Deep Single-Cell-Type Proteome Profiling of Mouse Brain by Nonsurgical AAV-Mediated Proximity Labeling

Benchmark of Data Integration in Single-Cell Proteomics

Single-cell proteomics (SCP) detected based on different technologies always involves batch-specific variations because of differences in sample processing and other potential biases. How to integrate SCP data effectively has become a great challenge. Integration of SCP data not only requires the conservation of true biological variances, but also realizes the removal of unwanted batch effects. In this study, benchmarking analysis of popular data integration methods was conducted to determine the most suitable method for SCP data. To comprehensively evaluate the performance of these integration methods, a novel evaluation system was proposed for integrating SCP data. This evaluation system consists of three objective measures from different perspectives: category (a), the efficacy of correcting batch effects; category (b), the power of conserving biological variances; and category (c), the ability to identify consistent markers. For this comprehensive evaluation, five benchmark data sets under different scenarios (containing substantial proteins, substantial cells, multiple batches, multiple cell types, and unbalanced data) were utilized for selecting the most suitable data integration method. As a result, three methods, ComBat, Scanorama, and Seurat version 3 CCA, were identified as the most recommended methods for integrating SCP data. Overall, this systematic evaluation might provide valuable guidance in choosing the appropriate method for data integration in the SCP.

Synaptic signatures and disease vulnerabilities of layer 5 pyramidal neurons

Cortical layer 5 (L5) intratelencephalic (IT) and pyramidal tract (PT) neurons are embedded in distinct information processing pathways. Their morphology, connectivity, electrophysiological properties, and role in behavior have been extensively analyzed. However, the molecular composition of their synapses remains largely uncharacterized. Here, we dissect the protein composition of the excitatory postsynaptic compartment of mouse L5 neurons in intact somatosensory circuits, using an optimized proximity biotinylation workflow with high spatial accuracy. We find distinct synaptic signatures of L5 IT and PT neurons that are defined by proteins regulating synaptic organization and transmission, including cell-surface proteins (CSPs), neurotransmitter receptors and ion channels. In addition, we find a differential vulnerability to disease, with a marked enrichment of autism risk genes in the synaptic signature of L5 IT neurons compared to PT neurons. These results align with human studies and suggest that the excitatory postsynaptic compartment of L5 IT neurons is susceptible in autism. Our approach is versatile and can be broadly applied to other neuron types to create a protein-based, synaptic atlas of cortical circuits.

Optimized Automated Workflow for BioID Improves Reproducibility and Identification of Protein-Protein Interactions

Proximity-dependent biotinylation is an important method to study protein-protein interactions in cells, for which an expanding number of applications has been proposed. The laborious and time-consuming sample processing has limited project sizes so far. Here, we introduce an automated workflow on a liquid handler to process up to 96 samples at a time. The automation not only allows higher sample numbers to be processed in parallel but also improves reproducibility and lowers the minimal sample input. Furthermore, we combined automated sample processing with shorter liquid chromatography gradients and data-independent acquisition to increase the analysis throughput and enable reproducible protein quantitation across a large number of samples. We successfully applied this workflow to optimize the detection of proteasome substrates by proximity-dependent labeling.

Flp-recombinase mouse line for genetic manipulation of ipRGCs

Light has myriad impacts on behavior, health, and physiology. These signals originate in the retina and are relayed to the brain by more than 40 types of retinal ganglion cells (RGCs). Despite a growing appreciation for the diversity of RGCs, how these diverse channels of light information are ultimately integrated by the ~50 retinorecipient brain targets to drive these light-evoked effects is a major open question. This gap in understanding primarily stems from a lack of genetic tools that specifically label, manipulate, or ablate specific RGC types. Here, we report the generation and characterization of a new mouse line (Opn4FlpO), in which FlpO is expressed from the Opn4 locus, to manipulate the melanopsin-expressing, intrinsically photosensitive retinal ganglion cells. We find that the Opn4FlpO line, when crossed to multiple reporters, drives expression that is confined to ipRGCs and primarily labels the M1-M3 subtypes. Labeled cells in this mouse line show the expected intrinsic, melanopsin-based light response and morphological features consistent with the M1-M3 subtypes. In alignment with the morphological and physiological findings, we see strong innervation of non-image forming brain targets by ipRGC axons, and weaker innervation of image forming targets in Opn4FlpO mice labeled using AAV-based and FlpO-reporter lines. Consistent with the FlpO insertion disrupting the endogenous Opn4 transcript, we find that Opn4FlpO/FlpO mice show deficits in the pupillary light reflex, demonstrating their utility for behavioral research in future experiments. Overall, the Opn4FlpO mouse line drives Flp-recombinase expression that is confined to ipRGCs and most effectively drives recombination in M1-M3 ipRGCs. This mouse line will be of broad use to those interested in manipulating ipRGCs through a Flp-based recombinase for intersectional studies or in combination with other, non-Opn4 Cre driver lines.

