand addition of recombinant Noggin. To identify a replacement to circumvent the technical and economic challenges associated with Noggin, we assessed dynamic gene expression trajectories of transforming growth factor-β signaling during FB-to-iN conversion. We identified ALK2 (ACVR1) of the bone morphogenic protein branch to possess the highest initial transcript abundance in FBs and the steepest decline during successful neuronal conversion. We thus assessed the efficacy of dorsomorphin homolog 1 (DMH1), a highly selective ALK2-inhibitor, for its potential to replace Noggin. Conversion media containing DMH1 (+DMH1) indeed enhanced conversion efficiencies over basic SMAD inhibition (tSMADi), yielding similar βIII-tubulin (TUBB3) purities as conversion media containing Noggin (+Noggin). Furthermore, +DMH1 induced high yields of iNs with clear neuronal morphologies that are positive for the mature neuronal marker NeuN. Validation of +DMH1 for iN conversion of FBs from 15 adult human donors further demonstrates that Noggin-free conversion consistently yields iN cultures that display high βIII-tubulin numbers with synaptic structures and basic spontaneous neuronal activity at a third of the cost.


Introduction
A ging is one of the most significant risk factors for many neurodegenerative diseases, and establishing predictive models for disease modeling and drug discovery that reflect aspects of old adult human brain cells is a challenge. The majority of cell-based models rely on differentiation from hu-man induced pluripotent stem cells (iPSCs), but during reprogramming of human somatic cells into iPSCs, the cells undergo a process of global cellular rejuvenation to early embryonic stages and lose signatures of aging (Lapasset et al., 2011;Miller et al., 2013;Sardo et al., 2017;Takahashi et al., 2007).
While several protocols for direct conversion of FBs to iNs exist, most studies utilize the overexpression of pioneer transcription factors such as Achaete-scute homolog 1 (Ascl1) and Neurogenin 2 (Ngn2), which together have been shown to yield the highest and most reliable conversion efficiencies in combination with cocktails of small molecular pathway modulators (Ladewig et al., 2012;Liu et al., 2013). Most small molecules result in stimulation of proneuronal pathways such as cyclic AMP signaling, and downregulation of antineuronal pathways such as glycogen synthase kinase 3 beta (GSK3b), hypoxia-inducible factor 1alpha (HIF1-a), integrin, or signal transducer and activator of transcription 3 signaling (STAT3) (Herdy et al., 2019;Ladewig et al., 2012;Liu et al., 2013).
One of these antineuronal pathways is transforming growth factor-b (TGFb) superfamily signaling, which uses the tyrosine kinase receptors of the anaplastic lymphoma kinase (ALK) family to phosphorylate SMAD proteins to signal to the nucleus (Miyazono, 2000). TGFb signaling can be divided into a TGFb1 arm (ALK4/5/7 and SMAD2/3) and a bone morphogenic protein (BMP) arm (ALK2/3/6 and SMAD1/5/8). The proneuronal activity of the inhibition of both arms was first demonstrated in a landmark discovery by Chambers et al. (2009), describing that inhibition by the small molecule SB-431542 (ALK4/5/7 inhibitor) and the BMP antagonist Noggin induces highly efficient neuralization of human iPSCs, a method known as dual-SMAD inhibition (Chambers et al., 2009), and that strategy has been quickly adopted also for direct iN conversion (Ladewig et al., 2012;Liu et al., 2013;Vadodaria et al., 2016).

Results and Discussion
High FB abundance and pronounced repression of ALK2-mediated BMP signaling during successful iN conversion Because TGFb superfamily signaling pathways are instrumental in orchestrating the neuronal conversion process in various in vivo and in vitro systems (Falk et al., 2008;Li et al., 2011;Yin et al., 2019), we investigated the dynamic transcriptional trajectories of TGFb signaling component genes during reprogramming in tSMADi plus Noggin condition (+Noggin). For this, we assessed gene expression trajectories of FB-to-iN conversion based on time series whole-transcriptome RNA-Seq data from FBs and converting cultures from three adult donors for 5, 10, 15, and 20 days (Herdy et al., 2019).
Comparison of FB and day 20 iN cultures identified 4863 highly significant differentially expressed genes (DEGs; padj <0.01). Of these 4863 DEGs, 2075 became downregulated and 2788 became upregulated during conversion, and the majority of which revealed a consistent and gradual repression or activation over time (Fig. 1A). Principal component analysis (PCA) of selected neuronal genes (Supplementary Table S1) and genes from Reactome pathway Neuronal System (R-HSA-112316) revealed that PC1 and PC2 together captured over 82% and 66% of the transcriptional variance, respectively, and illustrate a conversion trajectory along a combination of both components ( Fig. 1B; Supplementary  Fig. S1A-D).