Synaptic proteomics decode novel molecular landscape in the brain

Synapses play a pivotal role in forming neural circuits, with critical implications for brain functions such as learning, memory, and emotions. Several advances in synaptic research have demonstrated the diversity of synaptic structure and function, which can form thousands of connections depending on the neuronal cell types. Moreover, synapses not only interconnect neurons but also establish connections with glial cells such as astrocytes, which play a key role in the architecture and function of neuronal circuits in the brain. Emerging evidence suggests that dysfunction of synaptic proteins contributes to a variety of neurological and psychiatric disorders. Therefore, it is crucial to determine the molecular networks within synapses in various neuronal cell types to gain a deeper understanding of how the nervous system regulates brain function. Recent advances in synaptic proteome approaches, such as fluorescence-activated synaptosome sorting (FASS) and proximity labeling, have allowed for a detailed and spatial analysis of many cell-type-specific synaptic molecules in vivo. In this brief review, we highlight these novel spatial proteomic approaches and discuss the regulation of synaptic formation and function in the brain. This knowledge of molecular networks provides new insight into the understanding of many neurological and psychiatric disorders.

Multimodal joint deconvolution and integrative signature selection in proteomics

Deconvolution is an efficient approach for detecting cell-type-specific (cs) transcriptomic signals without cellular segmentation. However, this type of methods may require a reference profile from the same molecular source and tissue type. Here, we present a method to dissect bulk proteome by leveraging tissue-matched transcriptome and proteome without using a proteomics reference panel. Our method also selects the proteins contributing to the cellular heterogeneity shared between bulk transcriptome and proteome. The deconvoluted result enables downstream analyses such as cs-protein Quantitative Trait Loci (cspQTL) mapping. We benchmarked the performance of this multimodal deconvolution approach through CITE-seq pseudo bulk data, a simulation study, and the bulk multi-omics data from human brain normal tissues and breast cancer tumors, individually, showing robust and accurate cell abundance quantification across different datasets. This algorithm is implemented in a tool MICSQTL that also provides cspQTL and multi-omics integrative visualization, available at https://bioconductor.org/packages/MICSQTL .

In vivo identification of astrocyte and neuron subproteomes by proximity-dependent biotinylation

The central nervous system (CNS) comprises diverse and morphologically complex cells. To understand the molecular basis of their physiology, it is crucial to assess proteins expressed within intact cells. Commonly used methods utilize cell dissociation and sorting to isolate specific cell types such as neurons and astrocytes, the major CNS cells. Proteins purified from isolated cells are identified by mass spectrometry-based proteomics. However, dissociation and cell-sorting methods lead to near total loss of cellular morphology, thereby losing proteins from key relevant subcompartments such as processes, end feet, dendrites and axons. Here we provide a systematic protocol for cell- and subcompartment-specific labeling and identification of proteins found within intact astrocytes and neurons in vivo. This protocol utilizes the proximity-dependent biotinylation system BioID2, selectively expressed in either astrocytes or neurons, to label proximal proteins in a cell-specific manner. BioID2 is targeted genetically to assess the subproteomes of subcellular compartments such as the plasma membrane and sites of cell-cell contacts. We describe in detail the expression methods (variable timing), stereotaxic surgeries for expression (1-2 d and then 3 weeks), in vivo protein labeling (7 d), protein isolation (2-3 d), protein identification methods (2-3 d) and data analysis (1 week). The protocol can be applied to any area of the CNS in mouse models of physiological processes and for disease-related research.

Profiling Mouse Brain Single-Cell-Type Proteomes Via Adeno-Associated Virus-Mediated Proximity Labeling and Mass Spectrometry

Single-cell-type proteomics is an emerging field of research that combines cell-type specificity with the comprehensive proteome coverage offered by bulk proteomics. However, the extraction of single-cell-type proteomes remains a challenge, particularly for hard-to-isolate cells like neurons. In this chapter, we present an innovative technique for profiling single-cell-type proteomes using adeno-associated virus (AAV)-mediated proximity labeling (PL) and tandem-mass-tag (TMT) mass spectrometry. This technique eliminates the need for cell isolation and offers a streamlined workflow, including AAV delivery to express TurboID (an engineered biotin ligase) controlled by cell-type-specific promoters, biotinylated protein purification, on-bead digestion, TMT labeling, and liquid chromatography-mass spectrometry (LC-MS). We examined this method by analyzing distinct brain cell types in mice. Initially, recombinant AAVs were used to concurrently express TurboID and mCherry proteins driven by neuron- or astrocyte-specific promoters, which was validated through co-immunostaining with cellular markers. With biotin purification and TMT analysis, we successfully identified around 10,000 unique proteins from a few micrograms of protein samples with high reproducibility. Our statistical analyses revealed that these proteomes encompass cell-type-specific cellular pathways. By utilizing this technique, researchers can explore the proteomic landscape of specific cell types, paving the way for new insights into cellular processes, deciphering disease mechanisms, and identifying therapeutic targets in neuroscience and beyond.

Mapping the human brain proteome: opportunities, challenges, and clinical potential

INTRODUCTION

Due to the segmented functions and complexity of the human brain, the characterization of molecular profiles within specific areas such as brain structures and biofluids is essential to unveil the molecular basis for structure specialization as well as the molecular imbalance associated with neurodegenerative and psychiatric diseases.

AREAS COVERED

Much of our knowledge about brain functionality derives from neurophysiological, anatomical, and transcriptomic approaches. More recently, laser capture and imaging proteomics, technological and computational developments in LC-MS/MS, as well as antibody/aptamer-based platforms have allowed the generation of novel cellular, spatial, and posttranslational dimensions as well as innovative facets in biomarker validation and druggable target identification.

EXPERT OPINION

Proteomics is a powerful toolbox to functionally characterize, quantify, and localize the extensive protein catalog of the human brain across physiological and pathological states. Brain function depends on multi-dimensional protein homeostasis, and its elucidation will help us to characterize biological pathways that are essential to properly maintain cognitive functions. In addition, comprehensive human brain pathological proteomes may be the basis in computational drug-repositioning methods as a strategy for unveiling potential new therapies in neurodegenerative and psychiatric disorders.

Cooperative lipolytic control of neuronal triacylglycerol by spastic paraplegia-associated enzyme DDHD2 and ATGL

Intracellular lipolysis-the enzymatic breakdown of lipid droplet-associated triacylglycerol (TAG)-depends on the cooperative action of several hydrolytic enzymes and regulatory proteins, together designated as lipolysome. Adipose triglyceride lipase (ATGL) acts as a major cellular TAG hydrolase and core effector of the lipolysome in many peripheral tissues. Neurons initiate lipolysis independently of ATGL via DDHD domain-containing 2 (DDHD2), a multifunctional lipid hydrolase whose dysfunction causes neuronal TAG deposition and hereditary spastic paraplegia. Whether and how DDHD2 cooperates with other lipolytic enzymes is currently unknown. In this study, we further investigated the enzymatic properties and functions of DDHD2 in neuroblastoma cells and primary neurons. We found that DDHD2 hydrolyzes multiple acylglycerols in vitro and substantially contributes to neutral lipid hydrolase activities of neuroblastoma cells and brain tissue. Substrate promiscuity of DDHD2 allowed its engagement at different steps of the lipolytic cascade: In neuroblastoma cells, DDHD2 functioned exclusively downstream of ATGL in the hydrolysis of sn-1,3-diacylglycerol (DAG) isomers but was dispensable for TAG hydrolysis and lipid droplet homeostasis. In primary cortical neurons, DDHD2 exhibited lipolytic control over both, DAG and TAG, and complemented ATGL-dependent TAG hydrolysis. We conclude that neuronal cells use noncanonical configurations of the lipolysome and engage DDHD2 as dual TAG/DAG hydrolase in cooperation with ATGL.