The initial conversion trajectory (0-10 days) follows the loading vectors of basic neuronal genes such as neurofilament (NEFL) and bIII-tubulin (TUBB3), while later neuronal consolidation (10-20 days) follows loading vectors of AMPA and GABA receptor (GRIA2 and GABBR2) and NeuN (RBFOX3) genes ( Supplementary Fig. S1A). PC loadings of the Reactome pathway indicate changes toward neuronal identity for the majority of top-ranked genes during conversion ( Supplementary Fig. S1D).
Interestingly, genes of the TGFb signaling pathway (hsa04350) were almost exclusively highly enriched in FBs (Fig. 1C). Because TGFb signaling can be divided into two main arms, the TGFb1 arm and the BMP arm, we extracted the respective receptor genes, ALKs and SMADs, which are specific for either of the two arms from the time series data in a heatmap and performed linear regression analyses ( Fig. 1D-E). First, both signaling arms showed a significant repression over the time course of conversion, validating that pathway repression of both arms is indeed important for iN conversion (Fig. 1D-E). Variability between donors was assessed with mean standard deviation per time point, and indicated major changes during the first 10 days of conversion ( Supplementary Fig. S1H).
In addition, the average gene expression of BMP arm showed a steeper and more significant repression than genes specific to the TGFb1 arm (Fig. 1E), and among the BMP arm-specific components, ALK2 (ACVR1) showed the highest initial messenger RNA (mRNA) abundance in the FB stage, and the most significant drop in expression during conversion ( Fig. 1F; Supplementary Fig. S1G). These data indicate that BMP signaling through ALK2 receptor kinase signaling might be a promising signaling component where targeted inhibition might be beneficial for direct iN conversion from adult human FBs.
The ALK2-specific inhibitor dorsomorphin homolog 1 facilitates efficient iN conversion similar to Noggin on a tSMADi background In tSMADi, the compounds SB-431542 and A83-1 inhibit TGFb arm-specific ALK4, ALK5, and ALK7 receptors, and thus are not blocking signaling through the BMP arm of TGFb ( Fig. 2A). Furthermore, the tSMADi compound LDN-193189, which is similar to but more potent than dorsomorphin, unspecifically targets ALK2, ALK3, and ALK6 receptors of the BMP arm ( Fig. 2A). Because our time series transcriptome analysis suggests ALK2 inhibition to be the most promising target to enhance iN conversion, and tSMADi contains only LDN-193189 as a nonspecific ALK2 inhibitor in the absence of Noggin, we became interested in the specific ALK2 inhibitor dorsomorphin homolog 1 (DMH1) and its reported IC50s in the range of 10-100 nM (Hao et al., 2010;Mohedas et al., 2013;Neely et al., 2012).
To test the ALK2-specific inhibitor DMH1 for iN conversion, human adult FBs from two donors-CL01 (68 years, female) and CL02 (77 years, female)-were assessed. The cells were pooled into high densities in serum-containing dox-free media, and the media were changed to doxcontaining iN conversion media containing tSMADi only (further referred to as tSMADi), tSMADi plus 100 ng/mL recombinant human Noggin (+Noggin), or tSMADi plus 5 lM DMH1 (+DMH1). 5 lM DMH1 in +DMH1 media was previously confirmed as the optimal concentration with a titration curve (Supplementary Fig. S2A).
We observed that iNs formed in all three conditions, and that basal levels of toxicity did not appear to be elevated in any of the three conditions ( Fig. 2C; Supplementary  Fig. S2B). To quantitatively compare +DMH1 to +Noggin and tSMADi, the cells were fixed and analyzed at 7, 14, and 21 days of conversion and stained for the neuronal marker bIII-tubulin. All three conditions yielded iNs that displayed clear neuronal morphologies (Fig. 3A; Supplementary  Fig. S3A), and quantification revealed that at 14 and 21 days of conversion, +DMH1 and +Noggin significantly produced bIII-tubulin-positive cells, and +DMH1 showed less variability than +Noggin (Fig. 3B).
We next stained both lines with the mature neuronspecific marker NeuN ( Fig. 3C; Supplementary Fig. S3B), and found that both +DMH1 and +Noggin yielded approximately twofold more NeuN-positive iNs at 21 days than tSMADi, and only +DMH1 produced significantly more NeuN cells already at 14 days (Fig. 3D). These data suggest that +DMH1 and +Noggin are particularly important to generate fully converted cells from intermediate conversion stages. We thus asked whether +DMH1 and +Noggin can equally boost the transition of cells with intermediate transition morphologies toward fully converted iNs with mature neuronal morphologies.
We classified bIII-tubulin-positive cells into three types ( neuronal morphology), intermediate stage 2 (IM2, strong staining and no neuronal morphology), and neuron stage (iN, strong staining and neuronal morphology). Quantification confirmed that neither +DMH1 nor +Noggin had a strong and consistent effect on the formation of IM1 or IM2 cells (Fig. 4D, E), but specifically and significantly increased the number of iNs at 14 and 21 days (Fig. 4F). These data demonstrate that the small molecule DMH1 enhances the direct conversion of adult human FBs into mature iNs equally to recombinant Noggin, and that +DMH1 conversion is slightly less variable and specifically supports later stages of neuronal conversion.