The development of proximity labeling technology and its applications in mammals, plants, and microorganisms

Protein‒protein, protein‒RNA, and protein‒DNA interaction networks form the basis of cellular regulation and signal transduction, making it crucial to explore these interaction networks to understand complex biological processes. Traditional methods such as affinity purification and yeast two-hybrid assays have been shown to have limitations, as they can only isolate high-affinity molecular interactions under nonphysiological conditions or in vitro. Moreover, these methods have shortcomings for organelle isolation and protein subcellular localization. To address these issues, proximity labeling techniques have been developed. This technology not only overcomes the limitations of traditional methods but also offers unique advantages in studying protein spatial characteristics and molecular interactions within living cells. Currently, this technique not only is indispensable in research on mammalian nucleoprotein interactions but also provides a reliable approach for studying nonmammalian cells, such as plants, parasites and viruses. Given these advantages, this article provides a detailed introduction to the principles of proximity labeling techniques and the development of labeling enzymes. The focus is on summarizing the recent applications of TurboID and miniTurbo in mammals, plants, and microorganisms. Video Abstract.

Mass spectrometry-based proteomic characterization of the middle-aged mouse brain for animal model research of neuromuscular diseases

Neuromuscular diseases with primary muscle wasting symptoms may also display multi-systemic changes in the body and exhibit secondary pathophysiological alterations in various non-muscle tissues. In some cases, this includes proteome-wide alterations and/or adaptations in the central nervous system. Thus, in order to provide an improved bioanalytical basis for the comprehensive evaluation of animal models that are routinely used in muscle research, this report describes the mass spectrometry-based proteomic characterization of the mouse brain. Crude tissue extracts were examined by bottom-up proteomics and detected 4558 distinct protein species. The detailed analysis of the brain proteome revealed the presence of abundant cellular proteoforms in the neuronal cytoskeleton, as well as various brain region enriched proteins, including markers of the cerebral cortex, cerebellum, hippocampus and the olfactory bulb. Neuroproteomic markers of specific cell types in the brain were identified in association with various types of neurons and glia cells. Markers of subcellular structures were established for the plasmalemma, nucleus, endoplasmic reticulum, mitochondria and other crucial organelles, as well as synaptic components that are involved in presynaptic vesicle docking, neurotransmitter release and synapse remodelling.

Chemical and Biological Strategies for Profiling Protein-Protein Interactions in Living Cells

Protein-protein interactions (PPIs) play critical roles in almost all cellular signal transduction events. Characterization of PPIs without interfering with the functions of intact cells is very important for basic biology study and drug developments. However, the ability to profile PPIs especially those weak/transient interactions in their native states remains quite challenging. To this end, many endeavors are being made in developing new methods with high efficiency and strong operability. By coupling with advanced fluorescent microscopy and mass spectroscopy techniques, these strategies not only allow us to visualize the subcellular locations and monitor the functions of protein of interest (POI) in real time, but also enable the profiling and identification of potential unknown interacting partners in high-throughput manner, which greatly facilitates the elucidation of molecular mechanisms underlying numerous pathophysiological processes. In this review, we will summarize the typical methods for PPIs identification in living cells and their principles, advantages and limitations will also be discussed in detail.

Proximity Labeling in Plants

Proteins are workhorses in the cell; they form stable and more often dynamic, transient protein-protein interactions, assemblies, and networks and have an intimate interplay with DNA and RNA. These network interactions underlie fundamental biological processes and play essential roles in cellular function. The proximity-dependent biotinylation labeling approach combined with mass spectrometry (PL-MS) has recently emerged as a powerful technique to dissect the complex cellular network at the molecular level. In PL-MS, by fusing a genetically encoded proximity-labeling (PL) enzyme to a protein or a localization signal peptide, the enzyme is targeted to a protein complex of interest or to an organelle, allowing labeling of proximity proteins within a zoom radius. These biotinylated proteins can then be captured by streptavidin beads and identified and quantified by mass spectrometry. Recently engineered PL enzymes such as TurboID have a much-improved enzymatic activity, enabling spatiotemporal mapping with a dramatically increased signal-to-noise ratio. PL-MS has revolutionized the way we perform proteomics by overcoming several hurdles imposed by traditional technology, such as biochemical fractionation and affinity purification mass spectrometry. In this review, we focus on biotin ligase-based PL-MS applications that have been, or are likely to be, adopted by the plant field. We discuss the experimental designs and review the different choices for engineered biotin ligases, enrichment, and quantification strategies. Lastly, we review the validation and discuss future perspectives.

Cell-selective proteomics segregates pancreatic cancer subtypes by extracellular proteins in tumors and circulation

Cell-selective proteomics is a powerful emerging concept to study heterocellular processes in tissues. However, its high potential to identify non-cell-autonomous disease mechanisms and biomarkers has been hindered by low proteome coverage. Here, we address this limitation and devise a comprehensive azidonorleucine labeling, click chemistry enrichment, and mass spectrometry-based proteomics and secretomics strategy to dissect aberrant signals in pancreatic ductal adenocarcinoma (PDAC). Our in-depth co-culture and in vivo analyses cover more than 10,000 cancer cell-derived proteins and reveal systematic differences between molecular PDAC subtypes. Secreted proteins, such as chemokines and EMT-promoting matrisome proteins, associated with distinct macrophage polarization and tumor stromal composition, differentiate classical and mesenchymal PDAC. Intriguingly, more than 1,600 cancer cell-derived proteins including cytokines and pre-metastatic niche formation-associated factors in mouse serum reflect tumor activity in circulation. Our findings highlight how cell-selective proteomics can accelerate the discovery of diagnostic markers and therapeutic targets in cancer.

Proteomics-based synapse characterization: From proteins to circuits

The highly heterogeneous nature of neuronal cell types and their connections presents a major challenge to the characterization of neural circuits at the protein level. New approaches now enable an increasingly sophisticated dissection of cell type- and cellular compartment-specific proteomes, as well as the profiling of the protein composition of specific synaptic connections. Here, we provide an overview of these approaches and discuss how they hold considerable promise toward unravelling the molecular mechanisms of neural circuit formation and function. Finally, we provide an outlook of technological developments that may bring the characterization of synaptic proteomes at the single-synapse level within reach.

Recent advances in microfluidics for single-cell functional proteomics

Single-cell proteomics (SCP) reveals phenotypic heterogeneity by profiling individual cells, their biological states and functional outcomes upon signaling activation that can hardly be probed via other omics characterizations. This has become appealing to researchers as it enables an overall more holistic view of biological details underlying cellular processes, disease onset and progression, as well as facilitates unique biomarker identification from individual cells. Microfluidic-based strategies have become methods of choice for single-cell analysis because they allow facile assay integrations, such as cell sorting, manipulation, and content analysis. Notably, they have been serving as an enabling technology to improve the sensitivity, robustness, and reproducibility of recently developed SCP methods. Critical roles of microfluidics technologies are expected to further expand rapidly in advancing the next phase of SCP analysis to reveal more biological and clinical insights. In this review, we will capture the excitement of the recent achievements of microfluidics methods for both targeted and global SCP, including efforts to enhance the proteomic coverage, minimize sample loss, and increase multiplexity and throughput. Furthermore, we will discuss the advantages, challenges, applications, and future prospects of SCP.

LC-MS/MS-Based Proteomics Approach for the Identification of Candidate Serum Biomarkers in Patients with Narcolepsy Type 1

Narcolepsy type 1 (NT1) is the most common type of narcolepsy known to be caused by the loss of specific neurons responsible for producing peptide neurotransmitters (orexins/hypocretins), resulting in a sleep-wake cycle disorder. It is characterized by its association with cataplexy and abnormalities in rapid eye movement. To date, no cure has been established for this life-threatening condition. Misdiagnosis of NT1 is also quite common, although it is not exceedingly rare. Therefore, successfully identifying candidate serum biomarkers for NT1 would be a head start for accurate diagnosis and development of therapeutics for this disorder. This study aims to identify such potential serum biomarkers. A depletion protocol was employed for 27 human serum samples (16 NT1 and 11 healthy controls), followed by applying LC-MS/MS bottom-up proteomics analysis, then LC-PRM-MS for validation. The comparison of the proteome profiles of the low-abundant proteins in the samples was then investigated based on age, sex, sample groups, and the presence of the Human Leukocyte Antigen (HLA) DQB1*0602 allele. The results were tracked to gene expression studies as well as system biology to identify key proteins and understand their relationship in the pathogenesis of NT1. Our results revealed 36 proteins significantly and differentially expressed. Among the impaired pathways and bioprocesses, the complement activation pathway is impaired by six of the differentially expressed proteins (DEPs). They are coded by the genes C2, CFB, C5, C1R, C1S, and MASP1, while 11 DEPs are involved in Acute Phase Response Signaling (APRS), which are coded by the genes FN1, AMBP, APOH, CFB, CP, ITIH2, C5, C2, F2, C1, and ITIH4. The combined AUCs of the downregulated and upregulated DEPs are 0.95 and 0.76, respectively. Overall, this study reveals potential serum-protein biomarkers of NT1 and explains the possible correlation between the biomarkers and pathophysiological effects, as well as important biochemical pathways involved in NT1.