Consistent and economic production of high-quality iNs from a cohort of adult human FB using +DMH1.
Comparative iN-based disease modeling studies are based on relatively large cohorts of patient FBs, and demand several rounds of efficient and reliable conversion efficiencies into iNs with mature neuronal properties. Despite the use of LDN-193189 and other ALK inhibitors, many such studies that have assessed large FB cohorts critically depend on recombinant Noggin, which, however, is variable and expensive Mertens et al., 2021;Pircs et al., 2021).
To test if +DMH1 can indeed reliably yield iNs with high efficiency, dendritic morphologies, mature synaptic-like structures, and neuronal activity properties, we assessed FBs from 15 adult human donors (Supplementary Table S4). The FBs were converted into iNs using +DMH1, fixed and stained with bIII-tubulin at 21 days, and all 15 lines yielded healthy cultures of mature iNs with many of the cells being bIII-tubulin-positive (Fig. 5A). Staining with the mature dendritic marker MAP2 revealed complex dendritic networks, and the close proximity of synapsin puncta to the postsynaptic density (PSD) marker PSD-95, seen in the three- dimensional (3D) reconstruction and two-dimensional images, suggests the presence of synapses throughout these dendritic networks (Fig. 5B-D). Quantification of the number of PSD-95 puncta in the dendritic networks of +Noggin and +DMH1-iNs showed similar densities (Fig. 5E). To assess spontaneous functional activity by live-cell somatic calcium imaging, we next lentivirally transduced the +DMH1-iNs with GCaMP6m and imaged the cells at 35 days of conversion. Importantly, and similar to previous studies based on iNs generated with +Noggin , we observed frequent and spontaneous neuronal activity in all assessed iN cultures (Fig. 5F-H). +DMH1 thus allows for the consistent generation of mature, healthy, and functional iNs from a relatively large FB donor cohort, and facilitates studies that demand the routine production of high-quality iNs.
In our conventional conversion media containing tSMA-Di and recombinant Noggin (+Noggin), Noggin is the by far the biggest cost factor accounting for 37% of the cost, followed by Laminin (18%), N2 supplement (16%), and B27 supplement (12%) (Supplementary Fig. S4). In this study, we could demonstrate that the small molecular ALK2 inhibitor DMH1 can replace recombinant Noggin at 10% of the cost, lowering conversion media costs by over 30%, and without sacrificing conversion efficiencies or iN quality (Fig. 6).
Our finding will be helpful to facilitate future cohortbased studies using iN technology, which is particularly useful for better understanding aspects of human neuronal cell aging (Huh et al., 2016;Mertens et al., 2015) and human age-related diseases such as ALS ( Jovǐcić et al., 2015), Huntington's disease Victor et al., 2018), Parkinson's disease , or Alzheimer's disease (Ma et al., 2020;Mertens et al., 2021).

Bioinformatics analysis
RNA sequencing data from Herdy et al. (2019): libraries were prepared with TruSeq stranded mRNA sample prep kit (Illumina) and sequenced single-end 50 bp on Illumina HiSeq 2500 platform. Read trimming and mapping were performed using TrimGalore and STAR, respectively. Raw counts were generated using HTseq variance stabilizing transformation normalization (vst) and differential expression analysis was performed in DEseq2 with default Wald test for hypothesis testing. Cutoff for highly significant DEGs for this study was log2-fold change with a corresponding padj <0.01 (Herdy et al., 2019). Bioinformatics and statistics analyses of RNA-Seq data from Herdy et al. (2019) were performed with R version 4.1.1 and applied methods are indicated in figure legends. GSEA to compare gene expression changes (log2FC) between FBs and D20 iN in KEGG pathways was performed with R package clusterProfiler.

Calcium imaging
Following 21 days of conversion, iNs from CL05, CL12, and CL15 were isolated with fluorescence activated cell sorting (FACS) as previously described  and replated on Geltrex-coated dishes. Media were changed to BrainPhys-based maturation media containing N2, B27, glial cell line-derived neurotrophic factor (GDNF), brain-derived neurotrophic factor (BDNF) (both 20 ng/mL, R&D), dibutyryl cyclic AMP (500 lg/mL, Sigma-Aldrich), doxycycline (2 lg/mL, Sigma-Aldrich), and laminin (1 lg/mL, Thermo Fisher Scientific). On day 25 of conversion, iNs were transduced with lentiviral particles for CAG::GcAMP6m. Calcium imaging was performed on day 35 of conversion on Zeiss Cell Observer Spinning Disc Confocal Microscope with 25 · objective, at 488 nm excitation and controlled temperature (37°C), and CO 2 concentration (5%). Time-lapse images were captured using Hamamatsu camera at 2.5 frames per second. Cell lines were assessed with three fields per cell line. Analysis was performed as previously described .