An integrative machine-learning meta-analysis of high-throughput omics data identifies age-specific hallmarks of Alzheimer's disease

Alzheimer's disease (AD) is an incredibly complex and presently incurable age-related brain disorder. To better understand this debilitating disease, we collated and performed a meta-analysis on publicly available RNA-Seq, microarray, proteomics, and microRNA samples derived from AD patients and non-AD controls. 4089 samples originating from brain tissues and blood remained after applying quality filters. Since disease progression in AD correlates with age, we stratified this large dataset into three different age groups: < 75 years, 75-84 years, and ≥ 85 years. The RNA-Seq, microarray, and proteomics datasets were then combined into different integrated datasets. Ensemble machine learning was employed to identify genes and proteins that can accurately classify samples as either AD or control. These predictive inputs were then subjected to network-based enrichment analyses. The ability of genes/proteins associated with different pathways in the Molecular Signatures Database to diagnose AD was also tested. We separately identified microRNAs that can be used to make an AD diagnosis and subjected the predicted gene targets of the most predictive microRNAs to an enrichment analysis. The following key themes emerged from our machine learning and bioinformatics analyses: cell death, cellular senescence, energy metabolism, genomic integrity, glia, immune system, metal ion homeostasis, oxidative stress, proteostasis, and synaptic function. Many of the results demonstrated unique age-specificity. For example, terms highlighting cellular senescence only emerged in the earliest and intermediate age ranges while the majority of results relevant to cell death appeared in the youngest patients. Existing literature corroborates the importance of these hallmarks in AD.

Dynamic proteomic and phosphoproteomic atlas of corticostriatal axons in neurodevelopment

Mammalian axonal development begins in embryonic stages and continues postnatally. After birth, axonal proteomic landscape changes rapidly, coordinated by transcription, protein turnover, and post-translational modifications. Comprehensive profiling of axonal proteomes across neurodevelopment is limited, with most studies lacking cell-type and neural circuit specificity, resulting in substantial information loss. We create a Cre-dependent APEX2 reporter mouse line and map cell-type-specific proteome of corticostriatal projections across postnatal development. We synthesize analysis frameworks to define temporal patterns of axonal proteome and phosphoproteome, identifying co-regulated proteins and phosphorylations associated with genetic risk for human brain disorders. We discover proline-directed kinases as major developmental regulators. APEX2 transgenic reporter proximity labeling offers flexible strategies for subcellular proteomics with cell type specificity in early neurodevelopment, a critical period for neuropsychiatric disease.

Potential application of TurboID-based proximity labeling in studying the protein interaction network in plant response to abiotic stress

Abiotic stresses are major environmental conditions that reduce plant growth, productivity and quality. Protein-protein interaction (PPI) approaches can be used to screen stress-responsive proteins and reveal the mechanisms of protein response to various abiotic stresses. Biotin-based proximity labeling (PL) is a recently developed technique to label proximal proteins of a target protein. TurboID, a biotin ligase produced by directed evolution, has the advantages of non-toxicity, time-saving and high catalytic efficiency compared to other classic protein-labeling enzymes. TurboID-based PL has been successfully applied in animal, microorganism and plant systems, particularly to screen transient or weak protein interactions, and detect spatially or temporally restricted local proteomes in living cells. This review concludes classic PPI approaches in plant response to abiotic stresses and their limitations for identifying complex network of regulatory proteins of plant abiotic stresses, and introduces the working mechanism of TurboID-based PL, as well as its feasibility and advantages in plant abiotic stress research. We hope the information summarized in this article can serve as technical references for further understanding the regulation of plant adaptation to abiotic stress at the protein level.